A model predictive control method for a dual-module active power filter parallel system

Abstract

The application discloses a model prediction control method of a dual-module active power filter parallel system in the technical field of active power filtering, calculates a compensation current prediction value based on 64 synthesized voltage vectors and three-phase grid voltages and three-phase compensation currents, calculates 64 evaluation values through evaluation of a value function, and controls two inverters according to a switch state corresponding to a synthesized voltage vector corresponding to a minimum value in the 64 evaluation values; the active power filter based on the parallel three-phase inverters of the common DC bus is discretely mathematically modeled as a six-bridge-arm inverter system, the three-phase harmonic compensation currents of the two inverters are predicted, and the 64 synthesized voltage vectors of the parallel three-phase inverters are iteratively optimized through evaluation of the value function, so that the global optimization of the circulating current between the two inverters is realized, the circulating current of the parallel inverters is effectively inhibited, and the harmonic compensation accuracy and system efficiency of the active power filter are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of active power filter in power electronics, and particularly relates to a model predictive control technology of an active power filter parallel system. BACKGROUND

[0002] The wide use of power electronic devices in the power generation, power transmission, power distribution and power consumption links of a power system leads to the increasingly serious harmonic pollution problem of a public power grid. Compared with a passive power filter, an active power filter becomes a mainstream solution for improving the power quality of a modern power system due to its high steady-state accuracy, fast dynamic response, strong adaptability to power grid parameter fluctuation, good system stability, small device size and many other advantages. According to different system constitution modes, the active power filter is mainly divided into a series type, a parallel type and a hybrid type, among which the parallel type active power filter has the advantages of convenient access to a power grid and small influence on the power grid, and has obtained the most extensive industrial application.

[0003] With the large access of power electronic devices and other nonlinear loads to a power grid, the demand for large-capacity active power filters is increasingly strong in industrial sites. Limited by the capacity and cost of power switching devices and other factors, a large-capacity active power filter usually adopts an active filtering method of parallel operation of multi-module inverters. Compared with an independent DC bus multi-module parallel structure, a common DC bus multi-module active power filter parallel operation system has the advantages of small size, low cost, flexible harmonic compensation, good redundancy fault tolerance performance and easy modular production, and has a good application prospect in the field of harmonic control in future power electronic power systems. However, the parallel structure of common DC bus inverters leads to the inherent circulating current problem, which affects the harmonic compensation performance of the active power filter. Therefore, an effective control strategy must be taken to suppress the circulating current between the parallel inverters.

[0004] At present, for the circulating current suppression problem of the parallel connection system of the multi-module active power filter with common DC bus, document 1 (Research on circulating current suppression strategy of dual active power filter, Gessun, etc., Electrical Drive, 2017, 47 (1): 3-6) proposes a shoot-through circulating current suppression strategy, which suppresses the circulating current between the parallel inverters by adjusting the action time of the zero vector, but various nonlinear factors of the actual active power filter system will cause calculation deviation of the zero vector action time, thereby affecting the harmonic compensation accuracy. Document 2 (Research on circulating current suppression of parallel connection of modular multi-APF, Yu Huajun, etc., Electrical Automation, 2017, 39 (1): 41-44) proposes a circulating current suppression measure of positive sequence, negative sequence and zero sequence current coordinated control, but the PI controller therein cannot realize the shoot-through tracking effect of the harmonic compensation current, and the existence of the integrator also reduces the system response speed. Document 3 (Selection of design parameters to reduce the zero-sequence circulating current flow in parallel operation of dc linked multiple shunt APF units, Shafiuzzaman K.K., et al, Advances in Power Electronics, 2013, 381581: 1-13) improves the circulating current suppression performance of the parallel connection system of the dual-module active power filter with common DC bus by optimizing the control parameters, but in practice, it is difficult to effectively suppress the circulating current by adjusting the parameters only. Therefore, a control method that can effectively suppress the circulating current of the parallel inverters and improve the dynamic response speed of the active power filter is an urgent problem to be solved in the application field of high-power multi-module active power filter. SUMMARY

[0005] The purpose of the present application is to propose a model predictive control method for the parallel connection system of the dual-module active power filter in view of the deficiencies of the existing parallel inverter control technology, to suppress the circulating current of the parallel inverters, to improve the steady-state accuracy of harmonic compensation, to improve the dynamic response speed of harmonic compensation, to improve the system efficiency, and to reduce the cost and volume of the capacitor.

[0006] To achieve the above purpose, the technical scheme of the model predictive control method for the parallel connection system of the dual-module active power filter adopted by the present application is as follows:

[0007] The three-phase grid voltage and the three-phase compensation current of the two parallel inverters in dq0 coordinate system are detected and coordinate-transformed, respectively, 64 synthesized voltage vectors in dq0 coordinate system are obtained according to the switching states of the two inverters, and the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1), k is the current detection time;

[0008] The fundamental component is obtained by inputting the three-phase load current i Labc into a band-pass filter, and the compensation current given value i hdref in dq0 rotating coordinate system is obtained by subtracting the fundamental component from the three-phase load current and then coordinate-transforming. hqref h0ref ,

[0009] The voltage difference obtained by comparing the common DC bus voltage and the given value thereof is PI-regulated to obtain a loss current, and the loss current and the d-axis compensation current given value i hdref are added to obtain the d-axis total component of the compensation current given value i hdref_tatol .

[0010] 64 evaluation values are calculated by evaluating the value function:

[0011] K1 is the circulating current suppression weight coefficient of the parallel inverters, and K2 is the compensation current limiting coefficient of the active power filter;

[0012] The switching state corresponding to the synthesized voltage vector corresponding to the minimum value in the 64 evaluation values is used to control the two inverters.

[0013] Further, the parallel system is modeled according to the synthesized voltage vector, the continuous domain current state equation is obtained, and the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1).

[0014] The above technical scheme has the following beneficial effects:

[0015] ​1.The present application will be based on the active power filter of parallel three-phase inverter based on common DC bus as six-bridge arm inverter system for overall discrete mathematical modeling, the three-phase harmonic compensation current of two inverters is predicted, the 64 synthesized voltage vectors of parallel three-phase inverter are traversed and optimized by evaluating the value function, the optimal voltage vector of each inverter output is obtained, the global optimization of the circulating current between two inverters is realized, the circulating current of parallel inverters is effectively suppressed, and the harmonic compensation accuracy and system efficiency of the active power filter are improved.

[0016] 2.The present application adopts model predictive current control strategy to carry out compensation current control and switching state direct optimization selection of common bus double module active power filter, realizes compensation current closed loop prediction tracking control, and there is no need for traditional PI controller in the compensation current control loop, therefore the harmonic compensation response speed of the active power filter can be significantly improved, and the capacitor cost and volume can be reduced. BRIEF DESCRIPTION OF DRAWINGS

[0017] The technical solutions of the present application will be described in detail below in combination with the drawings and specific embodiments:

[0018] Figure 1 It is a model predictive control block diagram of double module active power filter parallel system;

[0019] Figure 2 It is Figure 1 It is a model predictive controller and its external structure block diagram;

[0020] Figure 3 It is an A-phase current simulation waveform diagram of a nonlinear load under simulation test of the model predictive control method of the present application;

[0021] Figure 4 It is an A-phase compensation current simulation waveform diagram of an active power filter under simulation test of the model predictive control method of the present application;

[0022] Figure 5 It is a circulating current simulation waveform diagram of parallel inverters under simulation test of the model predictive control method of the present application;

[0023] Figure 6 It is an A-phase current simulation waveform diagram of a power grid under simulation test of the model predictive control method of the present application;

[0024] Figure 7 It is an A-phase compensation current simulation waveform diagram before and after load mutation under simulation test of the model predictive control method of the present application. DETAILED DESCRIPTION

[0025] As Figure 1As shown, the dual-module active power filter parallel system comprises: two three-phase inverters, a DC bus capacitor, two sets of three-phase filter inductors, three sets of three-phase current sensors, a set of three-phase voltage sensors, and a DC bus voltage sensor. Among them, the two three-phase inverters are the first inverter 1 and the second inverter 2, and the two sets of three-phase filter inductors are the first set of three-phase filter inductors L h1 and the second set of three-phase filter inductors L h2 . The output end of each three-phase inverter is respectively connected in series with a set of three-phase filter inductors: the first inverter 1 output voltage u ha1 , u hb1 , u hc1 , the output end of the first inverter 1 is connected in series with the first set of three-phase filter inductors L h1 and its parasitic resistance R h1 ; the second inverter 2 output voltage u ha2 , u hb2 , u hc2 , the output end of the second inverter 2 is connected in series with the second set of three-phase filter inductors L h2 and its parasitic resistance R h2 . The two sets of three-phase filter inductors L h1 , L h2 are connected to the three-phase power grid in parallel after being connected in parallel, and the three-phase power grid voltage is u ga , u gb , u gc , and the three-phase power grid current is i ga , i gb , i gc . The three-phase power grid is connected to a three-phase nonlinear load, and the three-phase load current of the nonlinear load is i La , i Lb , i Lc . The positive pole P1 of the DC side of the first inverter 1 and the positive pole P2 of the DC side of the second inverter 2 are connected in parallel and then connected to the positive pole P of the DC bus capacitor C dc , and the negative pole N1 of the DC side of the first inverter 1 and the negative pole N2 of the DC side of the second inverter 2 are connected in parallel and then connected to the negative pole N of the DC bus capacitor C dc .

[0026] The set of three-phase voltage sensors detects the three-phase power grid voltage u ga , u gb , u gc , u gabc =u ga , u gb , u gc , the phase angle θ of the grid voltage is obtained through a phase-locked loop, and the three-phase power grid voltage u gabcA model predictive controller (hereinafter referred to as "model predictive controller") of a dual-module active power filter parallel system with grid voltage phase angle θ input.

[0027] A first set of three-phase current sensors is used to detect three-phase compensation currents i ha1 (k) of the first inverter 1 hb1 (k) of the first inverter 1 hc1 (k) of the first inverter 1 habc1 (k) of the first inverter 1 ha1 (k) of the first inverter 1 hb1 (k) of the first inverter 1 hc1 (k) of the first inverter 1 habc1 (k) of the first inverter 1 is input into the model predictive controller.

[0028] A second set of three-phase current sensors is used to detect three-phase compensation currents i ha2 (k) of the second inverter 2 hb2 (k) of the second inverter 2 hc2 (k) of the second inverter 2 habc2 (k) of the second inverter 2 ha2 (k) of the second inverter 2 hb2 (k) of the second inverter 2 hc2 (k) of the second inverter 2 habc2 (k) of the second inverter 2 is input into the model predictive controller, and k is the current detection time.

[0029] A third set of three-phase current sensors is used to detect three-phase load currents i La , i Lb , i Lc , i Labc = i La , i Lb , i Lc , and the three-phase load currents i Labc are input into the model predictive controller.

[0030] A DC bus voltage sensor is used to detect the voltage u dc of the common DC bus capacitor C dc of the two parallel three-phase inverters 1, 2, which is referred to as the common DC bus voltage u dc , and the common DC bus voltage u dc is input into the model predictive controller.

[0031] The output terminals of the model predictive controller are respectively connected to the two parallel three-phase inverters 1, 2 to drive the two three-phase inverters 1, 2 to work.

[0032] As Figure 2As shown, the model predictive controller comprises a band-pass filter, a current prediction module, a PI controller, a coordinate function evaluation module and four coordinate transformation modules.

[0033] Three-phase grid voltage u gabc and the phase angle θ thereof are input into the first coordinate transformation module, which performs coordinate transformation on the grid voltage in the abc stationary coordinate system to obtain the grid voltage u gd , u gq , u g0 in the dq0 rotating coordinate system.

[0034]

[0035] The obtained grid voltage u gd , u gq , u g0 and the grid voltage phase angle θ are input into the current prediction module.

[0036] The three-phase compensation current i habc1 (k) of the first inverter 1 and the grid voltage phase angle θ are input into the second coordinate transformation module, which performs coordinate transformation on the three-phase compensation current i habc1 (k) in the abc stationary coordinate system to obtain three compensation currents i hd1 (k), i hq1 (k), i h01 (k) in the dq0 rotating coordinate system.

[0037]

[0038] The obtained three compensation currents i hd1 (k), i hq1 (k), i h01 (k) are input into the current prediction module.

[0039] Similarly, the three-phase compensation current i habc2 (k) of the second inverter 2 and the grid voltage phase angle θ are input into the third coordinate transformation module, which performs coordinate transformation on the three-phase compensation current i habc2 (k) in the abc stationary coordinate system to obtain three compensation currents i hd2 (k), i hq2 (k), i h02 (k) in the dq0 rotating coordinate system.

[0040]

[0041] The three compensation currents i hd2 (k), i hq2 (k), i h02 (k) are input into the current prediction module.

[0042] The three-phase load current i Labc is input into the band-pass filter, and the fundamental component i fabc is obtained after the band-pass filter. fa , i fb , i fc The three-phase load current i Labc is subtracted from the fundamental component i fabc , and the compensation current given value i habcref of the active power filter is obtained. haref , i hbref , i hcref .

[0043] The compensation current given value i habcref and the grid voltage phase angle θ are input into the fourth coordinate transformation module, and the compensation current given value i habcref in the abc stationary coordinate system is coordinate-transformed based on the grid voltage phase angle θ, and the compensation current given value i hdref , i hqref , i h0ref in the dq0 rotating coordinate system is obtained. The coordinate transformation formula is as follows:

[0044]

[0045] The q-axis and 0-axis compensation current given values i hqref , i h0ref are input into the value function evaluation module.

[0046] The common DC bus voltage u dc is compared with the DC bus voltage given value U dcref , and the voltage difference is obtained. The voltage difference is input into the PI controller, and the loss current i loss is output after the PI controller adjustment. The loss current i loss and the d-axis component compensation current given value i hdref output by the fourth coordinate transformation module are added, and the d-axis total component i hdref_tatol of the compensation current given value is obtained. The d-axis total component i hdref_tatol of the compensation current given value is input into the value function evaluation module.

[0047] The current prediction module inputs the grid voltage u gd , u gq , u g0 , and the compensation current ihd1 (k), i hq1 (k), i h01 (k), i hd2 (k), i hq2 (k), i h02 (k) and grid voltage phase angle θ, to obtain the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1), and the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1) are input into the value function evaluation module.

[0048] Specifically:

[0049] The current prediction module first obtains 64 synthesized voltage vectors U APFα (j), U APFβ (j), U APF0 (j), j = 1 ~ 64 in the αβ0 coordinate system according to the switching states of the first inverter 1 and the second inverter 2 and their corresponding 8 basic voltage values in the αβ0 coordinate system.

[0050] Table 1

[0051]

[0052] Among them, the 64 synthesized voltage vectors U APFα (j), U APFβ (j), U APF0 (j) are constructed as shown in Table 2:

[0053] Table 2

[0054]

[0055]

[0056]

[0057] Then, for the 64 synthesized voltage vectors U APFα (j), U APFβ (j), U APF0 (j), based on the grid voltage phase angle θ, the synthesized voltage vectors U APFα (j), U APFβ (j), U APF0 (j) in the αβ0 coordinate system are coordinate-transformed to obtain the synthesized voltage vectors U APFd(j), U APFq (j), U APF0 (j), the coordinate transformation formula is as follows:

[0058]

[0059] Then, based on the synthesis voltage vector U APFd (j), U APFq (j), U APF0 (j) is carried out mathematical modeling of the dual-module active power filter parallel system in the dq0 coordinate system, and continuous domain current state equations are obtained:

[0060]

[0061] Wherein: i hd , i hq , i h0 is the total compensation current in the dq0 coordinate system: L h is the filter inductance, L h = L h1 = L h2 , R h is the parasitic resistance of the filter inductance, R h = R h1 = R h2 , T s is the sampling period of the dual-module active power filter parallel system, and ω is the angular frequency of the three-phase power grid voltage.

[0062] The continuous domain current state equation is forward Euler discretized to obtain the prediction model of the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1), and the prediction model is as follows:

[0063]

[0064] The 64 d-axis voltage components U APFd (j), the 64 q-axis voltage components U APFq (j), and the 64 0-axis voltage components U APF0 (j) are respectively brought into the prediction model of i hd (k+1), i hq (k+1), i h0 (k+1), and 64 compensation current prediction values i hd (k+1), i hq (k+1), i h0 (k+1) are obtained. All the compensation current prediction values i hd (k+1), ihq (k+1), i h0 (k+1) are input into the value function evaluation module.

[0065] The value function evaluation module processes all the input compensation current prediction values i hd (k+1), i hq (k+1), i h0 (k+1), the compensation current given value i hqref , i h0ref and the d-axis total component of the compensation current given value i hdref_tatol and calculates 64 value functions G hdq0 , that is, obtains 64 evaluation values:

[0066]

[0067] wherein K1 is the circulating current suppression weight coefficient of the parallel inverter, adjusting K1 can make the harmonic compensation and circulating current suppression optimal, K2 is the compensation current limiting coefficient of the active power filter, when the compensation current amplitude is less than the limiting value, K2=0, when the compensation current amplitude is greater than the limiting value, K2=∞.

[0068] Take the resultant voltage vector corresponding to the minimum value among the 64 evaluation values, the d-axis voltage component U APFd (j) of the minimum value corresponds to the optimal resultant voltage vector U APFq (j) of the q-axis voltage component U APF0 (j) of the 0-axis voltage component U APFd_opt . APFq_opt . APF0_opt .

[0069] According to the optimal resultant voltage vector U APFd_opt , the switching state of the two parallel inverters 1, 2 corresponding to U APFq_opt , U APF0_opt , the upper bridge arm drive signal S a1+ , S b1+ , S c1+ and the lower bridge arm drive signal S a1- , S b1- , S c1- , the upper bridge arm drive signal S a2+ , S b2+ , S c2+ and the lower bridge arm drive signal S a2- , S b2- , S c2- of the first inverter 1 are obtained. The drive signals S a1+ , S b1+ , Sc1+ a1- b1- c1- a2+ b2+ c2+ a2- b2- c2- The 12 switching tubes of the two parallel inverters 1, 2 are turned on and turned off to control the first inverter 1 to output three-phase voltage u ha1 hb1 hc1 and the second inverter 2 to output three-phase voltage u ha2 hb2 hc2 , wherein the three-phase voltage u ha1 hb1 hc1 acts on the filter inductance L h1 and its parasitic resistance R h1 to generate model predictive current control of the first inverter 1, and the three-phase voltage u ha2 hb2 hc2 acts on the filter inductance L h2 and its parasitic resistance R h2 to generate model predictive current control, thereby realizing model predictive current control of the dual-module active power filter parallel system.

[0070] In order to illustrate the model predictive control method of the dual-module active power filter parallel system of the application, simulation tests are carried out, Figure 1 The parallel system parameters are as follows: filter inductance L h = 1 mH, parasitic resistance R h of the filter inductance = 0.1 Ω, effective value U g of the grid voltage = 380 V, fundamental frequency f = 50 Hz, filter inductance L load of the three-phase uncontrolled bridge rectifier = 3 mH, DC load resistance R load of the three-phase uncontrolled bridge rectifier = 10 Ω, model predictive current control sampling period T s = 0.0001 s, DC bus capacitor C dc = 2200 μF, given voltage U dcref of the DC bus capacitor = 700 V. The simulation tests obtain Figures 3-7 , respectively, the simulation waveform diagrams of the nonlinear load A-phase current, the active power filter A-phase compensation current, the parallel inverter circulating current, the grid A-phase current, and the A-phase compensation current before and after the load mutation of the dual-module active power filter parallel system, at 0.5 s, the DC load resistance of the three-phase uncontrolled bridge rectifier is changed from R​​​​​​​​​​​​​​​​​load = 10 Ω to R load = 5 Ω. From Figures 3-7 It can be seen that the application can effectively inhibit the circulating current problem of the parallel inverter, and improve the harmonic compensation accuracy of the active power filter in the steady state and the harmonic compensation response speed in the load mutation case.

[0071] Although the preferred embodiments of the application have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the application, and the scope of the application is defined by the claims.

Claims

1. A model predictive control method of a dual-module active power filter parallel system, characterized by Comprise: The three-phase grid voltage and the three-phase compensation current of the two parallel inverters in dq0 coordinate system are detected and coordinate-transformed respectively, 64 synthesized voltage vectors in dq0 coordinate system are obtained according to the switching states of the two inverters, and the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1), k is the current detection time; The three-phase load current input band-pass filter obtains the fundamental component, and the three-phase load current i Labc Subtracting the fundamental component and then performing coordinate transformation obtains the compensation current given value i hdref 、 hqref 、 h0ref , The voltage difference obtained by comparing the public DC bus voltage and its given value is PI-regulated to obtain a loss current, and the loss current and a d-axis compensation current given value i hdref The d-axis total component i of the compensation current given value is obtained by adding hdref_tatol ; 64 evaluation values are calculated by evaluating the value function: K1 is a circulating current suppression weight coefficient of the parallel inverter, and K2 is a compensation current limiting coefficient of the active power filter. According to the minimum value of the 64 evaluation values, the switch state corresponding to the synthesized voltage vector is controlled to control the two inverters.

2. The model predictive control method of a dual-module active power filter parallel system according to claim 1, characterized in that: According to the modeling of the synthetic voltage vector pair on the parallel system, a continuous domain current state equation is obtained, and the compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1).

3. The model predictive control method of a dual-module active power filter parallel system according to claim 2, characterized in that: The continuous domain current state equation is: i hd , hq , h0 is the total compensation current in dq0 coordinate system, i hd1 (k), i hq1 (k), i h01 (k) and i hd2 (k), i hq2 (k), i h02 (k) are the three-phase compensation currents of two inverters in dq0 coordinate system, L h is the filter inductance, R h is the parasitic resistance of filter inductance, ω is the angular frequency of three-phase grid voltage, U APFd (j), U APFq (j), U APF0 (j) is the resultant voltage vector in dq0 coordinate system, j = 1 ~ 64.

4. The model predictive control method of a dual-module active power filter parallel system according to claim 3, characterized in that: The compensation current prediction value i hd (k+1), i hq (k+1), i h0 (k+1) is: T s is the sampling period.

5. The model predictive control method of a dual-module active power filter parallel system according to claim 1, characterized in that: According to the three-phase network voltage u ga , u gb , u gc and the phase angle θ thereof, the transformation by means of the formula results in the three-phase network voltage u gd , u gq , u g0 in the dq0 rotating coordinate system.

6. The model predictive control method of a dual-module active power filter parallel system according to claim 1, characterized in that: The three-phase compensation current i habc1 (k) and the grid voltage phase angle θ, to obtain the three-phase compensation current i of the inverter in the dq0 rotating coordinate system hd1 (k), i hq1 (k), i h01 (k):

7. The model predictive control method of a dual-module active power filter parallel system according to claim 1, characterized in that: According to the two inverter switch states and the corresponding eight basic voltage values in the αβ0 coordinate system, 64 synthesized voltage vectors U in the αβ0 coordinate system are obtained APFα (j), U APFβ (j), U APF0 (j), j = 1 ~ 64, the synthesized voltage vectors in the αβ0 coordinate system are coordinate-transformed based on the grid voltage phase angle θ, to obtain the synthesized voltage vectors in the dq0 rotating coordinate system.

8. The model predictive control method of a dual-module active power filter parallel system according to claim 7, characterized in that: According to the coordinate transformation formula, the resultant voltage vector U in the dq0 rotating coordinate system is obtained APFd (j), U APFq (j), U APF0 (j):

Images