A highly efficient control method for motor system based on model predictive control

Abstract

The present invention provides a motor system efficient control method based on model predictive control, belonging to the field of motor control. Aiming at the shortcomings of the traditional motor system efficiency optimization method, the present invention integrates the minimum motor loss control algorithm and the multi-step model predictive control algorithm, designs a current loop controller, and optimizes the target to reduce the inverter switching frequency and motor loss. The motor current value that minimizes the motor loss is obtained by analyzing the motor loss model, and is used as the current reference value in the multi-step model predictive control to achieve minimum motor loss control. At the same time, the number of voltage vector switching times that determines the switching frequency of the inverter is added to the cost function as an indicator, and the corresponding weight coefficient is set. The optimal voltage vector is solved based on the multi-objective cost function to achieve multi-objective optimization such as inverter switching loss, motor loss and dynamic tracking performance of the current loop, so that the motor system can operate efficiently.

Description

Technical Field

[0001] The present invention belongs to the field of motor control, and in particular relates to a motor system efficient control method based on model predictive control. Background Art

[0002] In the design and optimization of motor drive systems, improving the overall efficiency of the system is a critical goal. This efficiency is mainly affected by the losses in two key parts: first, the inverter loss, and second, the motor loss.

[0003] In terms of inverter losses, conduction loss is an important component, which mainly depends on the resistance value of the power device in the on state. This on-resistance value is determined by the physical properties and materials of the power device itself, so it is necessary to select suitable power devices in the design stage to ensure low on-resistance and reduce conduction losses. On the other hand, switching loss also occupies an important position in inverter losses. Switching loss is not only related to the characteristics of the power device itself, but also closely related to the switching frequency of the device. The higher the switching frequency, the greater the loss generated during each switching process. Therefore, in order to reduce switching losses, it is possible to consider appropriately reducing the switching frequency of the device. However, reducing the switching frequency may cause the dynamic response of the system to slow down, so it is necessary to find a balance between reducing losses and maintaining the system response speed. By optimizing the switching frequency, the device loss can be effectively reduced without significantly affecting the system performance, thereby improving the power density of the entire system.

[0004] In terms of motor losses, the operating conditions of the motor have a direct impact on its efficiency. In order to make the motor run in the best condition, it is necessary to deeply study the loss characteristics of the motor under different operating conditions. By analyzing the changes in the efficiency of the motor under different loads, different speeds and different temperatures, the operating conditions that minimize the motor losses can be determined. For example, by optimizing the control strategy of the motor and adjusting the supply voltage and current of the motor, it can be ensured that the motor operates at the optimal efficiency point. In addition, the operating efficiency of the motor can be further improved by improving the design of the motor, such as optimizing the magnetic circuit design of the motor, reducing iron loss and copper loss, and other measures.

[0005] In summary, the overall efficiency of the motor drive system can be significantly improved through in-depth analysis and optimization of the losses of the inverter and the motor. This requires not only the selection of appropriate power devices and optimization of the switching frequency, but also careful research and adjustment of the motor's operating conditions to ensure that the motor operates in the best condition, thereby achieving an efficient and energy-saving motor drive system. Summary of the invention

[0006] In view of the shortcomings of traditional efficiency optimization methods, the present invention proposes a high-efficiency control method for motor systems based on model predictive control. Based on a multi-step model predictive control algorithm, the present invention simultaneously optimizes efficiency from two aspects: inverter switching frequency and motor loss. A reasonable cost function is constructed in combination with the mathematical model of the motor current loop. The inverter switching loss, motor loss and dynamic tracking performance of the current loop are comprehensively considered to solve the optimal voltage vector and achieve high-efficiency operation of the motor system.

[0007] The technical solution adopted by the present invention is:

[0008] An efficient control method for a motor system based on model predictive control for current loop control includes the following steps:

[0009] Step 1, based on the loss model of the motor, calculate the d-axis reference current that minimizes the motor loss;

[0010] Step 2: Based on the motor mathematical model and the current state of the motor, predict the state variables of multiple steps in the future, obtain a set of voltage vector sequences corresponding to each set of state variables, and form multiple sets of voltage vector sequences;

[0011] Step 3, substitute the d-axis reference current, q-axis reference current, the predicted value of the state variable corresponding to each group of voltage vector sequences, and the number of inverter switches into the cost function, calculate the voltage vector sequence that minimizes the cost function value, that is, the optimal voltage vector sequence, output the first voltage vector in the optimal voltage vector sequence, and make the motor run according to the expected target.

[0012] Furthermore, in step 1, the d-axis reference current is i dref :

[0013]

[0014] Where, ω is the motor speed, L d is the d-axis inductance, ψ f is the permanent magnet flux, R s is the motor stator winding resistance, R i is the iron loss equivalent resistance.

[0015] Furthermore, the specific method of step 2 is:

[0016] Step 201, using the motor mathematical model, based on the three-phase current and the encoder angle, the state variable i at the current moment is calculated through coordinate transformation. d (k) and i q (k);

[0017] Step 202, based on the state variables at the current moment, respectively calculate the state variables at the next moment when using 8 different voltage vectors, and obtain the state prediction values ​​at the next moment corresponding to the 8 different voltage vectors:

[0018]

[0019] In the formula, k represents the current moment, k+1 represents the next moment, the superscript p represents the predicted value, and T s is the switching period, R s is the motor stator winding resistance, L d is the d-axis inductance, L q is the q-axis inductance, i d 、i q are the d-axis current and the q-axis current, respectively, f is the permanent magnet flux, u d (k),u q (k) are the d-axis voltage and q-axis voltage respectively, which are obtained by coordinate transformation of the adopted voltage vector, and ω is the motor speed;

[0020] Step 203, taking the k+1 moment as the current moment, returning to step 202 to calculate the state prediction value of the next moment again; through iterative calculation, the prediction of the state variables for the next N steps is realized, N>1;

[0021] Step 204: The state prediction values ​​calculated by using different voltage vectors in each step are combined into a set of state variables. Each set of state variables corresponds to a set of voltage vector sequences. A total of 8 N A set of state variables and voltage vector sequences that correspond to each other.

[0022] Furthermore, in step 3, the cost function is:

[0023]

[0024] In the formula, J c is the cost function value, N is the number of prediction steps, n is the accumulated variable, k is the current moment, i d 、i q are the d-axis current and q-axis current respectively. The superscript p indicates the predicted value. dref The d-axis reference current that minimizes motor loss, i qref is the q-axis reference current, λ d is the error weight coefficient of the d-axis reference current, λ q is the error weight coefficient of the q-axis reference current, λ s is the weight coefficient of the inverter switching times, where λ q =1,λ d The value range is 0.2<λ d <1,λs The value range is 0<λ s <1;N s is the number of inverter switching times, that is, the total number of switching changes of each adjacent voltage vector in each voltage vector sequence.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] Traditional algorithms only consider reducing the inverter switching frequency and reducing the motor loss when designing high-efficiency control of motor systems. The present invention adopts a multi-step model predictive control algorithm and designs a multi-objective optimization cost function that considers switching loss and motor loss. It integrates the two factors that affect the high-efficiency operation of the motor system, thereby further improving the operating efficiency of the motor system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the principle of an efficient control method for a motor system based on model predictive control;

[0028] Figure 2 Flowchart of an efficient control method for a motor system based on model predictive control. DETAILED DESCRIPTION

[0029] The present invention is further described in detail below with reference to the accompanying drawings.

[0030] An efficient control method for motor systems based on model predictive control for current loop control, such as Figure 1 As shown, the following steps are included:

[0031] (1) Calculation of minimum motor loss current: Based on the motor three-phase current and encoder angle, calculate the minimum loss reference current i dref .

[0032] This step is based on the loss model of the motor. By analyzing the relationship between copper loss, iron loss and motor parameters, the motor excitation flux or current is controlled to make the motor work in the minimum loss state while meeting the load demand, thereby reducing the motor loss and improving the motor efficiency.

[0033] (2) Multi-step model prediction: Based on the motor three-phase current and encoder angle, the state variable values ​​at the future k+N moments are predicted, and multiple sets of voltage vector sequences are obtained accordingly.

[0034] This step builds a multi-step prediction model based on the motor mathematical model, and predicts the state variables of the next N steps according to the current state of the motor. In the prediction process of each step, 8 voltage vectors (000, 001, 010, 011, 100, 101, 110, 111, corresponding to the 8 switching states of the three-phase inverter) are selected in turn to calculate the state variables. After N steps of prediction, multiple groups of voltage vector sequences are formed. Each group of sequences corresponds to a group of state prediction values ​​and the number of voltage vector switching times, among which the number of voltage vector switching times directly affects the switching frequency of the inverter. Therefore, the number of voltage vector switching times is used as an indicator for performance evaluation, and the switching frequency of the inverter can be reduced by selecting the optimal voltage vector.

[0035] (3) Multi-objective optimization cost function: combining the state prediction value corresponding to each set of voltage vector sequences and the number of voltage vector switching, the minimum loss reference current i dref and torque reference current i qref The optimal voltage vector is calculated to control the torque and flux of the motor.

[0036] In view of the shortcomings of the traditional motor system efficiency optimization method, this method integrates the minimum motor loss control algorithm and the multi-step model predictive control algorithm to design a current loop controller. The optimization goal is to reduce the inverter switching frequency and motor loss and ensure the dynamic tracking performance of the current loop. Among them, through the analysis of the motor loss model, the motor current value that minimizes the motor loss is obtained, which is used as the current reference value in the multi-step model predictive control to achieve minimum motor loss control; at the same time, the number of voltage vector switching times that determines the switching frequency of the inverter is added to the cost function as an indicator, and the corresponding weight coefficient is set. Finally, based on the multi-objective cost function, the optimal voltage vector is solved to achieve multi-objective optimization such as inverter switching loss, motor loss and dynamic tracking performance of the current loop, so that the motor system can operate efficiently.

[0037] Figure 2 The figure shows the workflow of multi-step current prediction control based on minimum motor loss in one control cycle. In the figure, time k represents the current time. The specific process is as follows:

[0038] (1) Calculate the minimum loss reference current i at the current moment based on the motor state dref ;

[0039] (2) Combine the current motor state information to perform multi-step model prediction, that is, use the motor mathematical model to calculate the state variables at the current moment (time k);

[0040] (3) Based on the state variables at the current moment, the state variables at the next moment (time k+1) when using different voltage vectors are calculated respectively to obtain the state prediction value at time k+1. Then, taking time k+1 as the current moment, the state prediction value at time k+2 is calculated using the same method. According to the above method, through iterative calculation, the state variable values ​​within the next N steps are predicted to form multiple sets of voltage vector sequences;

[0041] (4) Substitute the current reference value and the state prediction value and switching times corresponding to each group of voltage vector sequences into the multi-objective optimization cost function, calculate the voltage vector sequence that minimizes the cost function value, which is the optimal voltage vector sequence, and finally output the first voltage vector in the sequence to make the motor run according to the expected target.

[0042] Next, for Figure 2 The contents of minimum loss current calculation, multi-step model prediction and multi-objective optimization cost function are described in detail:

[0043] 1) Calculation of minimum loss current.

[0044] This method mainly considers the copper loss P Cu and iron loss P Fe , the sum of copper loss and iron loss is defined as the optimizable loss P opt , P opt It can be expressed as:

[0045]

[0046] Where ω is the motor speed, i d 、i q , L d , L q are the dq axis current and inductance respectively, ψ f is the permanent magnet flux, R s is the motor stator winding resistance, R i is the iron loss equivalent resistance.

[0047] Since the permanent magnet flux and iron loss equivalent resistance are related to the motor structure, and the q-axis current is related to the motor control performance, the loss minimization design is achieved by changing the d-axis current without affecting the motor control performance.

[0048] Because P opt with i d It forms a quadratic function relationship, and the coefficient of the quadratic term is positive, so let:

[0049]

[0050] At this time, P optThe minimum value of the d-axis current at the minimum loss is:

[0051]

[0052] 2) Multi-step model prediction.

[0053] The prediction model can be constructed as follows:

[0054]

[0055] Where: k represents the current moment, k+1 represents the next moment, the superscript p represents the predicted value, T s is the switching period, R s is the motor stator winding resistance, L d is the d-axis inductance, L q is the q-axis inductance, i d 、i q are the d-axis current and the q-axis current, respectively, f is the permanent magnet flux, u d (k) and u q (k) are the d-axis voltage and q-axis voltage respectively, which are obtained by coordinate transformation of the adopted voltage vector, and ω is the motor speed.

[0056] Based on the above prediction model, the dq-axis stator current value at time k can be used to calculate the predicted value of the dq-axis stator current at time k+1 under the action of any voltage vector. When performing multi-step prediction, after completing one step of prediction, the predicted value of the state variable at time k+1 is used as the current value. Using the above formula, the state prediction value at time k+N can be obtained by repeatedly calculating backward, and finally multiple groups of voltage vector sequences are obtained, each group of voltage vector sequences corresponds to a group of state prediction values ​​and the number of voltage vector switching times.

[0057] 3) Multi-objective optimization cost function.

[0058] In multi-step prediction, after predicting the state variables in the next N moments, it is necessary to analyze the predicted values ​​of each group of state variables and evaluate the application effects of different voltage vector sequences. In this process, the cost function needs to be designed. The error between the current predicted value and the given value under the action of each group of voltage vectors is squared, and then averaged and multiplied by the corresponding weight coefficient to adjust the proportion of each indicator. At the same time, the switching frequency is added as an indicator to the cost function, and finally the optimal voltage vector sequence that minimizes the cost function is selected. When designing the cost function, the first thing to consider is to achieve the controlled object's response to the torque current reference value i qref The tracking of the current is the following performance, followed by improving the performance indicators as much as possible, such as reducing the switching frequency and reducing the loss. According to the above analysis, the cost function is selected as follows:

[0059]

[0060] Where N is the number of prediction steps, N s is the number of inverter switches, that is, the total number of switches of each adjacent voltage vector in each voltage vector sequence. dref is the d-axis reference current obtained by the motor minimum loss algorithm, i qref is the torque reference current, λ d is the d-axis current error weight coefficient, λ q is the q-axis current error weight coefficient, λ s is the weight coefficient of the inverter switching times. The weight coefficient represents the importance of each indicator in the electric drive system. When the weight coefficient is selected for the first time, the principle is to achieve a numerical balance of each indicator after weighting. Among them, λ can be selected q =1,λ d The value range is 0.2<λ d <1,λ s The value range is 0<λ s <1, according to the system's requirements on loss d and λ s Make adjustments.

[0061] Substitute the relevant parameters into the cost function J c By enumerating, we can get J under different voltage vector sequences. c value, select J c The minimum voltage vector sequence outputs the first voltage vector in the optimal voltage vector sequence and converts it into the control signal of the inverter, thereby completing the entire control process.

Claims

1. A highly efficient control method for a motor system based on model predictive control for current loop control, characterized in that: The following steps are involved: Step 1, based on the loss model of the motor, calculate the d-axis reference current that minimizes the motor loss; Step 2: Based on the motor mathematical model and the current state of the motor, predict the state variables of multiple steps in the future, obtain a set of voltage vector sequences corresponding to each set of state variables, and form multiple sets of voltage vector sequences; Step 3, substitute the d-axis reference current, q-axis reference current, the predicted value of the state variable corresponding to each group of voltage vector sequences, and the number of inverter switches into the cost function, calculate the voltage vector sequence that minimizes the cost function value, that is, the optimal voltage vector sequence, output the first voltage vector in the optimal voltage vector sequence, and make the motor run according to the expected target.

2. The method for efficiently controlling a motor system based on model predictive control according to claim 1, characterized in that: In step 1, the d-axis reference current is i dref : Where, ω is the motor speed, L d is the d-axis inductance, ψ f is the permanent magnet flux, R s is the motor stator winding resistance, R i is the iron loss equivalent resistance.

3. The method for efficiently controlling a motor system based on model predictive control according to claim 2, characterized in that: The specific method of step 2 is: Step 201, using the motor mathematical model, based on the three-phase current and the encoder angle, the state variable i at the current moment is calculated through coordinate transformation. d (k) and i q (k); Step 202, based on the state variables at the current moment, respectively calculate the state variables at the next moment when using 8 different voltage vectors, and obtain the state prediction values ​​at the next moment corresponding to the 8 different voltage vectors: In the formula, k represents the current moment, k+1 represents the next moment, the superscript p represents the predicted value, and T s is the switching period, R s is the stator winding resistance of the motor, L d is the d-axis inductance, L q is the q-axis inductance, i d 、i q are the d-axis current and the q-axis current, respectively, f is the permanent magnet flux, u d (k),u q (k) are the d-axis voltage and q-axis voltage respectively, which are obtained by coordinate transformation of the adopted voltage vector, and ω is the motor speed; Step 203, taking the k+1 moment as the current moment, returning to step 202 to calculate the state prediction value of the next moment again; through iterative calculation, the prediction of the state variables for the next N steps is realized, N>1; Step 204: The state prediction values ​​calculated by using different voltage vectors in each step are combined into a set of state variables. Each set of state variables corresponds to a set of voltage vector sequences. A total of 8 N A set of state variables and voltage vector sequences that correspond to each other.

4. The method for efficiently controlling a motor system based on model predictive control according to claim 3, characterized in that: In step 3, the cost function is: In the formula, J c is the cost function value, N is the number of prediction steps, n is the accumulated variable, k is the current moment, i d 、i q are the d-axis current and q-axis current respectively. The superscript p indicates the predicted value. dref The d-axis reference current that minimizes motor loss, i qref is the q-axis reference current, λ d is the error weight coefficient of the d-axis reference current, λ q is the error weight coefficient of the q-axis reference current, λ s is the weight coefficient of the inverter switching times, where λ q =1,λ d The value range is 0.2<λ d <1,λ s The value range is 0<λ s <1;N s is the number of inverter switching times, that is, the total number of switching changes of each adjacent voltage vector in each voltage vector sequence.

Images