A double vector model predictive control method applied to a three-phase full-bridge inverter

Abstract

The application discloses a double-vector model predictive control method applied to a three-phase full-bridge inverter, and specifically comprises the following steps: sampling a load output voltage, a filter current and a load output current at a k moment; predicting a load output voltage value at k+2; calculating a first acting voltage vector with a minimum value function; distributing the first acting voltage vector with the acting time of seven different voltage vectors, synthesizing two voltage vectors to obtain seven groups of voltage vector combinations; predicting a load output voltage prediction value at k+2 under the voltage vector combination; comparing the value function values under the voltage vector combinations, applying two voltage vectors and the acting time which make the new value function value minimum to the three-phase full-bridge inverter, and outputting the switching state corresponding to the optimal voltage vector combination. The application takes into account the advantages of proportional integral current control, hysteresis current control and fuzzy control, and has the advantages of simple structure, strong anti-interference ability, strong robustness and low harmonic content.

Description

Technical Field

[0001] This invention belongs to the field of three-phase full-bridge inverter control technology, and particularly relates to a dual-vector model predictive control method applied to three-phase full-bridge inverters. Background Technology

[0002] The circuit structure of a three-phase full-bridge inverter is as follows: Figure 1 As shown. The entire inverter circuit is powered by an input-side DC voltage source. U dc A full bridge consisting of 6 MOSFET switches (S1-S6) and a filter inductor. L Filter capacitor C and purely resistive loads R composition.

[0003] Currently, the main control methods for three-phase full-bridge inverter circuits include proportional-integral current control, hysteresis current control, and fuzzy control. Proportional-integral current control has a simple structure and high control accuracy, but parameter tuning is difficult and anti-interference capability is weak; hysteresis control has a fast response speed and strong robustness, but requires a very high sampling frequency, so it is only suitable for applications where control accuracy is not high; fuzzy control does not require a precise mathematical model of the process and has strong robustness, but control accuracy is low and dynamic response is poor.

[0004] Model predictive control (MPC) is a closed-loop system optimization control strategy based on a mathematical model. Its core algorithm involves predicting the future dynamic model of the system, optimizing the model data online in real time, performing rolling real-time control, and providing feedback to correct deviations generated by the model. Therefore, MPC offers advantages such as high control accuracy and robustness. It can effectively overcome uncertainties, nonlinearities, and parallelism in the control process and easily handle various constraints on the controlled and manipulated variables. Consequently, MPC demonstrates excellent capabilities in addressing power quality issues such as output voltage and current in three-phase full-bridge inverters.

[0005] Model predictive control first establishes a mathematical model of the three-phase full-bridge inverter, predicts the value at the next moment based on the current measured value and the mathematical model, selects the optimal voltage vector by finding the minimum value of the cost function, and thus solves the optimal switching state.

[0006] Figure 2 This is a model predictive control block diagram of a three-phase full-bridge inverter. The various modules in the diagram form a closed-loop control circuit. The predictive control module and the value function module are the core of the control. By predicting the voltage value at the next moment and optimizing the value function, the optimal switching action can be obtained, thus completing the model predictive control of the three-phase full-bridge inverter.

[0007] For three-phase full-bridge inverters, the optimal voltage vector in traditional model predictive control (MPC) operates throughout the entire sampling period, with only seven sets of switching voltage vectors. This limitation is too great, and the number of optimization attempts is too small. When the predicted value after the optimal voltage vector is applied is less than or greater than the reference value, it will cause significant fluctuations, thereby increasing the system's harmonics. Therefore, many new research directions in model predictive control have emerged, such as delay compensation, value function optimization, and increasing the number of vectors. Furthermore, traditional MPC strategies generally only predict the load output voltage value at time k+1, resulting in a small prediction step. MPC, on the other hand, is a real-time rolling optimization control. Therefore, predicting the voltage value at time k+2 is more effective than predicting the voltage value at time k+1 in minimizing the perfect square difference between the predicted and reference values, thus improving the inverter's dynamic stability.

[0008] Traditional model predictive control, which involves a single cycle and only one switching action, has significant limitations. Therefore, increasing the number of switching actions within the same cycle has become a hot research topic in recent years. Some literature employs a model predictive control method that superimposes multiple voltage vectors to control the inverter. This method uses a dual-voltage vector duty cycle control strategy, combining a zero voltage vector with a non-zero voltage vector to change the amplitude of the regulating voltage vector. Furthermore, it utilizes the concept of SVPWM to combine adjacent effective voltage vectors and non-zero voltage vectors in each sector to control the inverter. However, this method suffers from insufficient control precision due to the limited number of optimization attempts. Summary of the Invention

[0009] This invention improves the output voltage steady-state characteristics and harmonic content of three-phase full-bridge inverters, and provides a dual-vector model predictive control method for three-phase full-bridge inverters.

[0010] The present invention provides a dual-vector model predictive control method for a three-phase full-bridge inverter, comprising the following steps: Step 1: Sample the load output voltage at time k on the αβ coordinate axis. u c (k) Filtering current i (k) and load output current i o (k).

[0011] Step 2: Predict the load output voltage value at sampling time k+2 u c (k+2): , In the formula, L For filtering inductors,C For filtering capacitors, T s The sampling period; V iα (k) and V iβ (k) represents the inverter output voltage at time k along the αβ coordinate axis. i α (k) and i β (k) represents the inductor current along the αβ coordinate axis at time k. u cα (k) and u cβ (k) represents the load output voltage along the αβ coordinate axis at time k. i oα (k) and i oβ (k) represents the load output current at time k along the αβ coordinate axis.

[0012] Substitute the predicted load output voltage value into the following formula to calculate the value function: In the formula, This is the reference value for the output voltage on the α coordinate axis; This is the reference value for the output voltage on the β axis.

[0013] Step 3: Compare the value function values ​​under the action of 7 different voltage vectors, and select the one that makes... g The smallest voltage vector is the first applied voltage vector at the next sampling time k+1.

[0014] Step 4: Allocate the application time of the first applied voltage vector and the seven different voltage vectors obtained in Step 3 according to the following formula: In the formula, t 1 and t 2 represents the duration of action of the first and second voltage vectors, respectively; g 1 and g 2 represents the value function values ​​under the first and second optimal voltage vectors, respectively.

[0015] The two voltage vectors and the application time are combined according to the following formula to obtain 7 sets of voltage vector combinations; In the formula, V 1 represents the optimal voltage vector during the first action time. V 2 represents the voltage vector during the second action time.

[0016] Step 5: Substitute the obtained voltage vector combination into the following formula to obtain the new predicted value of the load output voltage at time k+2; In the formula, V (k) represents sampling time k. V 1 and V 2 composite vector V The inverter output voltage.

[0017] Step 6: Compare the value function values ​​under the combined action of voltage vectors, and apply the two voltage vectors and their action times that minimize the new value function value to the three-phase full-bridge inverter to output the corresponding switching state of the optimal voltage vector combination.

[0018] Step 7: When the sampling time reaches k+1, repeat steps 1-6.

[0019] The beneficial technical effects of this invention are as follows: This invention adds one more voltage vector than traditional model predictive control within a sampling period, resulting in a synthesized voltage vector with the advantages of arbitrary direction and amplitude, thus reducing the error between the predicted and reference values. Furthermore, with prediction step N=2, compared to prediction step N=1 which predicts the output voltage value at time k+1, predicting the value at time k+2 and implementing the corresponding switching state at time k+1 allows the output voltage to better follow the fluctuations of the reference voltage, thereby reducing harmonic content.

[0020] The control method of the present invention combines the advantages of proportional-integral current control, hysteresis current control and fuzzy control, and has the advantages of simple structure, strong anti-interference ability, strong robustness and low harmonic content. Attached Figure Description

[0021] Figure 1 It is a three-phase full-bridge inverter circuit topology.

[0022] Figure 2 This is a model predictive control block diagram for a three-phase full-bridge inverter.

[0023] Figure 3 These are the implementation steps of the dual-vector model predictive control method of the present invention.

[0024] Figure 4 This is a schematic diagram of the improved two-vector model predictive control principle. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0026] The implementation steps of the dual-vector model predictive control method for a three-phase full-bridge inverter according to the present invention are as follows: Figure 3 As shown, it includes the following steps: Step 1: Sample the load output voltage at time k on the αβ coordinate axis. u c (k) Filtering current i (k) and load output current i o (k).

[0027] Step 2: Predict the load output voltage value at sampling time k+2 u c (k+2): , In the formula, L For filtering inductors, C For filtering capacitors, T s The sampling period; V iα (k) and V iβ (k) represents the inverter output voltage at time k along the αβ coordinate axis. i α (k) and i β (k) represents the inductor current along the αβ coordinate axis at time k. u cα (k) and u cβ (k) represents the load output voltage along the αβ coordinate axis at time k. i oα (k) and i oβ (k) represents the load output current at time k along the αβ coordinate axis.

[0028] Substitute the predicted load output voltage value into the following formula to calculate the value function: In the formula, This is the reference value for the output voltage on the α coordinate axis; This is the reference value for the output voltage on the β axis.

[0029] Step 3: Compare the value function values ​​under the action of 7 different voltage vectors, and select the one that makes... g The smallest voltage vector is the first applied voltage vector at the next sampling time k+1.

[0030] Step 4: Allocate the application time of the first applied voltage vector and the seven different voltage vectors obtained in Step 3 according to the following formula: In the formula, t 1 and t 2 represents the duration of action of the first and second voltage vectors, respectively; g 1 and g 2 represents the value function values ​​under the first and second optimal voltage vectors, respectively.

[0031] The two voltage vectors and the application time are combined according to the following formula to obtain 7 sets of voltage vector combinations; In the formula, V 1 represents the optimal voltage vector during the first action time. V 2 represents the voltage vector during the second action time.

[0032] Step 5: Substitute the obtained voltage vector combination into the following formula to obtain the new predicted value of the load output voltage at time k+2; Step 6: Compare the value function values ​​under the combined action of voltage vectors, and apply the two voltage vectors and their action times that minimize the new value function value to the three-phase full-bridge inverter to output the corresponding switching state of the optimal voltage vector combination.

[0033] Step 7: When the sampling time reaches k+1, repeat steps 1-6.

[0034] The improved dual-vector model predictive control principle of this invention: (1) Two voltage vectors act within one sampling period, and the voltage vector during the first action time is the voltage vector that minimizes the value function within one sampling period. V 1. Voltage vector during the second action time V 2 represents one of the seven different voltage vectors, where... V The selection criterion for 2 is based on the fact that within a sampling period V 1 and V 2. The combined effect of the synthesized voltage vectors minimizes the value function, such as... Figure 4 As shown, this solves the problems of traditional model predictive control, which uses only one voltage vector within a sampling period, resulting in poor steady-state characteristics of the output voltage and high harmonic content.

[0035] (2) The distribution of action time was considered when selecting the voltage vector combination, making the selection of the voltage vector combination more accurate.

[0036] (3) Directly control the output voltage and predict the step N=2, which can make the load output voltage better follow the fluctuation of the reference value.

[0037] The mathematical model of a three-phase full-bridge inverter in the αβ coordinate axis is: (1) (2) In the formula, L For filtering inductors, C For filtering capacitors, V iα and V iβ These represent the inverter output voltages on the α and β coordinate axes, respectively. i α and i β These are the inductor currents along the α and β coordinate axes, respectively. u cα and u cβ These represent the load output voltages along the α and β coordinate axes, respectively. i oα and i oβ These represent the load output currents on the α and β coordinate axes, respectively.

[0038] Equation (1) is discretized using Euler methods to obtain: (3) in, T s The sampling period.

[0039] Equation (3) at time k+2 is: (4) During the sampling period T s If the value is sufficiently small, it can be assumed that the load current does not change within one sampling period, i.e. .

[0040] Substituting equation (3) into equation (1) yields: (5) As can be seen from equation (5), the load output voltage value at time k+2 can be predicted by measuring the load output voltage value, filter current value and load output current value at time k, and the prediction step size is 2.

[0041] The idea behind model predictive control is to select the optimal voltage vector through a value function that minimizes the perfect square difference between the measured load output voltage and the reference load output voltage. The value function can then be expressed as: (6) In the formula, This is the reference value for the output voltage on the α coordinate axis; This is the reference value for the output voltage on the β axis.

[0042] In single-vector model predictive control, since only one voltage vector can be selected within a sampling period, the predicted load output voltage deviates significantly from the reference load output voltage, and voltage and current harmonics are also large. Therefore, it is considered to use two voltage vectors within a sampling period for synthesis.

[0043] When performing voltage vector synthesis, the most crucial aspect is the allocation of the application time. This study employs a modulation model predictive control method, assuming that the application times of the two voltage vectors are inversely proportional to the value functions, resulting in: (7) In the formula, t 1 and t 2 represents the duration of action of the first and second voltage vectors, respectively; g 1 and g 2 represents the value function values ​​under the first and second optimal voltage vectors, respectively.

[0044] After determining the duration of action of the two voltage vectors, they are synthesized using the principle of vector composition. The synthesized voltage vector is... V for: (8) In the formula, V 1 represents the optimal voltage vector during the first action time. V 2 represents the voltage vector during the second action time.

[0045] The predicted load output voltage changes under the improved dual-vector model predictive control as follows: (9) Under the action of the improved dual-vector MPC, the value function remains unchanged, as shown in equation (6).

[0046] Simulation experiment: A simulation platform for a three-phase full-bridge inverter based on Matlab / Simulink was built, using both the improved dual-vector model predictive control method and the traditional model predictive control method. The simulation parameters are shown in Table 1. Table 1. Parameters of the Simulation Platform for Three-Phase Full-Bridge Inverter By providing the same reference voltage and sampling frequency, the test conditions of the proposed method are consistent with those of the traditional vector model predictive control method, and the total harmonic distortion (THD) under the two control methods is measured.

[0047] Table 2 shows the simulation data of the present invention and the traditional vector model predictive control method.

[0048] Table 2 Simulation data of the present invention and the traditional vector model predictive control method The data in Table 2 shows that, at the same sampling frequency, the improved dual-vector model predictive control has lower total harmonic distortion and lower harmonic content than the traditional model predictive control, and the harmonic content also decreases as the frequency increases.

Claims

1. A dual-vector model predictive control method applied to a three-phase full-bridge inverter, characterized in that, Includes the following steps: Step 1: Sample the load output voltage at time k on the αβ coordinate axis. u c (k) Filtering current i (k) and load output current i o (k); Step 2: Predict the load output voltage value at sampling time k+2 u c (k+2): ， In the formula, L For filtering inductors, C For filtering capacitors, T s The sampling period; V iα (k) and V iβ (k) represents the inverter output voltage at time k along the αβ coordinate axis. i α (k) and i β (k) represents the inductor current along the αβ coordinate axis at time k. u cα (k) and u cβ (k) represents the load output voltage along the αβ coordinate axis at time k. i oα (k) and i oβ (k) represents the load output current under the αβ coordinate axis at time k; Substitute the predicted load output voltage value into the following formula to calculate the value function: In the formula, This is the reference value for the output voltage on the α coordinate axis; This is the reference value for the output voltage on the β axis; Step 3: Compare the value function values ​​under the action of 7 different voltage vectors, and select the one that makes... g The smallest voltage vector is the first applied voltage vector at the next sampling time k+1; Step 4: Allocate the application time of the first applied voltage vector and the seven different voltage vectors obtained in Step 3 according to the following formula: In the formula, t 1 and t 2 represents the duration of action of the first and second voltage vectors, respectively; g 1 and g 2 represents the value function values ​​under the first and second optimal voltage vectors, respectively; The two voltage vectors and the application time are combined according to the following formula to obtain 7 sets of voltage vector combinations; In the formula, V 1 represents the optimal voltage vector during the first action time. V 2 represents the voltage vector during the second action time; Step 5: Substitute the obtained voltage vector combination into the following formula to obtain the new predicted value of the load output voltage at time k+2; In the formula, V (k) represents sampling time k. V 1 and V 2 composite vector V The inverter output voltage is below; Step 6: Compare the value function values ​​under the action of the voltage vector combination, and apply the two voltage vectors and the action time that minimize the new value function value to the three-phase full-bridge inverter to output the corresponding switching state of the optimal voltage vector combination; Step 7: When the sampling time reaches k+1, repeat steps 1-6.

Images