Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Active Disturbance Rejection Control Design with Suppression of Sensor Noise Effects in Application to DC-DC Buck Power Converter

Active Disturbance Rejection Control Design with Suppression of Sensor Noise Effects in... This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS Active Disturbance Rejection Control Design with Suppression of Sensor Noise Effects in Application to DC-DC Buck Power Converter Krzysztof Łakomy, Rafal Madonski, Bin Dai, Jun Yang, Senior Member, IEEE, Piotr Kicki, Maral Ansari, Student Member, IEEE, Shihua Li, Fellow, IEEE Abstract —The performance of active disturbance rejec- strongly amplified measurement noise into the control signal tion control (ADRC) algorithms can be limited in practice calculated upon the state vector of ESO. This may cause by high-frequency measurement noise. In this work, this the decrease of control quality (e.g. when the noise-affected problem is addressed by transforming the high-gain ex- control signal hits the actuator saturation), higher energy tended state observer (ESO), which is the inherent element consumption, and quicker wear of the equipment. The HGO- of ADRC, into a new cascade observer structure. Set of experiments, performed on a DC-DC buck power converter based ADRC design and tuning often come down to a forced system, show that the new cascade ESO design, com- compromise between speed/accuracy of signals reconstruction pared to the conventional approach, effectively suppresses and sensitivity to noise [9]. Same compromise can be seen in the detrimental effect of sensor noise over-amplification the ADRC works for buck converters in which the measured while increasing the estimation/control performance. The system output (voltage) is oftentimes corrupted with high- proposed design is also analyzed with a low-pass filter at the converter output, which is a common technique for frequency noise [10]. Several types of solutions were proposed reducing measurement noise in industrial applications. to solve the problem of attenuating the effects of measurement noise in high-gain observers. They mainly address it by: Index Terms—noise suppression, power converter, high- gain observer, extended state observer, ESO employing nonlinear [4], [11] or adaptive techniques [12], redesigning the local behavior by combining different types of observers [13], using low-power structures [14]–[16], or I. INTRODUCTION modifying standard low-pass filters [17]. ENEWABLE energy sources, like fuel and photovoltaic Motivated by the above problem, a new cascade ESO-based cells, are rapidly evolving technologies for DC voltage error-domain ADRC solution is presented. Following the gen- generation, which results in proliferation of DC–DC buck eral idea shown in [18], we propose a virtual decomposition converters in power applications. Practically appealing results of the total disturbance present in the DC-DC buck converter on buck converter control using the idea of active disturbance system, allowing to design a cascade structure of ESO, where rejection control (ADRC) were recently reported in [1]–[3]. each level of the observer cascade is responsible for handling The key element in any ADRC scheme is the extended state a particular type and frequency range of estimated signal. The observer (ESO [4]), responsible for estimating the system state proposed topology enhances conventional state/disturbance vector and reconstructing the overall disturbance (also referred estimation performance while avoiding over-amplification of to as total disturbance) affecting the controlled variable [5]. the sensor noise. The user-defined number of cascade levels However, since the conventional form of ADRC uses a high- allows to customize the overall control system structure to gain observer (HGO) structure to estimate selected signals, meet certain disturbance rejection requirements. Although a its capabilities are intrinsically limited by the presence and multi-level cascade observer is proposed, a straightforward severity of high-frequency sensor noise, as discussed in [6]– design and implementation methodology is given, together [8]. The high gains of the observer cause the transfer of with intuitive tuning rules. The novelty of this work includes Manuscript received Month xx, 2xxx; revised Month xx, xxxx; ac- an experimental validation of the proposed cascade ESO- cepted Month x, xxxx. based ADRC structure, a proof of the input-to-state stability K. Łakomy is with the Poznan ´ University of Technology, Poznan, ´ 60- of the closed-loop system, and additional insights about the 965, Poland (e-mail: krzysztof.lakomy92@gmail.com). R. Madonski (corresponding author) is with the Energy Electricity Re- sensor noise suppressing effects in frequency domain. The search Center, International Energy College, Jinan University, Zhuhai, experimental study also addresses the impact of a low-pass 519070, P. R. China (e-mail: rafal.madonski@jnu.edu.cn). filter implemented at the converter output, which is a popular B. Dai, J. Yang and S. Li are with the School of Automation, Southeast University, Key Laboratory of Measurement and Control of approach for handling high-frequency sensor noise [19]. CSE, Ministry of Education, Nanjing, 210096, P. R. China (e-mail: Notation. Within this article, we treat R as a set of real fbin 1994/j.yang84/lshg@seu.edu.cn). numbers, R = {x ∈ R ∶ x > 0} as a set of positive real P. Kicki is with the Institute of Robotics and Machine Intelligence, ´ ´ Poznan University of Technology, Poznan, 60-965, Poland (e-mail: pi- numbers, R = {x ∈ R ∶ x > 0} as a set of non-negative real ≥0 otr.kicki@put.poznan.pl). numbers, Z as a set of integers,  (A A A) and  (A A A) are min max M. Ansari is with the Faculty of Engineering and Information Tech- respectively the minimal and maximal eigenvalues of matrix nology, University of Technology Sydney, Sydney NSW, 2007, Australia (e-mail: maral.ansari@student.uts.edu.au). A A A, while A A A ≻ 0 means that matrix A A A is positive definite. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. arXiv:2009.02948v2 [eess.SY] 4 Feb 2021 This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS Function f (x) ∶ R → R belongs to class K when it is strictly increasing and f (0) = 0. The expression ls ∶= lim sup is t→∞ used for the sake of notation compactness. II. PRELIMINARIES A. Simplified plant model and control objective Following [3], an average dynamic model of a DC-DC buck Fig. 1: Semiconductor realization of the considered DC-DC converter, depicted in Fig. 1, can be written as buck power converter, with diode V D and control switch V T . dv (t) ⎧ o 1 1 ⎪ = i (t) − v (t); L o ⎪ dt C CR di (t) L V 1 in ⎨ = [(t) + d(t)] − v (t); (1) o Combining the uncertain (or unknown) terms in (2), includ- dt L L ⎪ ing the imperfect identification of the input gain, results in a y (t) = v (t) + n(t); o o following form of the output voltage dynamics where  ∈ [0; 1] is the duty ratio, y [V] is the measured system ^ ^ ^ v  = a v + a v _ + b − b + bd+b = F (⋅) + b; (3) output that consists of the average capacitor voltage v [V] and o 2 o 1 o ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ the sensor noise n[V], i [A] is the average inductor current, F(t;v_ ;v ;;d) o o R[ ] is the load resistance of the circuit, L[H] is the filter where b ≠ 0 is a precise-enough estimate of the input gain b inductance, C [F] is the filter capacitance, V [V] is the input in from (2) and F (⋅) represents the total disturbance of (3). voltage source, and d(t) represents the unknown (possibly Since v (t) and its derivatives may not be known a priori, time-varying and nonlinear) external disturbance. which may lead to possible inability of constructing feedfor- The considered control objective is to force v (t) to follow ward signal in , let us reformulate (3) in error-domain a reference capacitor output voltage trajectory v (t)[V] by manipulating (t) with following assumptions applying. e  = v  − v  = v  − F (⋅)−b; (4) r o r Assumption 1: Following the limitations resulting from the ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ F (⋅;v ) physical properties of the considered electronic circuit, we may assume that the values of voltage and current are bounded, and where e(t) ≜ v (t)−v (t) is the control error signal and F (⋅) r o belong to some compact set such that sup Si (t)S < r and L i is the total disturbance in the error-domain [21]. In this article, t≥0 L sup Sv (t)S < r for r ;r > 0. o v i v we utilize a standard form of the ADRC controller t≥0 o L o Assumption 2: Output voltage v (t) is the only measur- −1 ∗ = b (F +  ); (5) able signal and is additionally corrupted by bounded, high- frequency measurement noise sup Sn(t)S < r for r > 0. n n which is constructed to simultaneously compensate the in- t≥0 Assumption 3: [20] The unknown external disturbance d(t) fluence of disturbance using the estimated value of total may have a countable number of first-class discontinuity disturbance (F ) and to stabilize system (4) in a close vicinity points at times t = T for i ∈ {1;:::;N }, N ∈ Z, 0 ≤ N < ∞, i d d d of the equilibrium point e = 0 using the output-feedback and 0 < inf (T −T ) < ∞ for N > 1. In all other i+1 i d stabilizing controller  . i∈{1;:::;N −1} 0 moments, the external disturbance function is bounded and has Assumption 5: Stabilizing controller  has a structure that bounded first time derivative, i.e., sup Sd(t)S < r and d guarantees the boundedness of  (⋅) and  _ (⋅). Although t≥0;t∈~{T } 0 0 sup Sd(t)S < r for some r ;r > 0 and i ∈ {1;:::;N }. this assumption may seem conservative, it is relaxed with the _ _ d d t≥0;t∈~{T } d d Assumption 4: The reference signal v (t) may have a previously introduced Assumptions 1, 3, and 4. Remark 1: Since the disturbance F and the control variable countable number of first-class discontinuity points at times have equal relative rank, with respect to the voltage v t = T for i ∈ {1;:::;N }, N ∈ Z, 0 ≤ N < ∞, and 0 < i r r r o representing the output of the original system (see (1)), the inf (T − T ) < ∞ for N > 1. There also exists i+1 i r i∈{1;:::;N −1} total disturbances affecting the error-domain system, described a positive constant r , such that v (t) and its specific time- v r (j) with (4), meet the so-called matching condition. The specific derivatives satisfy inequality sup šUv (t)UŸ ≤ r , r v t≥0;t∈~{T } r differences and control solutions for matched and mismatched for i ∈ {1;:::;N } and j ∈ {0; 1; 2; 3}. disturbances have been thoroughly discussed in [7]. We will first put the focus on precise and on-line estimation B. Application of the ADRC principle of perturbing term F (⋅), crucial for proper active disturbance Following the standard ADRC design, system model (1) is rejection. To calculate F , we first need to define the extended ⊺ ∗ ⊺ 3 reformulated, emphasizing its input-output relation state z = [z z z ] ≜ [e e _ F ] ∈ D , where D ≜ {xxx ∈ R ∶ 1 2 3 z z 2 YxY < r } for some r ∈ R . The dynamics of the state vector z z + d v (t) 1 dv (t) 1 V o o in = − − v (t) + [(t) + d(t)]: o z can be expressed, upon (4), as a state-space model dt CR dt CL CL ± ⎧ ² ² ^ _ z _ = A A Az −dddb +bbbF ; a a b ⎪ 1 2 ⎨ (6) (2) y = e − n = ccc z − n; Function f(x) ∶ R → R has first-class discontinuity at point x  if for 2×1 2×2 ⊺ ⊺ 000 III + − + − where A A A ≜  , ddd ≜ [0 1 0] , ccc ≜ [1 0 0] , and bbb ≜ 1×2 f ∶= lim + f(x) and f ∶= lim − f(x), it satisfies f ≠ f and x→x  x→x  0 000 + − ⊺ max{f ;f } ≤ r for some r > 0. [0 0 1] . Given (6), the output of this system y corresponds f f ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS to the control error e which, according to Assumption 2, is DC-DC (design parameter) influenced by the measurement noise n. converter (1) Remark 2: Control error e, together with its derivative e _ are p-th level - v o ... n v cascade ESO (7) - r bounded according to the Assumptions 1, 3, and 4, and the specific form of the system dynamics (1). 1 EE SS O O Remark 3: Under the Assumptions 1, 3 and 4, func- 1s1 t l seve t level l () ... tion F (t) is continuously differentiable, and thus, there μ μ Controller ESO exist bounded continuous functions ; such that 2 F _ ∗ . (5) 2nd level ∗ ∗ sup SF (t)S < (e;e; _ v ;v _ ;v  ;), sup SF (t)S < F r r r t≥0 t≥0 ... ⊺ 2 (e;e; _ v ;v _ ;v  ; v ;; _ ), for all [e e _] ∈ R . Both prac- d _ ∗ r r r r : tical and theoretical justifications of lumping selected com- ponents as parts of F (⋅), including control signal and state- z ξ ESO dependent variables, has been thoroughly discussed in [5]. p-th level III. MAIN RESULT: PROPOSED CASCADE ESO ADRC State selector (8) To calculate the estimated value of extended state vector z, Fig. 2: Proposed ADRC with sensor noise suppression via let us now introduce a novel p-level structure of a cascade cascade ESO structure for the DC-DC buck power converter. observer (p ∈ Z and p ≥ 2) in a following form _ ^ (t) = A A A (t) −dddb(t) +lll y(t) −ccc  (t) 1 1 1 i−1 and its inclusion in the overall estimate of the extended state ⎛ ⎞ (t) = A A A (t) +ddd −b(t) +bbb  (t) i j vector (8). The following observer levels are using the state ⎝ ⎠ j=1 vectors of previous observer levels instead of the measured +lll ccc [ (t) − (t)]; i ∈ {2;:::;p}; (7) i i−1 i signal, and thus, result in lower noise amplification than the single-level ESO with high bandwidth. Important part in the ⊺ 3 where  ≜ [   ] ∈ R is the state of a particular j j;1 j;2 j;3 utilized cascade observer structure is the state selector (8), 2 3 ⊺ 3 observer cascade level, lll ≜ [3! 3! ! ] ∈ R is the j oj oj oj which defines which estimated state variables (and from which j−1 observer gain vector with design parameter ! ≜ ! ∈ oj o1 observer level) participate in the controller synthesis (5) to R for > 1; ! ∈ R , and j ∈ {1;:::;p}. The estimate of + o1 + provide improved sensor noise effect suppression. z, resulting from the observer (7) can be expressed as Having z ^, the application of control action (5) to the system p−1 (4) results in a following second-order error dynamics ⊺ ⊺ 3 ^  bb z = [z ^ z ^ z ^ ] ≜  +bbbb Q  ∈ R : (8) 1 2 3 p j j=1 ∗ e  = F −  ; (9) Remark 4: It is worth noting, that if we reduce the observer ∗ ∗ ∗ ~ ^ where F ≜ F −F is the final residue of the total disturbance to a single level (p = 1), we would obtain a standard form resulting from the imperfect observation of F by observer (7). of a linear high-gain ESO, as seen in [22]. An introduction A block diagram of the proposed ADRC with cascade ESO of the subsequent cascade levels allows us to keep the same for the DC-DC buck power converter is shown in Fig. 2. observation quality with smaller values of ! , resulting in o1 Theorem 1: Under Assumptions 3-5, and by taking a stabi- a decrease of the measurement noise amplification visible in lizing proportional-derivative controller the state estimates, see (7). This effect will be depicted in the upcoming experiments. ≜ k y + k z ^ ; k ;k > 0; (10) 0 p d 2 p d The idea of cascade observer structure, proposed in (7) and illustrated in Fig. 2, is based on a specific choice of the observation errors of the extended state obtained with the the first level observer bandwidth ! , which should be large o1 p-level cascade observer, defined as enough to guarantee precise estimation of the first element p−1 of extended state vector z, and low enough to make the ⊺ ⊺ 3 z ~ = [z ~ z ~ z ~ ] ≜ z − z ^; = z − −bbbbbb  ∈ R ; (11) p p1 p2 p3 p j first level of the cascade to act as a low-pass filter for the j=1 noise. Latter elements of the extended state vector, i.e. z and together with the control error e, described with the dynamics z , usually have faster transients, and thus, are not estimated (9), are bounded. In other words precisely with the first level observer with a low ! value. o1 The consecutive observer levels are introduced to improve the ∀ ∀ ∃ ls Yz ~ (t)Y <  ∧ ls Se(t)S <  ; (12) t>t ! ;k>0  ; >0 ∞ p z~ ∞ e 0 o1 z~ e estimation quality of z and z using higher observer band- 2 3 where t = max{T ;T } results from Assumptions 3 and 4. widths ! (i > 1) and improve the observation performance by oi 0 N N d r incrementally extending the range of precisely estimated signal Remark 5: To keep the notational conciseness of the fol- frequencies. The introduction of additional cascade levels of lowing theoretical analysis and to reduce the overall number the observer can be interpreted as an attempt to estimate the of tuning parameters, we propose, following [22], to tune the total disturbance residue, that could not be precisely estimated stabilizing controller (10) with a single parameter k > 0, setting with the previous cascade levels due to limited bandwidth, the values of proportional and derivative gains, respectively, as ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. p = 1 p = 2 p This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS k = k and k = 2k. Chosen tuning procedure places the poles write down that Yz ~ Y ≤ ZZ ≤  (   ) YY and thus that the p d p max of control error dynamics (9) at value −k. asymptotic relation Proof of Theorem 1. The dynamics of the observation error ls Yz ~ (t)Y ≤  (   )ls Y(t)Y =∶  ; (19) ∞ p max  ∞ z~ defined for a particular cascade level, i.e. z ~ ≜ z −  − i i ⊺ i−1 3 bbbbbb  ∈ R for i ∈ {1;:::;p}, can be expressed (after ∑ which completes the proof of the observer part of (12). j=1 some algebraic transformations) as Remark 6: Upon the result (18), we can see that in the nominal conditions, when n(t) ≡ 0, the asymptotic relation ⊺ ∗ _ _ z ~ = (A A A −lll ccc )z ~ −lll n +bbbF ; 1 1 1 1 ls Y(t)Y → 0 as ! → ∞ resulting in the possibility of ∞ o1 ⊺ ⊺ ⊺ ⊺ ⊺ ∗ _ _ z ~ = (A A A −lll ccc )z ~ + (lll ccc −bbbbbb lll ccc )z ~ −bbbbbb lll n +bbbF getting an arbitrarily small value of  . i i i i i−1 i−1 1 z~ ⊺ 2 i−2 Let us define control error vector  = [e e _] ∈ R . The ⊺ ⊺ ⊺ −bbbbbb (lll ccc −lll ccc )z ~ ; for i ∈ {2;:::;p}: (13) Q j j+1 j application of feedback controller (10) to dynamics (9) gives j=1 0 1 0 0 0 0 _ =   +  z ~ −  n; (20) 2 2 Equations (13) allow us to write the dynamics of the −k −2k 0 2k 1 k ⊺ ⊺ 3p aggregated observation error  ≜ [z ~ ::: z ~ ] ∈ R in a form ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ 1 p K Z K K ZZ which can be transformed with substitution  =    """, where ~ ~ _ = H  + F + n; (14) −1 ≜ diag{k ; 1}, into where matrix H is lower triangular and its eigenvalues −1 −1 −1 """ _ =    K K K   """ +    Z Z Zz ~ −    n p " p " " " ∈ {−! ; − ! ; :::; − ! } for i ∈ {1;:::; 3p}, vector i o1 o1 o1 ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ = [bbb ::: bbb ] , and = [lll lll bbbbbb ::: lll bbbbbb ] . Introducing 0 1 1 1 1 " Z = k  "" +ZZz ~ −n: (21) ´¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¶ ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ −1 −2 p times p−1 times ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ the transformation  =     for    ≜ blkdiag{L L L ;:::;L L L } ∈ H H H 1 p " 3p×3p i−1 −2 i−1 −1 R where L L L ≜ diag{( ! ) ; ( ! ) ; 1} ∈ i o1 o1 Let us now introduce a Lyapunov function candidate V = 3×3 p×p R for i ∈ {1;:::;p} ∈ R , we can rewrite (14) to a form ⊺ 2 """ PPP """ ∶ R → R limited by  (PPP ) Y"""Y ≤ V (""") ≤ " ≥0 min " " −1 −1 ∗ −1 (PPP ) Y"""Y, where PPP ≻ 0 is the solution of Lyapunov _ max " " =   H    +   F +   n equation H H H PPP +PPP H H H = −III: The derivative " " " ∗ −1 = ! H H  + F +   n; (15) o1 ⊺ ⊺ ⊺ V = −k""" """ + 2""" PPP Z Z Zz ~ − 2""" PPP n " " p " where H H H is dependent only on parameter and its eigenval- " " P ≤ −k Y""Y + Y""Y (PP ) m Yz ~ Y + k SnS; (22) p max " Z p ues  ∈ {−1; − ;:::; − } for i ∈ {1;:::; 3p}. To conduct a stability analysis of the observation subsystem, let us introduce where m = max{1; 2k}, holds ⊺ 3p a Lyapunov function candidate V =  PPP  ∶ R → R lim- ≥0 2 2 V ≤ −(1 −  )k Y"""Y for " " ited by  (PPP ) YY ≤ V ≤  (PPP ) YY , where PPP ≻ 0 min   max 2 (PPP ) is the solution of Lyapunov equation H H H PPP + PPP H H H = −III: max " 2 Y"""Y ≥ m Yz ~ Y + k SnS (23) Z p The derivative of V , based on (15), can be written down as  k ⊺ ⊺ ∗ −1 The lower boundary of Y""Y is class K with respect to argu- _ _ V = −!   + 2 PP (F +   n) o1 ments Yz Y and SnS. According to the Remark 3, Assumption 2 ∗ 3 ≤ −! YY + 2 YY (PPP ) p ‰SF S + 3! SnSŽ (16) o1 max  and result (17), system (21) is ISS and satisfies o1 2 (PPP ) and holds max " "  ~ ls Y""(t)Y ≤  m ls Yz (t)Y + k r ∞ " Z ∞ p n V ≤ −(1 −  )! YY for o1 ⎡ √ P 2m  (PPP ) p √ √ 2 (PP ) Z max max " P P ≤   (⋅) 2 (PP ) p 6 (PP ) p! ∗ max  max  "  _ o1 ⎢ F k ! YY ≥ SF S + SnS; (17) ⎢ " o1 o1 2m  (PPP ) p! ⎥ Z max o1 2 where  ∈ (0; 1) is a chosen majorization constant. The + Œ + k ‘r ; (24) lower bound of YY is a class K function with respect to the _ » perturturbations SF S and SnS, so according to the Remark 3 where  =  (PPP )~ (PPP ). According to transforma- " max " min " and Assumption 2, system (15) is input-to-state stable (ISS), tion between original control error vector  and the trans- and according to [23], satisfies −1 formed """, we write Y _Y ≤ max{k ; 1} Y"""Y =∶ m Y"""Y and thus 2 (PPP ) p max ⎡ √ ls Y(t)Y ≤  (⋅) _ ∗ ⎢ 4 (PPP ) p ∞  2 (PPP ) F max " max ls Y(t)Y ≤ m   (⋅) o1  ∗ ∞ k "  _ ⎢ F k ⎢ ! " o1 6 (PPP ) p! ⎣ max o1 +  r ; (18) √ n ⎤ 4 (PPP ) p! ⎥ max o1 −1 2 + Œ + max{k ; 1}k ‘r =∶  ; (25) n e P P  ⎦ for  =  (PP )~ (PP ). Since  (  ) = max  min  max p−1 −2 max{1; ( ! ) } and z ~ is a subvector of  , we may which completes the proof of Theorem 1. o1 p ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS -2 0 2 4 6 8 10 10 10 10 10 10 10 10 Fig. 3: Laboratory setup, with a - buck converter, b - dSPACE controller, c - input voltage, d - oscilloscope, e - voltage sensor, Fig. 4: Bode diagram representing the module of G (j!). uy f - A/D converters, and g - PC with control software. controller bandwidth k, which is the range we expect the Remark 7: Similarly to the comment made in Remark 6, in closed-loop system to operate in, and the experiment sampling the case of n(t) ≡ 0 and upon the result (25), we can say that frequency ! . The green area represents the frequency range, ls Y(t)Y → 0 as ! → ∞ ∨ k → ∞, making it possible to ∞ o1 where CESO (p = 2 and p = 3) should react more rapidly get an arbitrarily small value of  . than the standard ESO, and red area is the range where only Remark 8: Upon the result (25), we may observe that the CESO p = 2 should provide quicker response with respect increasing gains of both observer and controller are amplify- to control errors. The points at the intersection of ! and ing measurement noise, thus, it is not recommended to use observers graphs indicate the amplification factors of high extremely high values of ! and k in practice. o1 frequency signals (e.g. measurement noise) within signal (t). Consequently, in the following experiments, we can expect IV. HARDWARE EXPERIMENT the measurement noise to be least amplified in CESO p = 3, followed by CESO p = 2, and finally in standard, single ESO. A. Testbed description The experimental setup used for the study is seen in Fig. 3. B. Test methodology The output voltage was measured by a Hall effect-based sensor and converted through a 16-bit A/D converter in the dSPACE The following experiments were conducted to test the platform. The output was recorded by a digital oscilloscope ADRC scheme with the proposed cascade ESO (CESO): and dedicated PC-based software. The sampling period was E1: Comparison with standard ESO (i.e. CESO with p = 1). set to T = 10 Hz. The physical parameters of the DC-DC E2: Influence of parameters ! (E2a), k (E2b), and (E2c). o1 converter, described with (1), were V = 20V, L = 0:01H, in E3: Impact of a low-pass filter (LPF) at the converter output. C = 0:001F, and R = 50 . This allowed to straightforwardly The control objective was to track a smooth voltage tra- calculate the system gain in (3) as b = V ~(CL) = 2 × 10 . in jectory v (t) despite the presence of a varying input-additive The tested control algorithm was first implemented in a external disturbance shown in Fig. 5. Such disturbance signal Matlab/Simulink-based model, from which a C code program is used here to test the robustness of the considered controllers was generated and run on the dSPACE controller in real-time. against different types of disturbances within one experimental Considering the above parameters of the utilized testbed and run. This specific shape of user-injected external disturbance the controller/observer structures introduced in (5), (10), and signal would not appear outside of a laboratory environment, (7), we can derive the transfer-function-based relation however, the character of disturbances designed in specific time intervals can be found in certain applications (e.g. [3]). U (j!) = G (j!) [E(j!) − N (j!)]; (26) uy ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ The reference trajectory was designed as a filtered and biased Y (j!) square signal with bias equal to 7V, amplitude of square signal where U (j!), E(j!), N (j!), and Y (j!) correspond re- equal to 6V, and period 1s. The filtering transfer function spectively to signals (t), e(t), n(t), and y(t) after Laplace applied to the square signal was G (s) = . f 2 0:025s +0:6s+4 transformation. The amplitude Bode diagram of G (j!), Although the most common control task in the control of buck uy obtained for the observer levels p ∈ {1; 2; 3} and tuned with converters is a set-point stabilization, trajectory following of the nominal parameters utilized in the experiment, is presented the output voltage can be occasionally seen in the literature, in Fig. 4. The vertical dashed lines represent the chosen e.g., [24]. Here, we consider a filtered piece-wise constant TABLE I: Used bandwidth parameterization of CESOs. TABLE II: Integral quality criteria for experiment E1. Cascade level Criterion p = 1 p = 2 p = 3 Observer type Bandwidth Se(t)Sdt S(t)Sdt S _ (t)Sdt ∫ ∫ ∫ 1st level ESO (! ) o1 Standard ESO (p = 1) 0.2310 0.5368 315.58 2nd level ESO (! ) – Cascade ESO (p = 2) 0.0467 0.5496 113.23 o2 3rd level ESO (! ) – –  Cascade ESO (p = 3) 0.0381 0.5545 29.11 o3 ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 0.15 Output 0.1 0.05 Reference 0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4 Control error Fig. 5: External disturbance applied in all experiments. reference to reach the desired level of the output voltage and avoid observer peaking caused by the discontinuities in v (t). -1 0 0.5 1 1.5 2 2.5 3 3.5 4 Control signal C. Experimental results The results of E1 are gathered in Fig. 6. The observer 0.5 bandwidth for the standard ESO (p = 1) was ! = 3600rad/s, o1 which was close to the maximum that could be obtained for a 0 0.5 1 1.5 2 2.5 3 3.5 4 10kHz sampling without observing any undesirable effects. Total disturbance For the comparison, only CESOs with p = 2 and p = 3 levels were utilized to maintain legibility of the results while not loosing their generality. In order to provide a systematic tuning methodology across tested observers, bandwidths of the CESOs were parameterized and set according to Table I with 0 0.5 1 1.5 2 2.5 3 3.5 4 = 3 and  = 3600rad/s. The controller gains from (10) were Fig. 6: Results of experiment E1. set to k = 6400 and k = 160 in each case, which corresponds p d to the controller bandwidth k = 80, introduced in Remark 5. One can notice from Fig. 6 that, with the applied tuning with the increase of ! , the amplitude of the control signal o1 methodology, all the tested controllers have realized the given increases but no visible improvement in the control accuracy task, however the standard ESO (p = 1) provided the worst can be observed. In other words, due to multiple factors performance in terms of tracking accuracy and noise sup- like maximum sampling frequency and noise characteristics, pression. On the other hand, with the increase of cascade increasing the observer bandwidth ! will at some point no o1 level p in CESO, better performance was achieved. This longer provide better performance. We can conclude that with observation is supported with the calculated integral quality the CESO one can achieve better control performance for indices in Table II. Besides the improvement of control error wider range of ! values, compared to the results of standard o1 performance, the transfer of sensor noise into the control signal ESO (p = 1) in Fig. 7(a). has decreased with the increase of parameter p thanks to the The results of E2b are depicted in Fig. 8. In the case of lower values of ! related to the first level of CESO. This o1 standard ESO (p = 1), it is clear that increasing the controller result is supported with the values of Su _ (t)Sdt criterion in bandwidth k improves the control accuracy while keeping a Table II, which represents the impact of rapid fluctuations of significant, undesired level of control signal and noise therein. the control signal, mostly caused by the amplified noise. In the case of proposed CESO (p = 2 and p = 3), increasing The initial premises formulated upon Fig. 4 have been the controller bandwidth k results in comparable control errors confirmed with the results in Fig. 6. As expected, the control while retaining similar level of control signal. Due to the signal with the lowest content of noise was obtained for CESO characteristics of CESO, it is possible to obtain better control p = 3, then CESO p = 2, and finally the standard ESO. performance for wider range of k values, compared to the Next, in order to provide potential CESO users with guide- results obtained for the standard ESO in Fig. 8(a). lines for its construction and tuning, the influence of its design The results of E2c are depicted in Fig. 9. In the case parameters was investigated. To this effect, the results of E2 of CESO (p = 2), increasing improves both the tracking are seen in Fig. 7-9. It should be noted that the estimated accuracy and noise suppression in the control signal. However, total disturbance is part of the control signal (see (5)) so its in the case of CESO (p = 3), increasing keeps improving influence is explicitly visible in the control signal. the noise suppression in the control signal but at some point The results of E2a are depicted in Fig. 7. In the case deterioration in the tracking accuracy can be spotted. It results of standard ESO (p = 1), the well-known relation from from a fact that, in this case, the observer bandwidth ! is set high-gain observers, discussed in the Introduction, can be o1 too small, which makes the observer not providing fast-enough noticed. Namely, with the increase of observer bandwidth ! , o1 and accurate-enough estimate of the first state variable of the significant noise amplification occurs in the control signal. At extended state vector. the same time, a slight improvement of the control error was obtained. In the case of proposed CESO (p = 2 and p = 3), Let us now focus on some frequency-domain insights con- ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS cerning experiment E3. An algebraic transformation of (5), 1 1 using (11), allows to write down the form of a generalized 0.8 controller utilizing p-level cascade observer, that is directly 0.5 dependent on the observation error of total disturbance z ~ , 0.6 p3 0 −1 i.e.,  = b (z − z ~ +  ). The transformation of (14) into 1 p3 0 0.4 Laplace domain allows to write that for every p ≥ 1 -0.5 0.2 -1 0 Z (j!) = G (j!)N (j!) + G (j!)Z (j!); (27) p3 z~ n z~ z 3 p3 p3 3 024 024 (a) standard ESO (p = 1) ~ where Z (j!) and Z (j!) are the Laplace-domain equiva- p3 3 lents of signals z ~ (t) and z (t). Application of the LPF p3 3 1 1 G = 1~(s + 1);  > 0; (28) LPF 0.8 0.5 0.6 0.4 at the output of the converter (in order to filter-out measure- -0.5 ment noise) affects the total disturbance signal with a filtered- 0.2 out parts of the measured signal z and results in a following -1 0 024 024 extended form of (27), i.e., (b) CESO (p = 2) Z (j!) = G (j!)N (j!) + G (j!)Z (j!) p3 z~ n z~ z 3 p3 p3 3 1 1 + G (j!)Z (j!); (29) z~ z 1 p3 1 0.8 0.5 where Z (j!) corresponds to signal z (t) after Laplace 1 1 0.6 transformation. According to [19], the high-gain extended 0.4 observer performance should not be substantially affected for -0.5 0.2 small enough values of time constant  of the low-pass filter. -1 0 We assume that  has been chosen appropriately, and hence 024 024 focus on the noise-connected characteristics of the ADRC with (c) CESO (p = 3) analyzed observers. The amplification of particular frequencies of the measurement noise using ESO and CESO (p = 2; 3) with Fig. 7: Results of experiment E2a. parameters = 3 and  = 3600rad/s (see Table I) has been presented in Fig. 10. The dashed lines represent the magnitude of G when a low-pass filter was applied while the regions z~ n p3 1.5 1 with corresponding colors illustrate the set of characteristics 0.8 that would be obtained for a practically useful set of values 0.5 0.6 ∈ [0:0001; 0:01]s, where the bottom edge corresponds to 0.4 = 0:01s and the top edge corresponds to  = 0:001s. -0.5 0.2 Looking at Fig. 10, one can notice that the maximal value of -1 YG Y for CESO (p = 3) without output filtering was similar, -1.5 0 z~ n p3 024 024 or smaller, compared to the characteristics obtained with the (a) standard ESO (p = 1) conventional ESO with LPF with  = 0:001s, so the expected 1.5 1 content of the measurement noise in signal z ~ affecting the p3 0.8 control signal should be similar, or lower. This observation was 0.5 0.6 validated by time-domain results of experiment E3, presented 0.4 in Fig. 11, where the amplitude of noise-dependent oscillations -0.5 0.2 -1 is  ≈  ≈ 0:05. The presented values of the control ESO+LPF CESO -1.5 0 error illustrate the essential difference in the measurement 024 024 noise handling by the CESO, compared to the use of a LPF. (b) CESO (p = 2) The proposed cascade observer structure suppresses the effect 1.5 1 of measurement noise amplification in the control signal but 0.8 does not change the noise level at the output, while the use of 0.5 0.6 0 a LPF decreases the level of measurement noise at the output 0.4 -0.5 but does not change the noise amplification feature of the high- 0.2 -1 gain ESO. In order to improve the overall performance of the -1.5 0 control system, in terms of robustness against measurement 024 024 noise, a LPF can be utilized along CESO. Such example is illustrated in Fig. 11, where the combination of CESO and (c) CESO (p = 3) LPF achieves the amplitude value  ≈ 0:02, which is CESO+LPF Fig. 8: Results of experiment E2b. smaller than the aforementioned  and  . ESO+LPF CESO ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] [-] [-] [-] [-] [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 1.5 1 1 1 0.8 0.8 0.5 0.5 0.6 0.6 0 0.4 0 0.4 -0.5 0.2 0.2 -0.5 -1 0 024 024 1 024 024 (a) CESO p = 2 0.82 1.5 1 1 0.8 0.78 0.5 0.6 0.74 0 0.4 0.46 0.5 0.54 0.58 0.62 -0.5 0.2 -1 0 Fig. 11: Results of experiment E3. 024 024 (b) CESO p = 3 Development, non-competitive project entitled “International Fig. 9: Results of experiment E2c. scholarship exchange of PhD students and academic staff” executed under the Activity 3.3 specified in the application for funding of project No. POWR.03.03.00-00-PN13 / 18. The work has been also supported by the Fundamental Research Funds for the Central Universities project no. 21620335. REFERENCES [1] R. Madonski, K. Łakomy, M. Stankovic, S. Shao, J. Yang, and S. Li, “Robust converter-fed motor control based on active rejection of multiple disturbances,” Control Engineering Practice, vol. 107, DOI 1 2 3 4 5 6 7 10 10 10 10 10 10 10 https://doi.org/10.1016/j.conengprac.2020.104696, p. 104696, 2021. [2] J. Yang, H. Wu, L. Hu, and S. Li, “Robust predictive speed regulation of converter-driven DC motors via a discrete-time reduced-order GPIO,” Fig. 10: Bode diagram representing the module of G (j!). z~ n p3 IEEE Trans. on Ind. Electronics, vol. 66, no. 10, pp. 7893–7903, 2019. [3] J. Yang, H. Cui, S. Li, and A. Zolotas, “Optimized active disturbance rejection control for DC-DC buck converters with uncertainties using a reduced-order GPI observer,” IEEE Transactions on Circuits and In order to summarize the results obtained in this work Systems I: Regular Papers, vol. 65, no. 2, pp. 832–841, 2018. and allow for their quick assessment, Table III compares the [4] J. Han, “From PID to active disturbance rejection control,” IEEE standard ESO with the proposed CESO using selected criteria. Transactions on Industrial Electronics, vol. 56, no. 3, pp. 900–906, 2009. [5] S. Chen and Z. Chen, “On active disturbance rejection control for a class of uncertain systems with measurement uncertainty,” IEEE Transactions V. CONCLUSIONS on Industrial Electronics, vol. 68, no. 2, pp. 1475–1485, 2021. [6] E. Sariyildiz, R. Oboe, and K. Ohnishi, “Disturbance observer-based An active disturbance rejection control with a novel cascade robust control and its applications: 35th anniversary overview,” IEEE extended state observer (CESO) for DC-DC buck converters Trans. on Industrial Electronics, vol. 67, no. 3, pp. 2042–2053, 2020. has been proposed. The validity of the new approach has [7] W. H. Chen, J. Yang, L. Guo, and S. Li, “Disturbance-observer-based control and related methods - an overview,” IEEE Transactions on been shown through a dedicated stability analysis and a Industrial Electronics, vol. 63, no. 2, pp. 1083–1095, 2016. set of hardware experiments. The comparison between the [8] K. Łakomy, R. Patelski, and D. Pazderski, “ESO architectures in the proposed cascade ESO-based ADRC and a standard single trajectory tracking ADR controller for a mechanical system: a compar- ison,” in Advanced, Contemporary Control, pp. 1323–1335, 2020. ESO-based ADRC showed that the former has stronger capa- [9] H. K. Khalil and L. Praly, “High-gain observers in nonlinear feedback bilities of sensor noise suppression and provides better control control,” International Journal of Robust and Nonlinear Control, vol. 24, performance (understood as tracking accuracy and energy no. 6, pp. 993–1015, 2014. [10] S. Sugahara and S. Matsunaga, “Fundamental study of influence of ripple efficiency). The structure of the proposed ADRC is bulkier noise from DC–DC converter on spurious noise of wireless portable than the conventional one but in return provides an additional equipment,” IEEE Transactions on Power Electronics, vol. 31, no. 3, pp. 2111–2119, 2016. and practically appealing degree of freedom in shaping the [11] A. A. Prasov and H. K. Khalil, “A nonlinear high-gain observer for influence of measurement noise on the observer/controller part. systems with measurement noise in a feedback control framework,” IEEE Trans. on Automatic Control, vol. 58, no. 3, pp. 569–580, 2013. [12] S. Battilotti, “Robust observer design under measurement noise with VI. ACKNOWLEDGMENT gain adaptation and saturated estimates,” Automatica, vol. 81, pp. 75– 86, 2017. The article was created thanks to participation in program [13] W. Xue, X. Zhang, L. Sun, and H. Fang, “Extended state filter based PROM of the Polish National Agency for Academic Ex- disturbance and uncertainty mitigation for nonlinear uncertain systems change. The program is co-financed from the European Social with application to fuel cell temperature control,” IEEE Transactions on Fund within the Operational Program Knowledge Education Industrial Electronics, vol. 67, no. 12, pp. 10 682–10 692, 2020. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] [-] [-] [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS TABLE III: Comparison between standard ESO- and proposed CESO-based control with selected criteria. Observer type Standard ESO (p = 1) Cascade ESO (p = 2) Cascade ESO (p = 3) Criterion Observer tuning methodology parameterization [22] parameterization [22] + Table I ^ ^ ^ Design parameters k;b;! k;b;! , k;b;! , o1 o1 o1 Total disturbance estimation quality (Fig. 6) — ↗ ↗↗ Control quality (« Se(t)Sdt in Table II) — ↗ ↗↗ Control effort (« S(t)Sdt in Table II) — ↗ ↗↗ Control jittering (« S _ (t)Sdt in Table II) — ↘ ↘↘ Noise content in control signal with output LPF (Fig. 10) ↘ ↘ ↘ Implementation complexity (no. of observer state variables) 3 6 9 Stability type (nominal conditions) asymptotic [5], [18] asymptotic (Remark 6 in Sect. III) Stability type (non-nominal conditions) practical [5], [18] practical (Theorem 1 in Sect. III) [14] Y. Wu, A. Isidori, and L. Marconi, “Achieving almost feedback- [20] Y. Huang and W. Xue, “Active disturbance rejection control: Method- linearization via low-power extended observer,” IEEE Control Systems ology and theoretical analysis,” ISA Transactions, vol. 53, no. 4, pp. Letters, vol. 4, no. 4, pp. 1030–1035, 2020. 963–976, 2014. [15] L. Wang, D. Astolfi, L. Marconi, and H. Su, “High-gain observers [21] R. Madonski, S. Shao, H. Zhang, Z. Gao, J. Yang, and S. Li, “Gen- with limited gain power for systems with observability canonical form,” eral error-based active disturbance rejection control for swift indus- Automatica, vol. 75, pp. 16–23, 2017. trial implementations,” Control Engineering Practice, vol. 84, DOI [16] H. K. Khalil, “Cascade high-gain observers in output feedback control,” 10.1016/j.conengprac.2018.11.021, pp. 218–229, 2019. Automatica, vol. 80, pp. 110–118, 2017. [22] Z. Gao, “Scaling and bandwidth-parameterization based controller tun- [17] D. Astolfi, M. Jungers, and L. Zaccarian, “Output injection filtering ing,” in American Control Conference, vol. 6, pp. 4989–4996, 2003. redesign in high-gain observers,” in Proc. European Control Conference, [23] H. K. Khalil, Nonlinear systems, 3rd edition. Prentice Hall, 2002. pp. 1957–1962, 2018. [18] K. Łakomy and R. Madonski, “Cascade extended state observer for [24] R. Silva-Ortigoza, V. M. Hernandez-Guzman, M. Antonio-Cruz, and active disturbance rejection control applications under measurement D. Munoz-Carrillo, “DC/DC buck power converter as a smooth starter noise,” ISA Transactions, DOI 10.1016/j.isatra.2020.09.007, 2020. for a DC motor based on a hierarchical control,” IEEE Transactions on [19] H. K. Khalil and S. Priess, “Analysis of the use of low-pass filters with Power Electronics, vol. 30, no. 2, pp. 1076–1084, 2015. high-gain observers,” IFAC, vol. 49, no. 18, pp. 488–492, 2016. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Computing Research Repository arXiv (Cornell University)

Active Disturbance Rejection Control Design with Suppression of Sensor Noise Effects in Application to DC-DC Buck Power Converter

Download PDF
9 pages

Loading next page...
 
/lp/arxiv-cornell-university/active-disturbance-rejection-control-design-with-suppression-of-sensor-CplminhZyg

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

ISSN
0278-0046
eISSN
ARCH-3344
DOI
10.1109/TIE.2021.3055187
Publisher site
See Article on Publisher Site

Abstract

This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS Active Disturbance Rejection Control Design with Suppression of Sensor Noise Effects in Application to DC-DC Buck Power Converter Krzysztof Łakomy, Rafal Madonski, Bin Dai, Jun Yang, Senior Member, IEEE, Piotr Kicki, Maral Ansari, Student Member, IEEE, Shihua Li, Fellow, IEEE Abstract —The performance of active disturbance rejec- strongly amplified measurement noise into the control signal tion control (ADRC) algorithms can be limited in practice calculated upon the state vector of ESO. This may cause by high-frequency measurement noise. In this work, this the decrease of control quality (e.g. when the noise-affected problem is addressed by transforming the high-gain ex- control signal hits the actuator saturation), higher energy tended state observer (ESO), which is the inherent element consumption, and quicker wear of the equipment. The HGO- of ADRC, into a new cascade observer structure. Set of experiments, performed on a DC-DC buck power converter based ADRC design and tuning often come down to a forced system, show that the new cascade ESO design, com- compromise between speed/accuracy of signals reconstruction pared to the conventional approach, effectively suppresses and sensitivity to noise [9]. Same compromise can be seen in the detrimental effect of sensor noise over-amplification the ADRC works for buck converters in which the measured while increasing the estimation/control performance. The system output (voltage) is oftentimes corrupted with high- proposed design is also analyzed with a low-pass filter at the converter output, which is a common technique for frequency noise [10]. Several types of solutions were proposed reducing measurement noise in industrial applications. to solve the problem of attenuating the effects of measurement noise in high-gain observers. They mainly address it by: Index Terms—noise suppression, power converter, high- gain observer, extended state observer, ESO employing nonlinear [4], [11] or adaptive techniques [12], redesigning the local behavior by combining different types of observers [13], using low-power structures [14]–[16], or I. INTRODUCTION modifying standard low-pass filters [17]. ENEWABLE energy sources, like fuel and photovoltaic Motivated by the above problem, a new cascade ESO-based cells, are rapidly evolving technologies for DC voltage error-domain ADRC solution is presented. Following the gen- generation, which results in proliferation of DC–DC buck eral idea shown in [18], we propose a virtual decomposition converters in power applications. Practically appealing results of the total disturbance present in the DC-DC buck converter on buck converter control using the idea of active disturbance system, allowing to design a cascade structure of ESO, where rejection control (ADRC) were recently reported in [1]–[3]. each level of the observer cascade is responsible for handling The key element in any ADRC scheme is the extended state a particular type and frequency range of estimated signal. The observer (ESO [4]), responsible for estimating the system state proposed topology enhances conventional state/disturbance vector and reconstructing the overall disturbance (also referred estimation performance while avoiding over-amplification of to as total disturbance) affecting the controlled variable [5]. the sensor noise. The user-defined number of cascade levels However, since the conventional form of ADRC uses a high- allows to customize the overall control system structure to gain observer (HGO) structure to estimate selected signals, meet certain disturbance rejection requirements. Although a its capabilities are intrinsically limited by the presence and multi-level cascade observer is proposed, a straightforward severity of high-frequency sensor noise, as discussed in [6]– design and implementation methodology is given, together [8]. The high gains of the observer cause the transfer of with intuitive tuning rules. The novelty of this work includes Manuscript received Month xx, 2xxx; revised Month xx, xxxx; ac- an experimental validation of the proposed cascade ESO- cepted Month x, xxxx. based ADRC structure, a proof of the input-to-state stability K. Łakomy is with the Poznan ´ University of Technology, Poznan, ´ 60- of the closed-loop system, and additional insights about the 965, Poland (e-mail: krzysztof.lakomy92@gmail.com). R. Madonski (corresponding author) is with the Energy Electricity Re- sensor noise suppressing effects in frequency domain. The search Center, International Energy College, Jinan University, Zhuhai, experimental study also addresses the impact of a low-pass 519070, P. R. China (e-mail: rafal.madonski@jnu.edu.cn). filter implemented at the converter output, which is a popular B. Dai, J. Yang and S. Li are with the School of Automation, Southeast University, Key Laboratory of Measurement and Control of approach for handling high-frequency sensor noise [19]. CSE, Ministry of Education, Nanjing, 210096, P. R. China (e-mail: Notation. Within this article, we treat R as a set of real fbin 1994/j.yang84/lshg@seu.edu.cn). numbers, R = {x ∈ R ∶ x > 0} as a set of positive real P. Kicki is with the Institute of Robotics and Machine Intelligence, ´ ´ Poznan University of Technology, Poznan, 60-965, Poland (e-mail: pi- numbers, R = {x ∈ R ∶ x > 0} as a set of non-negative real ≥0 otr.kicki@put.poznan.pl). numbers, Z as a set of integers,  (A A A) and  (A A A) are min max M. Ansari is with the Faculty of Engineering and Information Tech- respectively the minimal and maximal eigenvalues of matrix nology, University of Technology Sydney, Sydney NSW, 2007, Australia (e-mail: maral.ansari@student.uts.edu.au). A A A, while A A A ≻ 0 means that matrix A A A is positive definite. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. arXiv:2009.02948v2 [eess.SY] 4 Feb 2021 This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS Function f (x) ∶ R → R belongs to class K when it is strictly increasing and f (0) = 0. The expression ls ∶= lim sup is t→∞ used for the sake of notation compactness. II. PRELIMINARIES A. Simplified plant model and control objective Following [3], an average dynamic model of a DC-DC buck Fig. 1: Semiconductor realization of the considered DC-DC converter, depicted in Fig. 1, can be written as buck power converter, with diode V D and control switch V T . dv (t) ⎧ o 1 1 ⎪ = i (t) − v (t); L o ⎪ dt C CR di (t) L V 1 in ⎨ = [(t) + d(t)] − v (t); (1) o Combining the uncertain (or unknown) terms in (2), includ- dt L L ⎪ ing the imperfect identification of the input gain, results in a y (t) = v (t) + n(t); o o following form of the output voltage dynamics where  ∈ [0; 1] is the duty ratio, y [V] is the measured system ^ ^ ^ v  = a v + a v _ + b − b + bd+b = F (⋅) + b; (3) output that consists of the average capacitor voltage v [V] and o 2 o 1 o ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ the sensor noise n[V], i [A] is the average inductor current, F(t;v_ ;v ;;d) o o R[ ] is the load resistance of the circuit, L[H] is the filter where b ≠ 0 is a precise-enough estimate of the input gain b inductance, C [F] is the filter capacitance, V [V] is the input in from (2) and F (⋅) represents the total disturbance of (3). voltage source, and d(t) represents the unknown (possibly Since v (t) and its derivatives may not be known a priori, time-varying and nonlinear) external disturbance. which may lead to possible inability of constructing feedfor- The considered control objective is to force v (t) to follow ward signal in , let us reformulate (3) in error-domain a reference capacitor output voltage trajectory v (t)[V] by manipulating (t) with following assumptions applying. e  = v  − v  = v  − F (⋅)−b; (4) r o r Assumption 1: Following the limitations resulting from the ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ F (⋅;v ) physical properties of the considered electronic circuit, we may assume that the values of voltage and current are bounded, and where e(t) ≜ v (t)−v (t) is the control error signal and F (⋅) r o belong to some compact set such that sup Si (t)S < r and L i is the total disturbance in the error-domain [21]. In this article, t≥0 L sup Sv (t)S < r for r ;r > 0. o v i v we utilize a standard form of the ADRC controller t≥0 o L o Assumption 2: Output voltage v (t) is the only measur- −1 ∗ = b (F +  ); (5) able signal and is additionally corrupted by bounded, high- frequency measurement noise sup Sn(t)S < r for r > 0. n n which is constructed to simultaneously compensate the in- t≥0 Assumption 3: [20] The unknown external disturbance d(t) fluence of disturbance using the estimated value of total may have a countable number of first-class discontinuity disturbance (F ) and to stabilize system (4) in a close vicinity points at times t = T for i ∈ {1;:::;N }, N ∈ Z, 0 ≤ N < ∞, i d d d of the equilibrium point e = 0 using the output-feedback and 0 < inf (T −T ) < ∞ for N > 1. In all other i+1 i d stabilizing controller  . i∈{1;:::;N −1} 0 moments, the external disturbance function is bounded and has Assumption 5: Stabilizing controller  has a structure that bounded first time derivative, i.e., sup Sd(t)S < r and d guarantees the boundedness of  (⋅) and  _ (⋅). Although t≥0;t∈~{T } 0 0 sup Sd(t)S < r for some r ;r > 0 and i ∈ {1;:::;N }. this assumption may seem conservative, it is relaxed with the _ _ d d t≥0;t∈~{T } d d Assumption 4: The reference signal v (t) may have a previously introduced Assumptions 1, 3, and 4. Remark 1: Since the disturbance F and the control variable countable number of first-class discontinuity points at times have equal relative rank, with respect to the voltage v t = T for i ∈ {1;:::;N }, N ∈ Z, 0 ≤ N < ∞, and 0 < i r r r o representing the output of the original system (see (1)), the inf (T − T ) < ∞ for N > 1. There also exists i+1 i r i∈{1;:::;N −1} total disturbances affecting the error-domain system, described a positive constant r , such that v (t) and its specific time- v r (j) with (4), meet the so-called matching condition. The specific derivatives satisfy inequality sup šUv (t)UŸ ≤ r , r v t≥0;t∈~{T } r differences and control solutions for matched and mismatched for i ∈ {1;:::;N } and j ∈ {0; 1; 2; 3}. disturbances have been thoroughly discussed in [7]. We will first put the focus on precise and on-line estimation B. Application of the ADRC principle of perturbing term F (⋅), crucial for proper active disturbance Following the standard ADRC design, system model (1) is rejection. To calculate F , we first need to define the extended ⊺ ∗ ⊺ 3 reformulated, emphasizing its input-output relation state z = [z z z ] ≜ [e e _ F ] ∈ D , where D ≜ {xxx ∈ R ∶ 1 2 3 z z 2 YxY < r } for some r ∈ R . The dynamics of the state vector z z + d v (t) 1 dv (t) 1 V o o in = − − v (t) + [(t) + d(t)]: o z can be expressed, upon (4), as a state-space model dt CR dt CL CL ± ⎧ ² ² ^ _ z _ = A A Az −dddb +bbbF ; a a b ⎪ 1 2 ⎨ (6) (2) y = e − n = ccc z − n; Function f(x) ∶ R → R has first-class discontinuity at point x  if for 2×1 2×2 ⊺ ⊺ 000 III + − + − where A A A ≜  , ddd ≜ [0 1 0] , ccc ≜ [1 0 0] , and bbb ≜ 1×2 f ∶= lim + f(x) and f ∶= lim − f(x), it satisfies f ≠ f and x→x  x→x  0 000 + − ⊺ max{f ;f } ≤ r for some r > 0. [0 0 1] . Given (6), the output of this system y corresponds f f ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS to the control error e which, according to Assumption 2, is DC-DC (design parameter) influenced by the measurement noise n. converter (1) Remark 2: Control error e, together with its derivative e _ are p-th level - v o ... n v cascade ESO (7) - r bounded according to the Assumptions 1, 3, and 4, and the specific form of the system dynamics (1). 1 EE SS O O Remark 3: Under the Assumptions 1, 3 and 4, func- 1s1 t l seve t level l () ... tion F (t) is continuously differentiable, and thus, there μ μ Controller ESO exist bounded continuous functions ; such that 2 F _ ∗ . (5) 2nd level ∗ ∗ sup SF (t)S < (e;e; _ v ;v _ ;v  ;), sup SF (t)S < F r r r t≥0 t≥0 ... ⊺ 2 (e;e; _ v ;v _ ;v  ; v ;; _ ), for all [e e _] ∈ R . Both prac- d _ ∗ r r r r : tical and theoretical justifications of lumping selected com- ponents as parts of F (⋅), including control signal and state- z ξ ESO dependent variables, has been thoroughly discussed in [5]. p-th level III. MAIN RESULT: PROPOSED CASCADE ESO ADRC State selector (8) To calculate the estimated value of extended state vector z, Fig. 2: Proposed ADRC with sensor noise suppression via let us now introduce a novel p-level structure of a cascade cascade ESO structure for the DC-DC buck power converter. observer (p ∈ Z and p ≥ 2) in a following form _ ^ (t) = A A A (t) −dddb(t) +lll y(t) −ccc  (t) 1 1 1 i−1 and its inclusion in the overall estimate of the extended state ⎛ ⎞ (t) = A A A (t) +ddd −b(t) +bbb  (t) i j vector (8). The following observer levels are using the state ⎝ ⎠ j=1 vectors of previous observer levels instead of the measured +lll ccc [ (t) − (t)]; i ∈ {2;:::;p}; (7) i i−1 i signal, and thus, result in lower noise amplification than the single-level ESO with high bandwidth. Important part in the ⊺ 3 where  ≜ [   ] ∈ R is the state of a particular j j;1 j;2 j;3 utilized cascade observer structure is the state selector (8), 2 3 ⊺ 3 observer cascade level, lll ≜ [3! 3! ! ] ∈ R is the j oj oj oj which defines which estimated state variables (and from which j−1 observer gain vector with design parameter ! ≜ ! ∈ oj o1 observer level) participate in the controller synthesis (5) to R for > 1; ! ∈ R , and j ∈ {1;:::;p}. The estimate of + o1 + provide improved sensor noise effect suppression. z, resulting from the observer (7) can be expressed as Having z ^, the application of control action (5) to the system p−1 (4) results in a following second-order error dynamics ⊺ ⊺ 3 ^  bb z = [z ^ z ^ z ^ ] ≜  +bbbb Q  ∈ R : (8) 1 2 3 p j j=1 ∗ e  = F −  ; (9) Remark 4: It is worth noting, that if we reduce the observer ∗ ∗ ∗ ~ ^ where F ≜ F −F is the final residue of the total disturbance to a single level (p = 1), we would obtain a standard form resulting from the imperfect observation of F by observer (7). of a linear high-gain ESO, as seen in [22]. An introduction A block diagram of the proposed ADRC with cascade ESO of the subsequent cascade levels allows us to keep the same for the DC-DC buck power converter is shown in Fig. 2. observation quality with smaller values of ! , resulting in o1 Theorem 1: Under Assumptions 3-5, and by taking a stabi- a decrease of the measurement noise amplification visible in lizing proportional-derivative controller the state estimates, see (7). This effect will be depicted in the upcoming experiments. ≜ k y + k z ^ ; k ;k > 0; (10) 0 p d 2 p d The idea of cascade observer structure, proposed in (7) and illustrated in Fig. 2, is based on a specific choice of the observation errors of the extended state obtained with the the first level observer bandwidth ! , which should be large o1 p-level cascade observer, defined as enough to guarantee precise estimation of the first element p−1 of extended state vector z, and low enough to make the ⊺ ⊺ 3 z ~ = [z ~ z ~ z ~ ] ≜ z − z ^; = z − −bbbbbb  ∈ R ; (11) p p1 p2 p3 p j first level of the cascade to act as a low-pass filter for the j=1 noise. Latter elements of the extended state vector, i.e. z and together with the control error e, described with the dynamics z , usually have faster transients, and thus, are not estimated (9), are bounded. In other words precisely with the first level observer with a low ! value. o1 The consecutive observer levels are introduced to improve the ∀ ∀ ∃ ls Yz ~ (t)Y <  ∧ ls Se(t)S <  ; (12) t>t ! ;k>0  ; >0 ∞ p z~ ∞ e 0 o1 z~ e estimation quality of z and z using higher observer band- 2 3 where t = max{T ;T } results from Assumptions 3 and 4. widths ! (i > 1) and improve the observation performance by oi 0 N N d r incrementally extending the range of precisely estimated signal Remark 5: To keep the notational conciseness of the fol- frequencies. The introduction of additional cascade levels of lowing theoretical analysis and to reduce the overall number the observer can be interpreted as an attempt to estimate the of tuning parameters, we propose, following [22], to tune the total disturbance residue, that could not be precisely estimated stabilizing controller (10) with a single parameter k > 0, setting with the previous cascade levels due to limited bandwidth, the values of proportional and derivative gains, respectively, as ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. p = 1 p = 2 p This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS k = k and k = 2k. Chosen tuning procedure places the poles write down that Yz ~ Y ≤ ZZ ≤  (   ) YY and thus that the p d p max of control error dynamics (9) at value −k. asymptotic relation Proof of Theorem 1. The dynamics of the observation error ls Yz ~ (t)Y ≤  (   )ls Y(t)Y =∶  ; (19) ∞ p max  ∞ z~ defined for a particular cascade level, i.e. z ~ ≜ z −  − i i ⊺ i−1 3 bbbbbb  ∈ R for i ∈ {1;:::;p}, can be expressed (after ∑ which completes the proof of the observer part of (12). j=1 some algebraic transformations) as Remark 6: Upon the result (18), we can see that in the nominal conditions, when n(t) ≡ 0, the asymptotic relation ⊺ ∗ _ _ z ~ = (A A A −lll ccc )z ~ −lll n +bbbF ; 1 1 1 1 ls Y(t)Y → 0 as ! → ∞ resulting in the possibility of ∞ o1 ⊺ ⊺ ⊺ ⊺ ⊺ ∗ _ _ z ~ = (A A A −lll ccc )z ~ + (lll ccc −bbbbbb lll ccc )z ~ −bbbbbb lll n +bbbF getting an arbitrarily small value of  . i i i i i−1 i−1 1 z~ ⊺ 2 i−2 Let us define control error vector  = [e e _] ∈ R . The ⊺ ⊺ ⊺ −bbbbbb (lll ccc −lll ccc )z ~ ; for i ∈ {2;:::;p}: (13) Q j j+1 j application of feedback controller (10) to dynamics (9) gives j=1 0 1 0 0 0 0 _ =   +  z ~ −  n; (20) 2 2 Equations (13) allow us to write the dynamics of the −k −2k 0 2k 1 k ⊺ ⊺ 3p aggregated observation error  ≜ [z ~ ::: z ~ ] ∈ R in a form ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ 1 p K Z K K ZZ which can be transformed with substitution  =    """, where ~ ~ _ = H  + F + n; (14) −1 ≜ diag{k ; 1}, into where matrix H is lower triangular and its eigenvalues −1 −1 −1 """ _ =    K K K   """ +    Z Z Zz ~ −    n p " p " " " ∈ {−! ; − ! ; :::; − ! } for i ∈ {1;:::; 3p}, vector i o1 o1 o1 ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ ⊺ = [bbb ::: bbb ] , and = [lll lll bbbbbb ::: lll bbbbbb ] . Introducing 0 1 1 1 1 " Z = k  "" +ZZz ~ −n: (21) ´¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¶ ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ −1 −2 p times p−1 times ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ the transformation  =     for    ≜ blkdiag{L L L ;:::;L L L } ∈ H H H 1 p " 3p×3p i−1 −2 i−1 −1 R where L L L ≜ diag{( ! ) ; ( ! ) ; 1} ∈ i o1 o1 Let us now introduce a Lyapunov function candidate V = 3×3 p×p R for i ∈ {1;:::;p} ∈ R , we can rewrite (14) to a form ⊺ 2 """ PPP """ ∶ R → R limited by  (PPP ) Y"""Y ≤ V (""") ≤ " ≥0 min " " −1 −1 ∗ −1 (PPP ) Y"""Y, where PPP ≻ 0 is the solution of Lyapunov _ max " " =   H    +   F +   n equation H H H PPP +PPP H H H = −III: The derivative " " " ∗ −1 = ! H H  + F +   n; (15) o1 ⊺ ⊺ ⊺ V = −k""" """ + 2""" PPP Z Z Zz ~ − 2""" PPP n " " p " where H H H is dependent only on parameter and its eigenval- " " P ≤ −k Y""Y + Y""Y (PP ) m Yz ~ Y + k SnS; (22) p max " Z p ues  ∈ {−1; − ;:::; − } for i ∈ {1;:::; 3p}. To conduct a stability analysis of the observation subsystem, let us introduce where m = max{1; 2k}, holds ⊺ 3p a Lyapunov function candidate V =  PPP  ∶ R → R lim- ≥0 2 2 V ≤ −(1 −  )k Y"""Y for " " ited by  (PPP ) YY ≤ V ≤  (PPP ) YY , where PPP ≻ 0 min   max 2 (PPP ) is the solution of Lyapunov equation H H H PPP + PPP H H H = −III: max " 2 Y"""Y ≥ m Yz ~ Y + k SnS (23) Z p The derivative of V , based on (15), can be written down as  k ⊺ ⊺ ∗ −1 The lower boundary of Y""Y is class K with respect to argu- _ _ V = −!   + 2 PP (F +   n) o1 ments Yz Y and SnS. According to the Remark 3, Assumption 2 ∗ 3 ≤ −! YY + 2 YY (PPP ) p ‰SF S + 3! SnSŽ (16) o1 max  and result (17), system (21) is ISS and satisfies o1 2 (PPP ) and holds max " "  ~ ls Y""(t)Y ≤  m ls Yz (t)Y + k r ∞ " Z ∞ p n V ≤ −(1 −  )! YY for o1 ⎡ √ P 2m  (PPP ) p √ √ 2 (PP ) Z max max " P P ≤   (⋅) 2 (PP ) p 6 (PP ) p! ∗ max  max  "  _ o1 ⎢ F k ! YY ≥ SF S + SnS; (17) ⎢ " o1 o1 2m  (PPP ) p! ⎥ Z max o1 2 where  ∈ (0; 1) is a chosen majorization constant. The + Œ + k ‘r ; (24) lower bound of YY is a class K function with respect to the _ » perturturbations SF S and SnS, so according to the Remark 3 where  =  (PPP )~ (PPP ). According to transforma- " max " min " and Assumption 2, system (15) is input-to-state stable (ISS), tion between original control error vector  and the trans- and according to [23], satisfies −1 formed """, we write Y _Y ≤ max{k ; 1} Y"""Y =∶ m Y"""Y and thus 2 (PPP ) p max ⎡ √ ls Y(t)Y ≤  (⋅) _ ∗ ⎢ 4 (PPP ) p ∞  2 (PPP ) F max " max ls Y(t)Y ≤ m   (⋅) o1  ∗ ∞ k "  _ ⎢ F k ⎢ ! " o1 6 (PPP ) p! ⎣ max o1 +  r ; (18) √ n ⎤ 4 (PPP ) p! ⎥ max o1 −1 2 + Œ + max{k ; 1}k ‘r =∶  ; (25) n e P P  ⎦ for  =  (PP )~ (PP ). Since  (  ) = max  min  max p−1 −2 max{1; ( ! ) } and z ~ is a subvector of  , we may which completes the proof of Theorem 1. o1 p ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS -2 0 2 4 6 8 10 10 10 10 10 10 10 10 Fig. 3: Laboratory setup, with a - buck converter, b - dSPACE controller, c - input voltage, d - oscilloscope, e - voltage sensor, Fig. 4: Bode diagram representing the module of G (j!). uy f - A/D converters, and g - PC with control software. controller bandwidth k, which is the range we expect the Remark 7: Similarly to the comment made in Remark 6, in closed-loop system to operate in, and the experiment sampling the case of n(t) ≡ 0 and upon the result (25), we can say that frequency ! . The green area represents the frequency range, ls Y(t)Y → 0 as ! → ∞ ∨ k → ∞, making it possible to ∞ o1 where CESO (p = 2 and p = 3) should react more rapidly get an arbitrarily small value of  . than the standard ESO, and red area is the range where only Remark 8: Upon the result (25), we may observe that the CESO p = 2 should provide quicker response with respect increasing gains of both observer and controller are amplify- to control errors. The points at the intersection of ! and ing measurement noise, thus, it is not recommended to use observers graphs indicate the amplification factors of high extremely high values of ! and k in practice. o1 frequency signals (e.g. measurement noise) within signal (t). Consequently, in the following experiments, we can expect IV. HARDWARE EXPERIMENT the measurement noise to be least amplified in CESO p = 3, followed by CESO p = 2, and finally in standard, single ESO. A. Testbed description The experimental setup used for the study is seen in Fig. 3. B. Test methodology The output voltage was measured by a Hall effect-based sensor and converted through a 16-bit A/D converter in the dSPACE The following experiments were conducted to test the platform. The output was recorded by a digital oscilloscope ADRC scheme with the proposed cascade ESO (CESO): and dedicated PC-based software. The sampling period was E1: Comparison with standard ESO (i.e. CESO with p = 1). set to T = 10 Hz. The physical parameters of the DC-DC E2: Influence of parameters ! (E2a), k (E2b), and (E2c). o1 converter, described with (1), were V = 20V, L = 0:01H, in E3: Impact of a low-pass filter (LPF) at the converter output. C = 0:001F, and R = 50 . This allowed to straightforwardly The control objective was to track a smooth voltage tra- calculate the system gain in (3) as b = V ~(CL) = 2 × 10 . in jectory v (t) despite the presence of a varying input-additive The tested control algorithm was first implemented in a external disturbance shown in Fig. 5. Such disturbance signal Matlab/Simulink-based model, from which a C code program is used here to test the robustness of the considered controllers was generated and run on the dSPACE controller in real-time. against different types of disturbances within one experimental Considering the above parameters of the utilized testbed and run. This specific shape of user-injected external disturbance the controller/observer structures introduced in (5), (10), and signal would not appear outside of a laboratory environment, (7), we can derive the transfer-function-based relation however, the character of disturbances designed in specific time intervals can be found in certain applications (e.g. [3]). U (j!) = G (j!) [E(j!) − N (j!)]; (26) uy ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ The reference trajectory was designed as a filtered and biased Y (j!) square signal with bias equal to 7V, amplitude of square signal where U (j!), E(j!), N (j!), and Y (j!) correspond re- equal to 6V, and period 1s. The filtering transfer function spectively to signals (t), e(t), n(t), and y(t) after Laplace applied to the square signal was G (s) = . f 2 0:025s +0:6s+4 transformation. The amplitude Bode diagram of G (j!), Although the most common control task in the control of buck uy obtained for the observer levels p ∈ {1; 2; 3} and tuned with converters is a set-point stabilization, trajectory following of the nominal parameters utilized in the experiment, is presented the output voltage can be occasionally seen in the literature, in Fig. 4. The vertical dashed lines represent the chosen e.g., [24]. Here, we consider a filtered piece-wise constant TABLE I: Used bandwidth parameterization of CESOs. TABLE II: Integral quality criteria for experiment E1. Cascade level Criterion p = 1 p = 2 p = 3 Observer type Bandwidth Se(t)Sdt S(t)Sdt S _ (t)Sdt ∫ ∫ ∫ 1st level ESO (! ) o1 Standard ESO (p = 1) 0.2310 0.5368 315.58 2nd level ESO (! ) – Cascade ESO (p = 2) 0.0467 0.5496 113.23 o2 3rd level ESO (! ) – –  Cascade ESO (p = 3) 0.0381 0.5545 29.11 o3 ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 0.15 Output 0.1 0.05 Reference 0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4 Control error Fig. 5: External disturbance applied in all experiments. reference to reach the desired level of the output voltage and avoid observer peaking caused by the discontinuities in v (t). -1 0 0.5 1 1.5 2 2.5 3 3.5 4 Control signal C. Experimental results The results of E1 are gathered in Fig. 6. The observer 0.5 bandwidth for the standard ESO (p = 1) was ! = 3600rad/s, o1 which was close to the maximum that could be obtained for a 0 0.5 1 1.5 2 2.5 3 3.5 4 10kHz sampling without observing any undesirable effects. Total disturbance For the comparison, only CESOs with p = 2 and p = 3 levels were utilized to maintain legibility of the results while not loosing their generality. In order to provide a systematic tuning methodology across tested observers, bandwidths of the CESOs were parameterized and set according to Table I with 0 0.5 1 1.5 2 2.5 3 3.5 4 = 3 and  = 3600rad/s. The controller gains from (10) were Fig. 6: Results of experiment E1. set to k = 6400 and k = 160 in each case, which corresponds p d to the controller bandwidth k = 80, introduced in Remark 5. One can notice from Fig. 6 that, with the applied tuning with the increase of ! , the amplitude of the control signal o1 methodology, all the tested controllers have realized the given increases but no visible improvement in the control accuracy task, however the standard ESO (p = 1) provided the worst can be observed. In other words, due to multiple factors performance in terms of tracking accuracy and noise sup- like maximum sampling frequency and noise characteristics, pression. On the other hand, with the increase of cascade increasing the observer bandwidth ! will at some point no o1 level p in CESO, better performance was achieved. This longer provide better performance. We can conclude that with observation is supported with the calculated integral quality the CESO one can achieve better control performance for indices in Table II. Besides the improvement of control error wider range of ! values, compared to the results of standard o1 performance, the transfer of sensor noise into the control signal ESO (p = 1) in Fig. 7(a). has decreased with the increase of parameter p thanks to the The results of E2b are depicted in Fig. 8. In the case of lower values of ! related to the first level of CESO. This o1 standard ESO (p = 1), it is clear that increasing the controller result is supported with the values of Su _ (t)Sdt criterion in bandwidth k improves the control accuracy while keeping a Table II, which represents the impact of rapid fluctuations of significant, undesired level of control signal and noise therein. the control signal, mostly caused by the amplified noise. In the case of proposed CESO (p = 2 and p = 3), increasing The initial premises formulated upon Fig. 4 have been the controller bandwidth k results in comparable control errors confirmed with the results in Fig. 6. As expected, the control while retaining similar level of control signal. Due to the signal with the lowest content of noise was obtained for CESO characteristics of CESO, it is possible to obtain better control p = 3, then CESO p = 2, and finally the standard ESO. performance for wider range of k values, compared to the Next, in order to provide potential CESO users with guide- results obtained for the standard ESO in Fig. 8(a). lines for its construction and tuning, the influence of its design The results of E2c are depicted in Fig. 9. In the case parameters was investigated. To this effect, the results of E2 of CESO (p = 2), increasing improves both the tracking are seen in Fig. 7-9. It should be noted that the estimated accuracy and noise suppression in the control signal. However, total disturbance is part of the control signal (see (5)) so its in the case of CESO (p = 3), increasing keeps improving influence is explicitly visible in the control signal. the noise suppression in the control signal but at some point The results of E2a are depicted in Fig. 7. In the case deterioration in the tracking accuracy can be spotted. It results of standard ESO (p = 1), the well-known relation from from a fact that, in this case, the observer bandwidth ! is set high-gain observers, discussed in the Introduction, can be o1 too small, which makes the observer not providing fast-enough noticed. Namely, with the increase of observer bandwidth ! , o1 and accurate-enough estimate of the first state variable of the significant noise amplification occurs in the control signal. At extended state vector. the same time, a slight improvement of the control error was obtained. In the case of proposed CESO (p = 2 and p = 3), Let us now focus on some frequency-domain insights con- ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS cerning experiment E3. An algebraic transformation of (5), 1 1 using (11), allows to write down the form of a generalized 0.8 controller utilizing p-level cascade observer, that is directly 0.5 dependent on the observation error of total disturbance z ~ , 0.6 p3 0 −1 i.e.,  = b (z − z ~ +  ). The transformation of (14) into 1 p3 0 0.4 Laplace domain allows to write that for every p ≥ 1 -0.5 0.2 -1 0 Z (j!) = G (j!)N (j!) + G (j!)Z (j!); (27) p3 z~ n z~ z 3 p3 p3 3 024 024 (a) standard ESO (p = 1) ~ where Z (j!) and Z (j!) are the Laplace-domain equiva- p3 3 lents of signals z ~ (t) and z (t). Application of the LPF p3 3 1 1 G = 1~(s + 1);  > 0; (28) LPF 0.8 0.5 0.6 0.4 at the output of the converter (in order to filter-out measure- -0.5 ment noise) affects the total disturbance signal with a filtered- 0.2 out parts of the measured signal z and results in a following -1 0 024 024 extended form of (27), i.e., (b) CESO (p = 2) Z (j!) = G (j!)N (j!) + G (j!)Z (j!) p3 z~ n z~ z 3 p3 p3 3 1 1 + G (j!)Z (j!); (29) z~ z 1 p3 1 0.8 0.5 where Z (j!) corresponds to signal z (t) after Laplace 1 1 0.6 transformation. According to [19], the high-gain extended 0.4 observer performance should not be substantially affected for -0.5 0.2 small enough values of time constant  of the low-pass filter. -1 0 We assume that  has been chosen appropriately, and hence 024 024 focus on the noise-connected characteristics of the ADRC with (c) CESO (p = 3) analyzed observers. The amplification of particular frequencies of the measurement noise using ESO and CESO (p = 2; 3) with Fig. 7: Results of experiment E2a. parameters = 3 and  = 3600rad/s (see Table I) has been presented in Fig. 10. The dashed lines represent the magnitude of G when a low-pass filter was applied while the regions z~ n p3 1.5 1 with corresponding colors illustrate the set of characteristics 0.8 that would be obtained for a practically useful set of values 0.5 0.6 ∈ [0:0001; 0:01]s, where the bottom edge corresponds to 0.4 = 0:01s and the top edge corresponds to  = 0:001s. -0.5 0.2 Looking at Fig. 10, one can notice that the maximal value of -1 YG Y for CESO (p = 3) without output filtering was similar, -1.5 0 z~ n p3 024 024 or smaller, compared to the characteristics obtained with the (a) standard ESO (p = 1) conventional ESO with LPF with  = 0:001s, so the expected 1.5 1 content of the measurement noise in signal z ~ affecting the p3 0.8 control signal should be similar, or lower. This observation was 0.5 0.6 validated by time-domain results of experiment E3, presented 0.4 in Fig. 11, where the amplitude of noise-dependent oscillations -0.5 0.2 -1 is  ≈  ≈ 0:05. The presented values of the control ESO+LPF CESO -1.5 0 error illustrate the essential difference in the measurement 024 024 noise handling by the CESO, compared to the use of a LPF. (b) CESO (p = 2) The proposed cascade observer structure suppresses the effect 1.5 1 of measurement noise amplification in the control signal but 0.8 does not change the noise level at the output, while the use of 0.5 0.6 0 a LPF decreases the level of measurement noise at the output 0.4 -0.5 but does not change the noise amplification feature of the high- 0.2 -1 gain ESO. In order to improve the overall performance of the -1.5 0 control system, in terms of robustness against measurement 024 024 noise, a LPF can be utilized along CESO. Such example is illustrated in Fig. 11, where the combination of CESO and (c) CESO (p = 3) LPF achieves the amplitude value  ≈ 0:02, which is CESO+LPF Fig. 8: Results of experiment E2b. smaller than the aforementioned  and  . ESO+LPF CESO ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] [-] [-] [-] [-] [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 1.5 1 1 1 0.8 0.8 0.5 0.5 0.6 0.6 0 0.4 0 0.4 -0.5 0.2 0.2 -0.5 -1 0 024 024 1 024 024 (a) CESO p = 2 0.82 1.5 1 1 0.8 0.78 0.5 0.6 0.74 0 0.4 0.46 0.5 0.54 0.58 0.62 -0.5 0.2 -1 0 Fig. 11: Results of experiment E3. 024 024 (b) CESO p = 3 Development, non-competitive project entitled “International Fig. 9: Results of experiment E2c. scholarship exchange of PhD students and academic staff” executed under the Activity 3.3 specified in the application for funding of project No. POWR.03.03.00-00-PN13 / 18. The work has been also supported by the Fundamental Research Funds for the Central Universities project no. 21620335. REFERENCES [1] R. Madonski, K. Łakomy, M. Stankovic, S. Shao, J. Yang, and S. Li, “Robust converter-fed motor control based on active rejection of multiple disturbances,” Control Engineering Practice, vol. 107, DOI 1 2 3 4 5 6 7 10 10 10 10 10 10 10 https://doi.org/10.1016/j.conengprac.2020.104696, p. 104696, 2021. [2] J. Yang, H. Wu, L. Hu, and S. Li, “Robust predictive speed regulation of converter-driven DC motors via a discrete-time reduced-order GPIO,” Fig. 10: Bode diagram representing the module of G (j!). z~ n p3 IEEE Trans. on Ind. Electronics, vol. 66, no. 10, pp. 7893–7903, 2019. [3] J. Yang, H. Cui, S. Li, and A. Zolotas, “Optimized active disturbance rejection control for DC-DC buck converters with uncertainties using a reduced-order GPI observer,” IEEE Transactions on Circuits and In order to summarize the results obtained in this work Systems I: Regular Papers, vol. 65, no. 2, pp. 832–841, 2018. and allow for their quick assessment, Table III compares the [4] J. Han, “From PID to active disturbance rejection control,” IEEE standard ESO with the proposed CESO using selected criteria. Transactions on Industrial Electronics, vol. 56, no. 3, pp. 900–906, 2009. [5] S. Chen and Z. Chen, “On active disturbance rejection control for a class of uncertain systems with measurement uncertainty,” IEEE Transactions V. CONCLUSIONS on Industrial Electronics, vol. 68, no. 2, pp. 1475–1485, 2021. [6] E. Sariyildiz, R. Oboe, and K. Ohnishi, “Disturbance observer-based An active disturbance rejection control with a novel cascade robust control and its applications: 35th anniversary overview,” IEEE extended state observer (CESO) for DC-DC buck converters Trans. on Industrial Electronics, vol. 67, no. 3, pp. 2042–2053, 2020. has been proposed. The validity of the new approach has [7] W. H. Chen, J. Yang, L. Guo, and S. Li, “Disturbance-observer-based control and related methods - an overview,” IEEE Transactions on been shown through a dedicated stability analysis and a Industrial Electronics, vol. 63, no. 2, pp. 1083–1095, 2016. set of hardware experiments. The comparison between the [8] K. Łakomy, R. Patelski, and D. Pazderski, “ESO architectures in the proposed cascade ESO-based ADRC and a standard single trajectory tracking ADR controller for a mechanical system: a compar- ison,” in Advanced, Contemporary Control, pp. 1323–1335, 2020. ESO-based ADRC showed that the former has stronger capa- [9] H. K. Khalil and L. Praly, “High-gain observers in nonlinear feedback bilities of sensor noise suppression and provides better control control,” International Journal of Robust and Nonlinear Control, vol. 24, performance (understood as tracking accuracy and energy no. 6, pp. 993–1015, 2014. [10] S. Sugahara and S. Matsunaga, “Fundamental study of influence of ripple efficiency). The structure of the proposed ADRC is bulkier noise from DC–DC converter on spurious noise of wireless portable than the conventional one but in return provides an additional equipment,” IEEE Transactions on Power Electronics, vol. 31, no. 3, pp. 2111–2119, 2016. and practically appealing degree of freedom in shaping the [11] A. A. Prasov and H. K. Khalil, “A nonlinear high-gain observer for influence of measurement noise on the observer/controller part. systems with measurement noise in a feedback control framework,” IEEE Trans. on Automatic Control, vol. 58, no. 3, pp. 569–580, 2013. [12] S. Battilotti, “Robust observer design under measurement noise with VI. ACKNOWLEDGMENT gain adaptation and saturated estimates,” Automatica, vol. 81, pp. 75– 86, 2017. The article was created thanks to participation in program [13] W. Xue, X. Zhang, L. Sun, and H. Fang, “Extended state filter based PROM of the Polish National Agency for Academic Ex- disturbance and uncertainty mitigation for nonlinear uncertain systems change. The program is co-financed from the European Social with application to fuel cell temperature control,” IEEE Transactions on Fund within the Operational Program Knowledge Education Industrial Electronics, vol. 67, no. 12, pp. 10 682–10 692, 2020. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. [-] [-] [-] [-] This is the author’s version of an article that has been accepted for publication in this journal. Changes will be made to this version by the publisher prior to final publication. The official version of record is available at http://dx.doi.org/10.1109/TIE.2021.3055187 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS TABLE III: Comparison between standard ESO- and proposed CESO-based control with selected criteria. Observer type Standard ESO (p = 1) Cascade ESO (p = 2) Cascade ESO (p = 3) Criterion Observer tuning methodology parameterization [22] parameterization [22] + Table I ^ ^ ^ Design parameters k;b;! k;b;! , k;b;! , o1 o1 o1 Total disturbance estimation quality (Fig. 6) — ↗ ↗↗ Control quality (« Se(t)Sdt in Table II) — ↗ ↗↗ Control effort (« S(t)Sdt in Table II) — ↗ ↗↗ Control jittering (« S _ (t)Sdt in Table II) — ↘ ↘↘ Noise content in control signal with output LPF (Fig. 10) ↘ ↘ ↘ Implementation complexity (no. of observer state variables) 3 6 9 Stability type (nominal conditions) asymptotic [5], [18] asymptotic (Remark 6 in Sect. III) Stability type (non-nominal conditions) practical [5], [18] practical (Theorem 1 in Sect. III) [14] Y. Wu, A. Isidori, and L. Marconi, “Achieving almost feedback- [20] Y. Huang and W. Xue, “Active disturbance rejection control: Method- linearization via low-power extended observer,” IEEE Control Systems ology and theoretical analysis,” ISA Transactions, vol. 53, no. 4, pp. Letters, vol. 4, no. 4, pp. 1030–1035, 2020. 963–976, 2014. [15] L. Wang, D. Astolfi, L. Marconi, and H. Su, “High-gain observers [21] R. Madonski, S. Shao, H. Zhang, Z. Gao, J. Yang, and S. Li, “Gen- with limited gain power for systems with observability canonical form,” eral error-based active disturbance rejection control for swift indus- Automatica, vol. 75, pp. 16–23, 2017. trial implementations,” Control Engineering Practice, vol. 84, DOI [16] H. K. Khalil, “Cascade high-gain observers in output feedback control,” 10.1016/j.conengprac.2018.11.021, pp. 218–229, 2019. Automatica, vol. 80, pp. 110–118, 2017. [22] Z. Gao, “Scaling and bandwidth-parameterization based controller tun- [17] D. Astolfi, M. Jungers, and L. Zaccarian, “Output injection filtering ing,” in American Control Conference, vol. 6, pp. 4989–4996, 2003. redesign in high-gain observers,” in Proc. European Control Conference, [23] H. K. Khalil, Nonlinear systems, 3rd edition. Prentice Hall, 2002. pp. 1957–1962, 2018. [18] K. Łakomy and R. Madonski, “Cascade extended state observer for [24] R. Silva-Ortigoza, V. M. Hernandez-Guzman, M. Antonio-Cruz, and active disturbance rejection control applications under measurement D. Munoz-Carrillo, “DC/DC buck power converter as a smooth starter noise,” ISA Transactions, DOI 10.1016/j.isatra.2020.09.007, 2020. for a DC motor based on a hierarchical control,” IEEE Transactions on [19] H. K. Khalil and S. Priess, “Analysis of the use of low-pass filters with Power Electronics, vol. 30, no. 2, pp. 1076–1084, 2015. high-gain observers,” IFAC, vol. 49, no. 18, pp. 488–492, 2016. ©2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Journal

Computing Research Repository – arXiv (Cornell University)

Published: Sep 7, 2020

There are no references for this article.