Inter national J our nal of P o wer Electr onics and Dri v e Systems (IJPEDS) V ol. 12, No. 2, Jun 2021, pp. 858 869 ISSN: 2088-8694, DOI: 10.11591/ijpeds.v12.i2.pp858-869 858 An efcient pr edicti v e curr ent contr oller with adapti v e parameter estimation in 3- Φ in v erter Haddar Mabr ouk, Allaoua Boumediene Smart Grids & Rene w able Ener gy (SGRE) Laboratory , Uni v ersity of T ahri Mohammed, Bechar , Algeria Article Inf o Article history: Recei v ed Dec 1, 2020 Re vised Feb 14, 2021 Accepted Mar 19, 2021 K eyw ords: Filter Grid-connected In v erter MRAS observ er Predicti v e control ABSTRA CT In this paper , a detail design and descri ption of a predicti v e current control scheme are adopted for three-phase grid-connected tw o-le v el in v erter and its application in wind ener gy con v ersion sys tems. Despite its adv antages, the predicti v e current controller is v ery sensiti v e to parameter v ariations which could e v entually af fected on system sta- bility . T o solv e this problem, an estimation technique proposed to identify the v alue of harmonic lter parameter based on Model reference adapti v e system (MRAS). L ya- puno v stability theory is selected to guarantee a rob ust adaptation and stable response o v er lar ge system parameter v ariation. The simulation results sho ws the ef cienc y of the proposed techniques to impro v e the current tracking performance. This is an open access article under the CC BY -SA license . Corresponding A uthor: Haddar Mabrouk Department of Electrical Engineering SGRE Laboratory , Uni v ersity of T ahri Mohammed, Bechar , Algeria B.P 417 route k enadsa 08000, Bechar , Algeria Email: haddar .mabrouk@uni v-bechar .dz 1. INTR ODUCTION In recent years, the grid-connected in v erters ha v e made a giant strides in the v arious applications of industries and rene w able ener gies , that embody by introduce a ne w concepts of adv anced control in detriment to other con v entional controls. The grid-connected in v erters is widely used in wind ener gy systems. The simplied structure of the grid-connected in v erters is illustrated in Figure 1. Where, it’ s possible to replace the maximum amount of po wer that can be withdra wn from the wind turbine, generator , and rectier o wed through the in v erter by adequate v ariable DC-current source without causing an y damage in system caracteristic [1]-[6]. (a) (b) Figure 1. Grid-connected in v erter in a wind ener gy concersion system, (a) General structure, (b) Simplied structure J ournal homepage: http://ijpeds.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Po w Elec & Dri Syst ISSN: 2088-8694 859 On the other side, The progress of the micro processor unit allo wed the possibility to impro v e the controllers design with v ery short sampling time and good performance. The model predicti v e control ( MPC) is one of the most important miles tones of de v elopment, as it recently occupied a lar ge part of studies and the researchers’ ef fort as promise control methods because of its: (1 simple theory concept, (2 easy to im- plement (3 e xible and promising for digital control, (4 f ast dynamic response, [2] ,[3]. Ho we v er , despite of its adv antages, the MPC is sensi ti v e to the system parameter v ariations during the operating conditions. This results of perturbations inuences on the o v erall stability of the system, which mak es MPC need to adaptation or estimation algorithm [7]. The MPC applied for po wer con v erters and electrical dri v es include tw o main cate gories: Continuous Control Set MPC with a modulation stage and Finite Control Set MPC (FCS-MPC) without a modulation stage [8], [9]. In FCS-MPC, the control objecti v es in cost function may be either main or additional [7]-[11], the rst one using ph ysics quantities such as current, v oltage, po wer , torque, etc. It e v alu- ates error tracking between an y predicted v ariable v alue and its reference. While the second one is considering as sub-objecti v es control for e xample switching frequenc y reduce. The additional control objecti v e multiplying by suitable v alues of weighting f actors witch’ s reecting its importance in optimization criteria. The algorithm of FCS-MPC minimizes the cost function and applies the generated switching signal at end each the sampling time directly on po wer con v erter [11], [12]. The rob ustness of FSC-MPC ag ainst parameter v ariations still major concern in man y studies [12]. Addressing the issue, a man y control strate gies ha v e been proposed to estimate the parameters v ariable in po wer con v erter applications. In [9], an analytical approach is proposed to e xam the inuence of model parametric uncertainties on the prediction of FSC-MPC error for current control in three phase tw o-le v el in v erter . In [13], the authors proposed adapti v e observ er to accurately estimate tw o v ariable: speed and ux, this observ er based encoderless FCS-PTC (predicti v e torque control) helps k eeping the stability for induction machine. The Model Reference Adapti v e System (MRAS) observ er is used in [14], where the authors estimate line inductance to impro v e rob ustness for sensorless predicti v e control of Acti v e front End (AFE) rectiers. In [15], a control scheme based on a computationally nite-set model predicti v e po wer control (FS-MPPC) for grid-connected photo v oltaic systems is proposed, the stability of (FS-MPPC) ag ainst inductance v ariation is v eried by means of no v el online nite-set model inductance estimati on technique. In [16], an e xtended Kalman lter (EKF) is proposed to estimate the system parameter ( lter grid impedance) for AFE. Ho we v er , the MRAS technique has been appro v ed to be a po werful tool for parameter estimation, the MRAS is widely used in po wer electronics application and motor dri v es system to mainly estimate: (1 the lter parameter [16], (2) the machine state v ariables (speed, ux,...) for sensorless dri v e systems [18], [19],...etc, b ut there are limited researches based on MRAS for grid-connected in v erter [16]. The MRAS consists of tw o models, the reference one and the adapti v e one. The dif ference between the outputs of these tw o models is then used in an adaptation mechanism such as lyapuno v theory to adjust the parameters in the adapti v e model until the response of the main tw o models become consistent and the tracking error con v er ges to zero [19], [20]. On this light, the performance of predicti v e current controller for the On this light, the perf ormance of predicti v e current controller for the three-phase grid-connected tw o-le v el in v erter entirely depends on its mathematical model as wel l as the accurac y of the parameters that may assumed in theory constant and kno wn b ut pract ically not so. In this paper , the lter parameters may easily af fected due to the heat risen, magnetic saturation and system lifetime, etc. This paper presents a structure that combines tw o types of controller; the rst one is a FCS-MPC witch generate a switching states for t he grid-connected in v erter b ut it is inuenced by parameters v ariation, this could yield ne g ati v e results as inaccurate prediction of the fut ure beha vior due to an incorrect system model and so the selection of the i ncorrect switching states. The second is a classic which relies on the MRAS observ er to cope with the aforementioned dra wback through a parameter estimation method based on L yapuno v stability theory . Therefore, it is possible to a v oid the deterioration of the control performance. There are a fe w researches report on the applic ation the MRAS observ er to grid-connected in v erter [17]. So, this no v el structure brings great adv antages such as simple control scheme, No inner control loop, and No modulator . Additionally , although the proposed estimation t echnique is classic, it can mak e the system operate rob ustly and independently from mathematical model. This paper is arranged as follo ws: Section (2 discusses the system structure and modeling. Section (3 and section (4, present the theory behind the v oltage oriented v ector and the predicti v e current control strate gy , respecti v ely . The identication of the lter parameter based on MRAS observ er are presented is gi v en in Section (5. Section (6 contains The simulation results and analysis, and nally conclusions are discussed in Section (7. An ef cient pr edictive curr ent contr oller with adaptive par ameter estimation... (Haddar Mabr ouk) Evaluation Warning : The document was created with Spire.PDF for Python.
860 ISSN: 2088-8694 2. SYSTEM STR UCTURE AND MODELING The con v entional po wer circuit of a three-phase grid-connected tw o-le v el in v erter applied to wind ener gy is modeled as sho wn in Figure 2. Where, a harmonic lter ( r g , L g ) tak es place between the grid and the in v erter . The models of the three-phase in v erter and grid current dynamic are briey described in the follo wing subsections.. 2.1. In v erter model The tw o-le v el in v erter is composed of six bidirectional switches. Since there are three phases with tw o operating modes in each phase, the in v erter is able to generate only 2 3 = 8 states as dif ferent possible output v oltages. The operating modes of the in v erter are summarized as follo ws [2]: S x + ¯ S x = 1 for x { a, b, c } (1) Figure 2. T opology of grid-connected tw o-le v el in v erter in wind ener gy con v ersion systems By means of the space v ector tool, The output v oltage generated by t he in v erter can e xpressed in terms of the dif ferent switching states by v αi v β i = V dc 2 3 1 1 2 1 2 0 3 2 3 2 S a S b S c (2) Where V dc is DC link v oltage, v αi and v β i are α and β component of the in v erter output v oltage v ectors. Figure 3 sho ws the eight output in v erter v ol tage v ectors in space v ectors corresponding the all possible switching states which including six dif ferent acti v e v oltage v ectors ( V 1 to V 6 ) and tw o other zero v ectors ( V 0 and V 7 ). Figure 3. Space v ectors generated by the in v erter The in v erter v oltage component e xpressed in the dq -frame rotating at angle θ g can be related to the α β component by v di v q i = cos θ g sin θ g sin θ g cos θ g v αi v β i (3) Where v di and v q i are dq -axis components of the in v erter output v oltage. Int J Po w Elec & Dri Syst, V ol. 12, No. 2, Jun 2021 : 858 869 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Po w Elec & Dri Syst ISSN: 2088-8694 861 2.2. Grid curr ent dynamic model The continuous-time grid current dynamic equation of three-phase grid-tied in v erter can be written in the dq -frame as [1]-[4] di dg dt = r g L g i dg + ω g i q g + 1 L g ( v di v dg ) di q g dt = ω g i dg r g L g i q g + 1 L g ( v q i v q g ) (4) Where v dg and v q g are dq -axis components of the grid v oltage and ω g is the angular fraquenc y of the grid, r g lter resistance and L g is lter inductance. The grid acti v e and reacti v e po wers are e xpressed by [19]: P g = 3 2 { v dg i dg + v q g i q g } Q g = 3 2 { v q g i dg v dg i q g } (5) 3. V OL T A GE ORIENTED VECT OR (V OC) In the literature, the phase-lock ed loop (PLL) is needed not only to nd the grid v oltage angle θ g for the grid synchronization, b ut also to realize the v oltage oriented control (V OC). Where, the dq -axis components of the grid v oltage is oriented to be align only with its d-axis component and eliminate the second component. That means, the acti v e po wer equation and reacti v e po wer equation are simplied to [20] P g = 3 2 v dg i dg Q g = 3 2 v dg i q g (6) As described pre viously , It is noted that the d -axis current reference is i dg related to the grid acti v e po wer . It can be adjusted through PI controller which maintains the measured DC-link v oltage v dc at its gi v en reference v alue v dc , we can write i dg = ( k p + k i s )( v dc v dc ) (7) The q -axis current reference i q g can be calculated from reference reacti v e po wer of the grid Q g as i q g ( k ) = Q g 1 . 5 v dg (8) 4. PREDICTIVE CURRENT CONTR OL (PCC) According to the 8 switching possible states for tw o-le v el in v erter , the predicti v e current control ler e xploit the discrete-time model with one-step for grid-connected in v erter to predict at the ne xt instant the future beha vior of the controlled grid curr ent. It consists of thr ee major subsystems: e xtrapolation of reference currents, predicti v e model, and cost function minimization. At each sampling time, cost function v alues are calculated for all of possible commutation states, based on predened crit erion. Then, the smallest v alue of the cost function will be used to determine the optimal switching state applied to the in v erter at the ne xt period [2]. Approximating the grid current deri v ati v es by [1]-[3] di g dt i g ( k + 1) i g ( k ) T s (9) An ef cient pr edictive curr ent contr oller with adaptive par ameter estimation... (Haddar Mabr ouk) Evaluation Warning : The document was created with Spire.PDF for Python.
862 ISSN: 2088-8694 The discrete-time model on dq -frame of grid currents at ( k + 1) state is e xpressed as follo ws: i dg ( k + 1) i q q ( k + 1) = 1 r g T s /L g 0 0 1 r g T s /L g i dg ( k ) i q g ( k ) + L g 0 0 L g v di ( k ) v dg ( k ) v q i ( k ) v q g ( k ) (10) Ho we v er , the future reference of the grid current i g ,dq ( k + 1) can be estimated by means of current and pre vious v alue of the reference current with the help of second-order Lagrange e xtrapolation as follo ws [2], [3]. i dg ( k + 1) i q q ( k + 1) = 2 i dg ( k ) i q g ( k ) i dg ( k 1) i q q ( k 1) (11) In t his paper , the rst tar get of the predicti v e current controller is to achie v e the smallest current error between the e xtrapolated reference currents and the predicted current. The second tar get is to reduce signicantly the switching frequenc y by adjusting the commutation number between tw o successi v e sampling instants. In situations where the switching loses are important. In brief, These all objecti v es can be e xpressed in the form of a cost function g to be mi nimized. The cost function summarizes the desired beha vior of the in v erter can be obtained using the error squared such as : g = i dg ( k + 1) i dg ( k + 1) 2 + i q g ( k + 1) i q g ( k + 1) 2 + λ sw ( S ( k ) S ( k 1)) 2 (12) Where λ sw is weighting f actors for switching frequenc y reduction, S ( k ) and S ( k 1) are the present switching state and past applied switching state respecti v ely . Figure 4 sho ws the proposed predicti v e current controller for the grid-connected tw o-le v el in v erter . In this scheme, getting the angle θ g from the measured grid v oltage is done via PLL method to mak e v dg equal to V g and also to transforme the v ariables measured ( v g and i g ) from abc -frame to dq -frame. Once reference grid currents i g ,dq ( k ) is calculated, the PCC procedure is established to select the optimal switching signal applied to the in v erter . Moreo v er , the scheme in v olv e s the MRAS observ er to impro v e the rob usrness of PCC by identifying the lter parameter . Figure 4. Proposed predicti v e current controller scheme for grid-connected tw o-le v el in v erter . 5. IDENTIFICA TION OF THE HARMONIC FIL TER P ARAMETERS B ASED ON MRAS In the real time, Both v alues of harmonic lter parameters (inductance and resistance) link ed between the in v erter and the grid are changed under the operating conditions that reduce system ef cienc y in term of current quality . So that, these parameters are estimated based on the model reference adapti v e system (MRAS). Int J Po w Elec & Dri Syst, V ol. 12, No. 2, Jun 2021 : 858 869 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Po w Elec & Dri Syst ISSN: 2088-8694 863 Figure 5 presents the basic struct ure of MRAS, wich consi sts of three parts: a reference model , adapti v e model, and an adapti v e controller . Figure 5. Basic structure of MRAS The equation of state for reference model: (4) d i g dt = A i g + B u (13) Where i g = i dg i q g T , A = α ω g ω g α , α = r g L g , B = 1 L g , u = u d u q T , u d = v di v dg , u q = v q i v q g The adapti v e model is gi v en as follo ws: d ˆ i g dt = A ˆ i g + B u (14) Where ˆ i g = ˆ i dg ˆ i q g T , ˆ A = ˆ α ω g ω g ˆ α , ˆ α = ˆ r g ˆ L g , ˆ B = 1 ˆ L g , u = u d u q T i g and ˆ i g are the output current of reference model and adapti v e model respecti v ely . The measured grid current is tuning as the reference model witch is strongly reliant on the rated lter parameters ( r g and L g ), and the estimated grid current is tuning as the adapti v e model witch uses a v alue of estimated lter parameters ( ˆ r g and ˆ L g ), and then, the output current error of tw o models is used by adapti v e controller to adjust the parameter in adapti v e controller until the identication becomes asymptotically stable and the current error approaches zero as small as possible. The current error is dened as e = i g ˆ i g = [ e d e q ] T , the error state equation can be written: ˙ e = A e + ( A ˆ A ) ˆ i g + ( B ˆ B ) r (15) Letting aI = ( A ˆ A ) , b = ( B ˆ B ) = 1 L g 1 ˆ L g , Z = [ ˆ i g u ] T and T = [ a b ] T The (15) can be written as ˙ e = A e + T Z (16) Considering a L yapuno v function candidate as follo w: V ( e , t ) = 1 2 ( e T P e + T Z ∆) (17) Where P = I 2 × 2 and A = λ 1 0 0 λ 2 1 λ 1 and λ 2 are arithmetic number . Starting from L yapuno v stability concept, the L yapuno v function must respond the follo wing three necessary conditions to ensure that the identication is asymptotically stable [22]-[25]: 1) V ( e , t ) > 0 , An ef cient pr edictive curr ent contr oller with adaptive par ameter estimation... (Haddar Mabr ouk) Evaluation Warning : The document was created with Spire.PDF for Python.
864 ISSN: 2088-8694 2) ˙ V ( e , t ) < 0 , 3) V ( e , t ) 8 as | e | 8 . It is e vident that the rst and third condit ions are met. The second condition can be discussed as follo ws: ˙ V ( e , t ) = 1 2 ( ˙ e T P e + e T P ˙ e + ˙ T Z + T Z ˙ ∆) = 1 2 ( e T ( P A + P A T ) e + a ( ˆ i d e d + ˆ i q e q + ˙ a λ 1 ) + b ( u d e d + u q e q + ˙ b λ 2 ) (18) F or P A + P A T = 2 α I < 0 it is clear that e T ( P A + P A T ) is ne g ati v e denite. The term ˙ V ( e, t ) must be ne g ati v e denite, this yields ˙ a λ 1 + ˆ i dg ( i dg ˆ i dg ) + ˆ i q g ( i q g ˆ i q g ) = 0 (19) ˙ b λ 2 + u d ( i dg ˆ i dg ) + u q ( i q g ˆ i q g ) = 0 (20) The adapti v e la w can be easily e xpressed as follo wing: ˆ r g ˆ L g = r g L g λ 1 Z t 0 n ˆ i dg ( i dg ˆ i dg ) + ˆ i q g ( i q g ˆ i q g ) o dt (21) 1 ˆ L g = 1 L g + λ 2 Z t 0 n u d ( i dg ˆ i dg ) + u q ( i q g ˆ i q g ) o dt (22) 6. SIMULA TION RESUL TS T o e v aluate the ef cienc y of the considered predicti v e current control model with MRAS oberv er for grid connected tw o-le v el in v erter under v arious v alues of parameter lter , the whole simulation studies are implemented by means of MA TLAB/simulink tools. The main parameters are indicated in T able 1. T able 1. Simulation parameters P arameter Nomenclature V alue Rated Line-to-Line V oltage (rms) [V] V g 690 Rated current (rms) [A] I g 627.6 Rated acti v e po wer [kw] P g 750 DC-link v oltage [V] V dc 1220 DC-link capacitor [ µ F] C dc 16714 Filter resistance [m ] r g 95.25 Filter inductance [mH] L g 0.3368 Sampling time [ µ s] f s 20 Frequenc y of the grid [Hz] f g 50 The results presented in three scenario: Ideal Case operation: In this case , the lter parameter are setting on rated v alues Non-ideal Case operation without MRAS: In this case , the parameter lter are changed to sho w the inuence of parameter v ariation on whole performance of system. Non-ideal Case operation with MRAS: In this case , the MRAS observ er is used to estimate t he lter parameter and to enhance the performance of system Int J Po w Elec & Dri Syst, V ol. 12, No. 2, Jun 2021 : 858 869 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Po w Elec & Dri Syst ISSN: 2088-8694 865 6.1. Ideal case operation The performance of the proposed predicti v e current controller scheme for three-phase gri d-connected in v erters considering the rated v alue of lter parameter is tested by applying a DC-link v oltage reference V dc = 1220 V , and a zero reference po wer reacti v e Q g = 0 to unitary the po wer f actor . Figure 6 (a) sho ws the weighting f actor impact λ sw on the grid current quality via T otal Harmonic Distortion (THD) and the a v erage switching frequenc y f sw . It’ s clearly noted when the weighting f actor is increasing, the of the grid current is changed progress i v ely from 1.94 to 3.57 %, while the a v erage switching frequenc y is reduced considerably from 8.86 to 3.56 [KHz] . F or this reason, the weighting f actor can be adjusting at 1700 for operating the in v erter at f sw = 3.703 [KHz] and for ha ving a suitable v alue of THD at 3.30 %. (a) (b) (c) (d) Figure 6. Simulation results for case ideal operation, (a) THD % and f sw v ersus λ sw v ariation, (b) DC-link v oltage [V], (c) dq -axis grid current [A] (peak), (d) w a v eforms of phase-a grid current i ag [A] (peak) . Figure 6 (b) sho ws the obtained simulation result where the DC-link v oltage is controlled to maint ain at its reference v alue 1200 V and it is stable during steady state operation. This simulated response is obtained with a PI controller ha ving a k p = 6 . 90 and a k i = 1 . 952 10 3 . The dq -axis grid current components either the reference i g ,dq or the measured i g ,dq are tracking each other perfectly as sho wn in Figure 6 (c). Consequently , The amplitude of the reference d -axis grid current component i dg created from the v oltage PI controller is equal to 887.5 A (peak). In addition, the grid-connected in v erter is control led to supply zero reacti v e po wer corresponding the zero reference q -axis grid current. Figure 6 (d) illustrates the w a v eform of a-phase grid current, where the peak v alue of i ag is equal to i dg . 6.2. Non-ideal case operation without MRAS It’ s the case where the v ariation in lter parameter has been considered as follo ws: Filter resistance and inductance are 50 % lo wer than its nominal v a lue Filter resistance and inductance is also 100 % of its nominal v alue Filter resistance and inductance are 150 % higher than its nominal v al ue These v alues are applying at the follo wing moments 0.1 and 0.2 (s) respecti v ely . Figure 7 sho ws the ef fect of parameters v ariation of both r g and L g on the o v erall performance of the controller as well as its stability . Under these conditions, it is clear to note the tracking errors between the measured v alues of i dg , i q g , and V dc and their references, which leads to unbalanced v alue of THD especially when both the resistance and the inductance are higher than the nominal v alues as presented in Figure 8. An ef cient pr edictive curr ent contr oller with adaptive par ameter estimation... (Haddar Mabr ouk) Evaluation Warning : The document was created with Spire.PDF for Python.
866 ISSN: 2088-8694 (a) (b) Figure 7. Simulation results for case non-ideal operation without MRAS, (a) DC-link v oltage [V], (b) dq -axis grid current [A] (peak). (a) (b) (c) Figure 8. Simulation results e xplains the inuence of lter parameters v ariation on the performance of PCC r g and L g are: (a) 50 % lo wer than its nominal v alue , (b) aquals to nominal v alue, (c) 150 % higher than its nominal v alue. 6.3. Non-ideal case operation with MRAS Figure 9 sho ws the dynamics performance of PCC scheme when the MRAS observ er based on lya- puno v function is enable under parameters v ariation. The adapti v e g ains are tuned using a tuning method (trial and error) to gi v e suitable performance. After some tuning, the adapti v e g ains are setting as follo wing: λ 1 = 0 . 625 and λ 2 = 6 . 795 . the simulation results consists of tw o parts: in the rst one, the lter parameters are increasing and switching from 50 % of its nominal v alue to nominal v alue, and in the second one the lter parameters are decreasing and switching from 150 % of its nominal v alue to nominal v al ue. Int J Po w Elec & Dri Syst, V ol. 12, No. 2, Jun 2021 : 858 869 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Po w Elec & Dri Syst ISSN: 2088-8694 867 (a) (b) Figure 9. Simulation results for R L lter parameters estimation based on MRAS on serv er (a) starts from 50 % of its nominal v alue to nominal v alue; (b) sta rts from 150 % of its nominal v alue to nominal v alue The switching is applied at the moment 0.15 (s). In the both simul ation, the MRAS observ er track and mak e the estimated v alues con v er ges to parameter v ariation. Moreo v er , the estimated current ˆ i g and the mea- sured current i g tracks each other . Despite lar ge error at the same time where the switching is applied. the error current e con v er ges to zero. It is w orth noting that the estimate v alue of the lter resistance is more accurate than the estimated v alue of the lter inductance with steady-state estimation error in range of 1 . 763 10 4 and 0 . 1 % respecti v ely . T o sho w the rob ustness of PCC controller with a MRAS observ er for a three-phase grid-connected in v erters, a comparison of performance can be done through tw o cases: The rst is in agreement with the second scenario, where the lter parameters are changed in the absence of a MRAS observ er , while PCC controller is operating on the normal v alues. As for the second case, which corresponds to the third scenario, the change of the lter parameters are in the presence of the MRAS observ er , which estimates the lter parameters so that the PCC controller uses them. The comparison w as done by measuring THD and f sw , as sho wn in Figure 10. Although the lter parameters are changed, the MRAS observ er helps the PCC controller to select a switching states that corresponds to the minimal v alues of f sw with slight dif ference in THD. (a) (b) Figure 10. Comparison of system performance by measuring; (a) THD %; (b) f sw [KHz]. An ef cient pr edictive curr ent contr oller with adaptive par ameter estimation... (Haddar Mabr ouk) Evaluation Warning : The document was created with Spire.PDF for Python.