A three-phase inverter current sensorless model predictive control method and device

Abstract

The application discloses a kind of three-phase inverter current sensorless model predictive control method and device, comprising the following steps: constructing three-phase inverter model predictive controller based on capacitor current observer;Three-phase inverter current sensorless model predictive control algorithm based on neural network is constructed;State quantity is calculated using neural network, and three-phase inverter model predictive control based on capacitor current is implemented.The designed method and device can realize model predictive control to three-phase inverter without current sensor and observer, with the advantages of low cost, easy to deploy and strong adaptability.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of model predictive control for three-phase inverters and neural network applications, specifically to a sensorless model predictive control method and device for three-phase inverters based on neural networks. Background Technology

[0002] In the Finite Control Set Model Predictive Control (FCSMPC) algorithm for three-phase inverters, the key to implementing model predictive control lies in accurately acquiring the state information of the three-phase inverter. Generally, for a two-level three-phase voltage source inverter with an LC filter, a complete implementation of FCSMPC requires at least nine voltage and current sensors. Although measures such as load current observers can reduce the number of current sensors by three, six sensors are still needed. This makes the control performance of the FCSMPC algorithm susceptible to sensor influences, and a larger number of sensors also increases system hardware costs and reduces reliability.

[0003] To address the issue of numerous sensors in finite control set model predictive control (FMCC) algorithms for three-phase inverters, existing technologies, such as Chinese patent application CN121283226A, disclose a sensor-saving FMCC method and apparatus for three-phase inverters. This method utilizes the fitting characteristics of neural networks to achieve FMCC based on two voltage sensors and two current sensors. However, within the existing FCSMPC algorithm architecture, it is difficult to further reduce the number of sensors. Therefore, to further reduce the number of sensors in the three-phase inverter FMCC, this solution employs a neural network to fit the three-phase inverter capacitor current observer, achieving FMCC without current sensors and observers. The designed method and apparatus offer advantages such as low cost, ease of deployment, and strong adaptability. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a low-cost, easy-to-deploy, and highly adaptable sensorless model predictive control method and apparatus for three-phase inverters.

[0005] The technical solution of this invention to solve the above problems is: a sensorless model predictive control method and device for a three-phase inverter, comprising the following steps: A1. Construct a predictive controller for a three-phase inverter model based on a capacitor current observer; A2. Construct a sensorless model predictive control algorithm for a three-phase inverter based on neural networks; A3. Use neural networks to calculate state variables and implement model predictive control of the three-phase inverter based on capacitor current.

[0006] The advantages of this invention are as follows: Addressing the problem of multi-sensor configuration affecting the control performance and reliability of model predictive control algorithms for three-phase inverters, this invention provides a low-cost, easy-to-deploy, and highly adaptable sensorless model predictive control method and device for three-phase inverters. The method's advantages include: requiring only two voltage sensors to achieve model predictive control of the three-phase inverter, improving the reliability of the control system and reducing hardware costs; eliminating the need for a current observer, reducing the impact of observer design and parameter tuning, and improving the robustness of system state acquisition; and using a neural network to directly calculate system state variables, achieving seamless integration of state variable calculation and control algorithm implementation, improving algorithm computational efficiency, reducing the computational performance requirements of the microcontroller, and facilitating algorithm deployment and implementation. Attached Figure Description

[0007] Figure 1 A flowchart of the control method in this invention.

[0008] Figure 2 The block diagram of the classic FCSMPC algorithm structure for a three-phase inverter in this invention.

[0009] Figure 3 The block diagram of the three-phase inverter model predictive controller based on the capacitor current observer in this invention.

[0010] Figure 4 The block diagram of the sensorless model predictive control algorithm for a three-phase inverter based on neural networks in this invention.

[0011] Figure 5 The neural network structure diagram in this invention.

[0012] Figure 6 The output voltage waveform of the three-phase inverter under resistive load conditions in this invention.

[0013] Figure 7 The output voltage waveform of the three-phase inverter under no-load conditions in this invention.

[0014] Figure 8 The output voltage waveform of the three-phase inverter under resistive-inductive load conditions in this invention.

[0015] Figure 9 The output voltage waveform of the three-phase inverter under nonlinear load conditions in this invention.

[0016] Figure 10 The structural diagram of the sensorless model predictive control device for a three-phase inverter in this invention. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0018] The present invention discloses a sensorless model predictive control method and apparatus for a three-phase inverter, the flowchart of which is shown below. Figure 1 As shown, it includes the following steps: A1. Construct a predictive controller for a three-phase inverter model based on a capacitor current observer.

[0019] Figure 2 The diagram shows the classic FCSMPC algorithm block diagram for a three-phase inverter. The three-phase inverter in the diagram is a two-level three-phase voltage source inverter with an LC filter. The model predictive controller uses a finite control set model predictive control algorithm to achieve output voltage tracking control of the three-phase inverter. Figure 2 In the classic FCSMPC algorithm shown, implementing model predictive control requires at least nine sensors to acquire the three-phase inductor current ( i sa-b-c ), three-phase capacitor voltage ( v oa-b-c ) and three-phase load current ( i oa-b-c ).

[0020] The process of constructing a predictive controller for a three-phase inverter model based on a capacitor current observer is as follows: A1.1 For the predictive control algorithm of the finite control set model of a three-phase inverter, an output voltage prediction model based on capacitor current is designed.

[0021] exist Figure 2 In the middle, with inductor current i s and capacitor voltage v o As state variables, the state equations of the three-phase inverter are obtained as follows: (1) In the formula, L and C These are the filter inductor and capacitor, respectively. r The equivalent resistance of the line. i s , v o , i o , v s These are the three-phase inductor current vector, the three-phase capacitor voltage vector, the three-phase load current vector, and the three-phase bridge arm output voltage vector, respectively.

[0022] Three-phase abc State variables in coordinate system ( x a , xb , x c )arrive αβ Vector in coordinate system x αβ The general formula for calculation is: (2) In the formula, the operator .

[0023] for Figure 2 In a two-level three-phase voltage source inverter, the switching function combination is defined. S =( S a , S b , S c ),in S a , S b , S c This indicates the state of the upper bridge arm in the three-phase bridge arm of the inverter. A value of 1 indicates that the upper bridge arm is on and the lower bridge arm is off, while a value of 0 indicates that the upper bridge arm is off and the lower bridge arm is on. In equation (1) , V dc This is the DC side voltage.

[0024] According to equation (1) by control cycle T s After discretization, the discrete state equations are obtained: (3) Ignoring the equivalent resistance of the circuit, let equation (1) r =0. (Based on the capacitance current equation) i c ( k )= i s ( k )- i o ( k ), to obtain based on capacitor current vector i c ( k ) and capacitor voltage vector v o ( k Three-phase inverter output voltage prediction model: (4) A1.2 Construct a capacitor current observer to obtain a three-phase inverter model predictive controller based on the capacitor current observer.

[0025] The equations for constructing the capacitor current observer are as follows: (5) In the formula, the state variable x ( k )=[ v o ( k ) i c ( k )] T Observations of state variables ,matrix , , .parameter k 1. k The formula for calculating 2 is: (6) The capacitor current observation value obtained from equation (5) Substituting into equation (4), we obtain the three-phase inverter output voltage prediction model based on capacitor current observations: (7) Furthermore, based on the finite control set model predictive control algorithm, we obtain... k +2 time interval and switching function combination S j Corresponding predicted output voltage value: (8) For a two-level three-phase voltage source inverter, in equation (8) j ={0,…,7} For combination with switching functions S j The corresponding three-phase bridge arm output voltage vector.

[0026] For the tracking control of the output voltage of a three-phase inverter, the cost function is designed as follows: (9) In the formula, , These represent the real and imaginary parts of the output voltage reference value, respectively. , These are the predicted output voltage values. The real and imaginary parts.

[0027] For equation (9), iterate through and calculate the cost functions corresponding to all 8 switch function combinations, and take the switch function combination that minimizes the cost function as the minimum cost function. k The control inputs of the three-phase inverter at time +1. The constructed predictive controller for the three-phase inverter model based on the capacitor current observer is as follows: Figure 3As shown.

[0028] A2. Construct a sensorless model predictive control algorithm for a three-phase inverter based on neural networks.

[0029] Figure 4 The diagram shows the constructed sensorless model predictive control algorithm for a three-phase inverter. This algorithm only requires two voltage sensors and does not require a current sensor or observer. The algorithm uses a three-phase inverter model predictive controller based on a capacitor current observer and replaces the capacitor current observer with a neural network to calculate the capacitor current state variables.

[0030] The constructed neural network is a three-layer backpropagation (BP) neural network, such as... Figure 5 As shown. This neural network is used to fit the capacitor current observer and the three-phase capacitor voltage coordinate transformation process, directly based on the three-phase inverter. a , b Phase capacitor voltage v oa , v ob Calculate the state variables used in equations (7) and (8). Figure 5 The neural network shown contains an input layer, hidden layers, and an output layer, with 8, 7, and 6 neurons in each layer, respectively. The input to the neural network is... The output is ,in, , Capacitor current observation value The real and imaginary parts, , Capacitor voltage observation value The real and imaginary parts, , Let represent the real and imaginary parts of the output voltage of the three-phase bridge arm. , Let represent the real and imaginary parts of the measured capacitor voltage. The sample data used to train this neural network comes from... Figure 3 The three-phase inverter model shown is a predictive controller based on a capacitor current observer.

[0031] A3. Use neural networks to calculate state variables and implement model predictive control of the three-phase inverter based on capacitor current.

[0032] The state variables are calculated using a neural network, and model predictive control of the three-phase inverter based on capacitor current is implemented. The process is as follows: ① At the current moment... k The voltage of the three-phase inverter is obtained by two voltage sensors. a , b Phase capacitor voltage v oa , vob ② The capacitor voltage and capacitor current state variables used for output voltage prediction are calculated by the trained neural network. ③ Based on the capacitor voltage and capacitor current state variables, the predicted output voltage value is calculated by equations (7) and (8). ④ Based on the predicted output voltage value, the cost function is calculated by equation (9). Finally, the control quantity of the three-phase inverter is obtained, and... k +1 is implemented at time.

[0033] Based on the above process, for different load conditions of the three-phase inverter, the following results are obtained: Figures 6 to 9 The diagram shows the output voltage waveform of a three-phase inverter. Figure 6 This is a waveform diagram of the output voltage of a three-phase inverter under resistive load conditions. Figure 7 This is a waveform diagram of the output voltage of a three-phase inverter under no-load conditions. Figure 8 This is a waveform diagram of the output voltage of a three-phase inverter under resistive-inductive load conditions. Figure 9 The output voltage waveform of the three-phase inverter under nonlinear load conditions is shown above. As can be seen from the output voltage waveform, the designed sensorless model predictive control algorithm for the three-phase inverter based on neural networks can achieve tracking control of the output voltage of the three-phase inverter under different load conditions. Furthermore, it requires no current sensor or observer, and does not require matching neural network parameters based on load type, thus exhibiting wide adaptability.

[0034] Figure 10 The diagram shows the structure of a sensorless model predictive control device for a three-phase inverter, including two voltage sensors, an STM32H743 microcontroller, and a two-level three-phase voltage source inverter main circuit with an LC filter. The two voltage sensors are used to acquire the voltage readings of the three-phase inverter. a , b Phase capacitor voltage sampling signal. The STM32H743 microcontroller is responsible for... a , b The phase capacitor voltage sampling signal is subjected to analog-to-digital conversion (AD), and BP neural network calculation and three-phase inverter model predictive control based on capacitor current are implemented to finally generate 6 PWM pulse signals to control the three-phase inverter.

Claims

1. A sensorless model predictive control method for a three-phase inverter, characterized in that, Includes the following steps: A1. Construct a predictive controller for a three-phase inverter model based on a capacitor current observer; A2. Construct a sensorless model predictive control algorithm for a three-phase inverter based on neural networks; A3. Use neural networks to calculate state variables and implement model predictive control of the three-phase inverter based on capacitor current.

2. The sensorless model predictive control method for a three-phase inverter according to claim 1, characterized in that, In step A1, the three-phase inverter refers to a two-level three-phase voltage source inverter with an LC filter, and the model predictive controller refers to a controller that uses a finite control set model predictive control algorithm to achieve output voltage tracking control of the three-phase inverter. The process of constructing a three-phase inverter model predictive controller based on a capacitor current observer is as follows: A1.1 For the predictive control algorithm of the finite control set model of a three-phase inverter, an output voltage prediction model based on capacitor current is designed; With inductor current i s and capacitor voltage v o As state variables, the state equations of the three-phase inverter are obtained as follows: （1） In the formula, L and C These are the filter inductor and capacitor, respectively. r The equivalent resistance of the line. i s , v o , i o , v s These are the three-phase inductor current vector, the three-phase capacitor voltage vector, the three-phase load current vector, and the three-phase bridge arm output voltage vector, respectively. abc State variables in coordinate system ( x a , x b , x c )arrive αβ Vector in coordinate system x αβ The general formula for calculation is: （2） In the formula, the operator ; For a two-level three-phase voltage source inverter, define the switching function combination. S =( S a , S b , S c ),in S a , S b , S c The state of the upper bridge arm in the three-phase bridge arm of the inverter is represented by 1, which indicates that the upper bridge arm is on and the lower bridge arm is off, and 0 indicates that the upper bridge arm is off and the lower bridge arm is on. In equation (1) , V dc This is the DC side voltage; According to equation (1) by control cycle T s After discretization, the discrete state equations are obtained: （3） Ignoring the equivalent resistance of the circuit, let equation (1) r =0; combined with the capacitance current equation i c ( k )= i s ( k )- i o ( k ), to obtain based on capacitor current vector i c ( k ) and capacitor voltage vector v o ( k Three-phase inverter output voltage prediction model: （4） A1.2 Construct a capacitor current observer to obtain a three-phase inverter model predictive controller based on the capacitor current observer; The equations for constructing the capacitor current observer are as follows: （5） In the formula, the state variable x ( k )=[ v o ( k ) i c ( k )] T Observations of state variables ,matrix , , ,parameter k 1. k The formula for calculating 2 is: （6） The capacitor current observation value obtained from equation (5) Substituting into equation (4), we obtain the three-phase inverter output voltage prediction model based on capacitor current observations: （7） Furthermore, based on the finite control set model predictive control algorithm, we obtain... k +2 time interval and switching function combination S j Corresponding predicted output voltage value: （8） For a two-level three-phase voltage source inverter, in equation (8) j ={0,…,7} For combination with switching functions S j The corresponding three-phase bridge arm output voltage vector; For the tracking control of the output voltage of a three-phase inverter, the cost function is designed as follows: （9） In the formula, , These represent the real and imaginary parts of the output voltage reference value, respectively. , These are the predicted output voltage values. The real and imaginary parts; For equation (9), iterate through and calculate the cost functions corresponding to all 8 switch function combinations, and take the switch function combination that minimizes the cost function as the minimum cost function. k +1 is the control quantity of the three-phase inverter.

3. The sensorless model predictive control method for a three-phase inverter according to claim 1, characterized in that, In step A2, the constructed sensorless model predictive control algorithm for the three-phase inverter only requires two voltage sensors, eliminating the need for current sensors and observers. The algorithm employs a three-phase inverter model predictive controller based on a capacitor current observer, replacing the capacitor current observer with a neural network to calculate the capacitor current state variables. The constructed neural network is a three-layer BP neural network, used to fit the capacitor current observer and the three-phase capacitor voltage coordinate transformation process, directly based on the three-phase inverter. a , b Phase capacitor voltage v oa , v ob The state variables used in equations (7) and (8) are calculated; the constructed neural network includes an input layer, hidden layers, and an output layer, with 8, 7, and 6 neurons in each layer, respectively. The input of the neural network is... The output is ,in, , Capacitor current observation value The real and imaginary parts, , Capacitor voltage observation value The real and imaginary parts, , Let represent the real and imaginary parts of the output voltage of the three-phase bridge arm. , represents the real and imaginary parts of the capacitor voltage measurement; the sample data used to train this neural network comes from a three-phase inverter model predictive controller based on a capacitor current observer.

4. The sensorless model predictive control method for a three-phase inverter according to claim 1, characterized in that, In step A3, a neural network is used to calculate the state variables, and three-phase inverter model predictive control based on capacitor current is implemented. The process is as follows: ① At the current moment k The voltage of the three-phase inverter is obtained by two voltage sensors. a , b Phase capacitor voltage v oa , v ob ② The capacitor voltage and capacitor current state variables used for output voltage prediction are calculated by the trained neural network. ③ Based on the capacitor voltage and capacitor current state variables, the predicted output voltage value is calculated by equations (7) and (8). ④ Based on the predicted output voltage value, the cost function is calculated by equation (9). Finally, the control quantity of the three-phase inverter is obtained, and... k +1 is implemented at time.

5. A sensorless model predictive control device for a three-phase inverter for implementing the method of any one of claims 1-4, characterized in that... The device includes two voltage sensors, an STM32H743 microcontroller, and a two-level three-phase voltage source inverter main circuit with an LC filter; wherein, the two voltage sensors are used to acquire the voltage of the three-phase inverter. a , b Phase capacitor voltage sampling signal; STM32H743 microcontroller is responsible for... a , b The phase capacitor voltage sampling signal is converted from analog to digital, and BP neural network calculation and three-phase inverter model predictive control based on capacitor current are implemented to finally generate 6 PWM pulse signals to control the three-phase inverter.

Images