An adaptive optimal model predictive control method for mmmc

Abstract

The application discloses an adaptive optimal model predictive control method of MMC (Modular Multilevel Converter). The application is based on the differentiated requirements of the steady state stage and the transient state stage of the MMC system, and adaptively switches the optimization strategy of the model predictive control: in the steady state stage, a local optimization strategy is adopted, only the adjacent control quantity of the control quantity at the last moment is evaluated, so that the cycle calculation amount is significantly reduced, and two voltage levels with different current slopes are output in each control period, and the current control error minimization is realized through linear combination; in the transient state stage, a global optimization strategy based on a grouping optimization mechanism is adopted, the inter-group optimization is carried out in small groups first, the approximate interval of the optimal level is determined, and then the optimal level is determined by further optimization in the approximate interval. While guaranteeing the tracking accuracy and dynamic performance of the system, the application greatly reduces the calculation burden of the processor, and makes it have application feasibility in high-voltage power transmission occasions with a large number of sub-modules.

Description

Technical Field

[0001] This invention belongs to the field of modular multilevel matrix converter (MMMC) control technology, specifically an adaptive optimization model predictive control method for MMMC. Background Technology

[0002] In offshore wind power transmission technology, low-frequency AC transmission (LFAC) demonstrates good techno-economic efficiency because it avoids the capacitance effect problem of high-voltage AC transmission and eliminates the need for expensive offshore converter stations. Modular multilevel matrix converters (MMMCs, or M3Cs) are a topology capable of direct three-phase AC / AC conversion, eliminating the DC link and some bridge arms, resulting in a simpler structure. More importantly, MMMCs can effectively adapt to low-frequency, high-power operating scenarios by adjusting the power balance between bridge arms, thus their application potential in LFAC systems has attracted attention.

[0003] Currently, traditional control strategies for M3C often employ PI / PR linear controllers, using closed-loop design for a single objective and cascading structures to achieve multi-objective control. While this method can achieve basic control effects, the complex and nested control loop structure leads to problems such as low tracking accuracy and slow dynamic response. Furthermore, the large number and coupling of PI / PR linear controllers result in complex parameter tuning and significant debugging difficulties.

[0004] Model predictive control (MPC) predicts changes in state variables based on a dynamic model of the system, solves for the optimal control variable corresponding to the cost function, and applies it to the system. This process is repeated at the next time step to achieve rolling optimization. MPC has advantages such as simple structure, flexible control, and fast dynamic response, and it does not require complex parameter tuning, making it suitable for modular multilevel converters. However, traditional MPC needs to traverse all possible level combinations, resulting in a high computational burden. Although some studies have proposed evaluating only combinations adjacent to the current level in high-voltage applications to reduce computation, this sacrifices the controller's dynamic performance. Furthermore, traditional MPC outputs only a single level per control cycle, and its control accuracy needs further improvement. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide an adaptive optimization model predictive control (MPC) method for MMMC. Based on the differentiated requirements of the steady-state and transient phases of the system, the optimization strategy of model predictive control is adaptively switched. While ensuring the tracking accuracy and dynamic performance of the system, it effectively reduces the computational burden of the processor, making it feasible for application in high-voltage power transmission applications with a large number of sub-modules.

[0006] To this end, the present invention adopts the following technical solution: an adaptive optimization model predictive control method for MMMC, which is based on the differentiated requirements of the steady-state and transient phases of the MMMC system and adaptively switches the optimization strategy of model predictive control: In the steady-state phase, a local optimization strategy is adopted, which only evaluates the neighboring control variables of the control variable at the previous moment, and outputs two voltage levels with different current slopes in each control cycle, and minimizes the current control error by linearly combining them; In the transient phase, a global optimization strategy based on a group optimization mechanism is adopted, which first performs inter-group optimization in groups to determine the approximate range of the optimal level, and then further optimizes within the approximate range to determine the optimal level.

[0007] Furthermore, the adaptive optimization model predictive control method described herein belongs to finite set model predictive control, and the control target is the arm current; the identification criteria for the steady-state stage and the transient stage are: whether the control reference value of the MMMC system is maintained within the allowable threshold. If so, the MMMC system is in the steady-state stage; if not, the MMMC system is in the transient stage.

[0008] Furthermore, the aforementioned adaptive optimization model predictive control method is applied to an MMMC system comprising nine arms, each arm consisting of... N The system consists of a submodule cascaded with a bridge arm inductor, and its discrete mathematical model is as follows:

[0009] In the formula, , indicating in The slope of the current under action; For the current moment k bridge arm xy Output voltage; , Each represents the current time. k The input and output voltages of the MMMC system x = a, b, c, y = u, v, w; and These are the inductance values ​​of the bridge arms and their parasitic resistances, respectively. T s The sampling period; , Each represents the current time. k bridge arm xy Output current and The predicted current value at the next moment.

[0010] Furthermore, when the MMMC system is in steady state, the optimal control quantity varies little between each control cycle. Therefore, a local optimization strategy is implemented: the maximum change in the number of submodules is limited to a certain value. The finite set of control at this time for:

[0011] In the formula, For bridge arm xy The average value of the capacitor voltage in the submodule; For the previous moment k -1 bridge arm xy Output voltage.

[0012] Furthermore, within each control cycle, from a finite set of controls... Two optimal voltage levels with different current slopes are selected as outputs and linearly combined to improve current control error. The expression for the current prediction value is updated as follows:

[0013] In the formula, , These are the current slopes corresponding to the first output level and the last output level, respectively; The duration of the initial output level.

[0014] Furthermore, the first output voltage in each control cycle continues the last output level of the previous cycle to reduce the number of voltage level switching; simultaneously, the signs of the tracking error values ​​generated by the two output levels should be opposite to eliminate tracking error; the duration of the first output level... The calculation is as follows:

[0015] In the formula, ,and ; To obtain from a finite set of controls The optimal combination of voltage level and first output level is selected using the following first cost function. Conduct an assessment:

[0016] In the formula, for The reference current, , These are all weighting coefficients in the first cost function. The first term is used to evaluate tracking accuracy; the second term is to prevent... du / dt The first is too large; the second is to reduce the fluctuation of the average switching frequency; the last two values ​​are small and will not affect the judgment of current tracking error.

[0017] Furthermore, when the control reference value changes beyond the threshold range, the MMMC system will enter a transient operating state. In order to achieve rapid current tracking and avoid current oscillation, it will switch to a global optimization strategy, expanding the finite control set to all levels and executing a group optimization mechanism, as follows: Step 1: Sample and initialize the key system variables; Step two, connect each bridge arm N The sub-modules are divided into G Groups, each group contains the following number of submodules. M j ; Step 3: Treat each group as a whole and generate a control set for candidate levels. ; Step four, based on the control set in step three. Inter-group optimization is performed, and the predicted current value, current error value, and corresponding second cost function are calculated under the action of each control variable. ; Step 5: Record the optimal error value Optimal cost function value and optimal control quantity ; Step six, set the first judgment condition: if the optimal cost function value at this time is... Less than minimum error ε ,or If the corresponding current error values ​​have the same sign, the optimization result is considered to be accurate enough, and the result is directly output as the final control command, then proceed to step nine; otherwise, proceed to step seven. Step 7: Set the second judgment condition: Based on the optimal control quantity and its control error, determine the interval where the global optimal level is located, and update the control set as follows. ; Step 8: Update the control set based on the results obtained in Step 7. Perform intra-group optimization and calculate The current prediction value and second cost function under the action of each control quantity are calculated, and the optimal control quantity corresponding to the second cost function is output. Step nine: After completing the calculation for one control cycle, return to the starting point and wait for the next sampling cycle to begin.

[0018] Furthermore, in step three, under the capacitor voltage balancing control strategy, the average value of the submodule capacitor voltage of the bridge arm is set to... Therefore, each group is treated as a whole, generating a control set of candidate levels for inter-group optimization. as follows:

[0019] In the formula, Round down to the nearest integer; . Furthermore, in step seven, if the current optimal control quantity is... The first control variable, or its corresponding optimal error value Error value with the previous control quantity If the signs are the same, it indicates that the global optimal solution is located at the current optimal control quantity. With the next control quantity For submodule groups between these groups, the finite control set should be updated as follows:

[0020] Otherwise, the global optimal solution will be located at the current optimal control quantity. Compared with the previous control quantity For the submodule groups between them, the finite control set should be updated as follows: .

[0021] Furthermore, the computational cost of global optimization is reduced to (2 G +1)~(2 G +max( M j The inter-group optimization phase only requires 2 G +1 calculation, the intra-group optimization phase will increase by at most max( M j The calculation is performed in 1-1 steps, and the two stages can be connected by simple logical judgments, thus effectively reducing the computational complexity while ensuring the optimization accuracy. To minimize the computational load, the number of groups is... G and the number of submodules within each group M j The value of should follow the following: 1) Order and ; 2) Calculate separately and ; 3) Select The corresponding number of groups is used as the group number. G Number of submodules M j According to the number of groups G Calculated.

[0022] The beneficial effects of this invention are as follows: 1. The adaptive optimization model predictive control method proposed in this invention employs a local optimization strategy when the system is in steady state, and the computational cost per control cycle is only 2. Δmax This method is independent of the number of submodules, significantly reducing the computational burden on the processor. It outputs two optimal voltage levels with different current slopes in each control cycle and minimizes the current control error by linearly combining them. Furthermore, this method effectively reduces fluctuations in the average switching frequency and improves the harmonic distribution of the output current.

[0023] 2. The adaptive optimization model predictive control method proposed in this invention, when the system is in the transient phase, adopts a global optimization strategy based on a group optimization mechanism, reducing the computational cost of each bridge arm to (2). G +1)~(2 G +max( M j This improvement makes the global optimization strategy feasible for practical application in high-voltage transmission scenarios with a large number of sub-modules.

[0024] 3. The adaptive optimization model predictive control method proposed in this invention is based on the differentiated requirements of the steady-state and transient stages of the system, and adaptively switches the optimization strategy of model predictive control. While ensuring the system's tracking accuracy and dynamic performance, it effectively reduces the computational burden on the processor. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a flowchart of an adaptive optimization model predictive control method for MMMC according to the present invention; Figure 2 This is a topology diagram of the MMMC system of the present invention; Figure 3 This is a schematic diagram of the current path for the bridge arm dual output level combination during the steady-state phase of the present invention. Figure 4 This is a flowchart illustrating the adaptive optimization model predictive control method for the steady-state phase of this invention. Figure 5 This is a flowchart illustrating the process of the grouped global optimization method in the transient phase of this invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0028] This invention relates to an adaptive optimization model predictive control (MPC) method for MMMC, such as... Figure 1 As shown, based on the differentiated requirements of the steady-state and transient phases of the MMMC system, the optimization strategy of adaptive switching model predictive control is as follows: In the steady-state phase, a local optimization strategy is adopted, which only evaluates the neighboring control variables of the control variable at the previous moment, thereby significantly reducing the cyclic computation. At the same time, two voltage levels with different current slopes are output in each control cycle, and the current control error is minimized by linearly combining them, which also effectively suppresses the fluctuation of the average switching frequency. In the transient phase, a global optimization strategy based on a group optimization mechanism is adopted. First, inter-group optimization is performed on a group basis to determine the approximate range of the optimal level, and then further optimization is performed within this approximate range to determine the optimal level.

[0029] To predict the state variables of an MMMC system, a mathematical model of the MMMC system must first be established. Figure 2 The topology of the MMMC system is shown, which consists of 9 arms, each arm being composed of... N Each submodule is cascaded with a bridge arm inductor. Each bridge arm is connected to the low-frequency side of the power grid and the mains frequency side of the power grid, respectively, causing the bridge arm modulation wave to exhibit alternating positive and negative AC characteristics. Therefore, the submodule generally adopts a full-bridge structure. The bridge arm inductor is mainly used to suppress circulating currents caused by power imbalances between bridge arms and to reduce the impact of current on circuit equipment during system faults.

[0030] According to Kirchhoff's laws, the mathematical model of MMMC in Cartesian coordinates can be established as follows: (1) In the formula, , ( x =a, b, c) are the system input voltage and current, respectively; , ( y =u, v, w) represent the system output voltage and current, respectively; , bridge arms xy Voltage and current; L arm and R arm These are the inductance values ​​of the bridge arms and their parasitic resistances, respectively. The voltage difference between the input and output neutral points is the voltage difference when the system operates in three-phase symmetrical mode. =0.

[0031] For the bridge arm current control loop, this invention adopts an adaptive optimization model predictive control method. This method is based on a finite set model predictive control framework. It constructs a control set of bridge arm levels and selects the output level closest to the target value from it to act on the bridge arm, thereby achieving the control effect. Therefore, equation (1) can be simplified to the following form: (2) Equation (2) describes the voltage of the bridge arm. Under action, the bridge arm current The slope of the change. Considering that the sampling period is usually small, it can be discretized using the forward Euler method, thus obtaining an approximate nonlinear discrete mathematical model at that moment: (3) In the formula, T s The sampling period; This is the predicted value of the bridge arm current at the current moment for the next moment.

[0032] To evaluate whether the predicted value is close to the target value, a suitable cost function needs to be constructed, and the optimal bridge arm level needs to be selected based on this function. The constructed cost function is as follows: (4) In the formula, This represents the target value for the bridge arm current.

[0033] To address the technical problems of traditional MPC, this invention employs an adaptive optimization MPC method, which configures dedicated MPC strategies based on the differentiated requirements of the system's steady-state and transient phases. The following explanation will use a network-based MMMC as an example: 1) Steady-state stage If the power reference value remains within the allowable threshold, the MMMC system will maintain steady-state operation, where the optimal control variable varies little between control cycles. Therefore, MPC executes a local optimization strategy, limiting the maximum change in the number of submodules to Δ. max The finite set of control at this point is: (5) In the formula, For bridge arm xy The average value of the capacitor voltage in the submodule. To appropriately increase the search range, Δ can be set. max = 3.

[0034] Although the optimal output level minimizes the following error, the error still exists; and this error further increases as the sampling period increases. As shown in equation (3), different output levels can change the current change slope. Based on this, the present invention outputs two voltage levels with different current slopes in each control cycle, and improves the current control error by linearly combining these two levels, such as... Figure 3 As shown. Therefore, the expression for the predicted current value is updated to: (6) In the formula, , These are the current slopes corresponding to the first output level and the last output level, respectively; The duration of the initial output level.

[0035] It should be noted that, in order to reduce the number of voltage level switching times, the first output level in each control cycle in this embodiment continues the last output level of the previous cycle. At the same time, to eliminate tracking errors, the signs of the tracking error values ​​generated by the two output levels should be opposite.

[0036] Combining equations (4) and (6), the first level action time The calculation is as follows: (7) In the formula, ,and .

[0037] Depend on Figure 3 It can be seen that when the tracking error is zero, the combination of levels satisfying the condition may not be unique. Therefore, an additional evaluation index needs to be introduced into the cost function to select the optimal combination. The updated first cost function... The update is as follows: (8) (9) In the formula: , All are first cost functions J st The weighting coefficients in the equation. The first term is to prevent... du / dt The first option is too large; the second option is to reduce the fluctuation of the average switching frequency. Note that the values ​​of these two options are relatively small and will not affect the judgment of current tracking error.

[0038] Figure 4 This describes the workflow of the method described in this invention during the steady-state phase. The system first processes the bridge arm current. Input voltage Output voltage The relevant variables are sampled. Then, the algorithm enters a level combination traversal loop: the algorithm sequentially reads the current control set. Candidate levels and A combination is formed, and its corresponding current slope is calculated according to equation (2). Then, the first-level duration of the combination is calculated according to equation (7). The control performance is further evaluated using equation (8). After the loop is completed, the combination with the smallest cost function is selected from all candidate combinations as the optimal control output. This process is repeated in the next sampling period to achieve rolling optimization control.

[0039] 2) Transient phase When the power reference value changes beyond the threshold range, the MMMC system enters a transient operating state. If the MPC still uses a local optimization strategy at this time, the system's dynamic performance will significantly degrade. To achieve rapid current tracking and avoid current oscillations, the MPC switches to a global optimization strategy, expanding the limited control set to all levels.

[0040] (10) However, traditional global optimization requires traversing all level combinations, resulting in a huge computational burden, especially in high-voltage transmission scenarios with hundreds of submodules, making real-time implementation difficult. Therefore, this invention proposes a grouped global optimization method: grouping each bridge arm... N The sub-modules are divided into G Groups, each group contains the following number of submodules: M j This method first performs inter-group optimization on a small-group basis; then, it determines the approximate range of the optimal level; finally, it further optimizes within this range to determine the optimal level.

[0041] Figure 5 The flowchart for the grouped global optimization method is as follows: (1) Step 1, the system first checks the bridge arm current. Input voltage Output voltage Sampling was performed on key variables.

[0042] (2) Step 2, connect each bridge arm N The sub-modules are divided into G Groups, each group contains the following number of submodules. M j .

[0043] (3) Step 3, under the capacitor voltage equalization control strategy, let the average value of the capacitor voltage of the bridge arm submodule be... Therefore, by treating each group as a whole, a control set can be generated for inter-group optimization. as follows: (11) In the formula: (Round down); .

[0044] (4) Step 4: Perform inter-group optimization. According to equations (4) and (11), traverse the control set. And calculate the predicted current value, error value, and corresponding second cost function under the action of each control variable. .

[0045] (5) Step 5: Record the optimal error value Optimal cost function value and optimal control quantity .

[0046] (6) Step 6, according to equations (3) and (4), it can be seen that if the optimal cost function at this time is < ε (set up ε =1×10 -8 ),or If the corresponding current error values ​​have the same sign, it is considered that the inter-group optimization result is accurate enough, and the result is directly output as the final control command, then proceed to step 9; otherwise, proceed to step 7.

[0047] (7) Step 7, if the current optimal control quantity corresponds to the first control quantity in equation (11), or its corresponding optimal error value Error value with the previous control quantity If the signs are the same, it indicates that the global optimal solution is located at the current optimal control quantity. With the next control quantity Submodule groups between. At this point, the finite control set should be updated to: (12) Otherwise, the global optimal solution will be located at the current inter-group optimal control quantity. Compared with the previous control quantity The submodule groups between them. The finite control set should then be updated as follows: (13) (8) Step 8: Perform intra-group optimization. According to equations (4), (12) and (13), iterate and update the control set. and calculate Predicted current values ​​and second cost function under the action of various control variables It also outputs the optimal control quantity corresponding to the second cost function. .

[0048] (9) Step 9: Complete the calculation of one control cycle, output the control quantity, and return the process to the starting point to wait for the next sampling cycle to begin.

[0049] As can be seen from equation (10), traditional global MPC requires traversing all level combinations, and the computational load for each bridge arm is as high as 2. N +1, which increases sharply with the number of submodules, resulting in a heavy computational burden. In contrast, the adaptive optimization MPC proposed in this invention reduces the computational cost of global optimization to (2) G +1)~(2 G +max( M j Specifically, the inter-group optimization phase only requires 2... G +1 calculation, the intra-group optimization phase will increase by at most max( M j It requires only 1 calculation, and the two stages can be connected by simple logical judgment, thus effectively reducing the computational complexity while ensuring the optimization accuracy.

[0050] To minimize the computational cost of this strategy, the number of groups... G and the number of modules in each group M j The value of should follow these steps: (1) Let (round down) and (Round up); (2) Calculate separately and ; (3) Select The corresponding number of groups is used as the actual number of groups. G , M j It can then be calculated.

[0051] For example: Assume the number of bridge arm submodules N =110, then the computation amount of each bridge arm in the traditional global MPC is as high as 221 times, while the computation amount of each bridge arm is only 30 times at most after adopting the method proposed in this invention, which significantly reduces the computational burden and thus improves the overall performance of the algorithm.

[0052] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. It will be apparent to those skilled in the art that various modifications can be made to the above embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. An adaptive optimization model predictive control method for MMMC, characterized in that, Based on the differentiated requirements of the steady-state and transient phases of the MMMC system, the optimization strategy of adaptive switching model predictive control is as follows: In the steady-state phase, a local optimization strategy is adopted, which only evaluates the neighboring control variables of the control variable at the previous moment, and outputs two voltage levels with different current slopes in each control cycle, and minimizes the current control error by linearly combining them; In the transient phase, a global optimization strategy based on a group optimization mechanism is adopted, which first performs inter-group optimization on a group basis to determine the approximate range of the optimal level, and then further optimizes within the approximate range to determine the optimal level. This method is applied to an MMMC system containing nine bridge arms, each bridge arm consisting of... N The system consists of a submodule cascaded with a bridge arm inductor, and its discrete mathematical model is as follows: In the formula, , indicating in The slope of the current under action; For the current moment k bridge arm xy Output voltage; , Each represents the current time. k The input and output voltages of the MMMC system x = a, b, c, y = u, v, w; and These are the inductance values ​​of the bridge arms and their parasitic resistances, respectively. T s The sampling period; , Each represents the current time. k bridge arm xy Output current and The predicted current value at the next moment.

2. The adaptive optimization model predictive control method according to claim 1, characterized in that, This method belongs to finite set model predictive control, and the control target is the arm current. The identification criteria for the steady-state stage and the transient stage are: whether the control reference value of the MMMC system is maintained within the allowable threshold. If so, the MMMC system is in the steady-state stage; if not, the MMMC system is in the transient stage.

3. The adaptive optimization model predictive control method according to claim 1, characterized in that, When the MMMC system is in steady state, it executes a local optimization strategy: limiting the maximum change in the number of submodules to a certain value. The finite set of control at this time for: In the formula, For bridge arm xy The average value of the capacitor voltage in the submodule; For the previous moment k -1 bridge arm xy Output voltage.

4. The adaptive optimization model predictive control method according to claim 3, characterized in that, Within each control cycle, from a finite set of controls Two optimal voltage levels with different current slopes are selected as outputs and linearly combined to improve current control error. The expression for the current prediction value is updated as follows: In the formula, , These are the current slopes corresponding to the first output level and the last output level, respectively; The duration of the initial output level.

5. The adaptive optimization model predictive control method according to claim 4, characterized in that, The first output level in each control cycle continues the level of the last output level of the previous cycle, and the signs of the tracking error values ​​generated by the two output levels should be opposite. The duration of the first output level is... The calculation is as follows: In the formula, ,and ; To obtain from a finite set of controls The optimal combination of voltage level and first output level is selected using the following first cost function. Conduct an assessment: In the formula, for The reference current, , All of these are weighting coefficients in the first cost function.

6. The adaptive optimization model predictive control method according to claim 2, characterized in that, When the control reference value changes beyond the threshold range, the MMMC system will enter a transient operating state, switch to a global optimization strategy, expand the limited control set to all levels, and execute a group optimization mechanism, as follows: Step 1: Sample and initialize the key system variables; Step two, connect each bridge arm N The sub-modules are divided into G Groups, each group contains the following number of submodules: M j ; Step 3: Treat each group as a whole and generate a control set for candidate levels. ; Step four, based on the control set in step three. Inter-group optimization is performed, and the predicted current value, current error value, and corresponding second cost function are calculated under the action of each control variable. ; Step 5: Record the optimal error value Optimal cost function value and optimal control quantity ; Step six, set the first judgment condition: if the optimal cost function value at this time is... Less than minimum error ε ,or If the corresponding current error values ​​have the same sign, then the optimization result is considered to be accurate enough, and the result is directly output as the final control command, and then proceed to step nine. Otherwise, proceed to step seven; Step 7: Set the second judgment condition: Based on the optimal control quantity and its control error, determine the interval where the global optimal level is located, and update the control set as follows. ; Step 8: Update the control set based on the results obtained in Step 7. Perform intra-group optimization and calculate The current prediction value and second cost function under the action of each control quantity are calculated, and the optimal control quantity corresponding to the second cost function is output. Step nine: After completing the calculation for one control cycle, return to the starting point and wait for the next sampling cycle to begin.

7. The adaptive optimization model predictive control method according to claim 6, characterized in that, In step three, under the capacitor voltage balancing control strategy, the average value of the submodule capacitor voltage of the bridge arm is set to... Therefore, each group is treated as a whole, generating a control set of candidate levels for inter-group optimization. as follows: In the formula, Round down to the nearest integer; .

8. The adaptive optimization model predictive control method according to claim 7, characterized in that, In step seven, if the current optimal control quantity is The first control variable, or its corresponding optimal error value Error value with the previous control quantity If the signs are the same, it indicates that the global optimal solution is located at the current optimal control quantity. With the next control quantity For submodule groups between these groups, the finite control set should be updated as follows: Otherwise, the global optimal solution will be located at the current optimal control quantity. Compared with the previous control quantity For the submodule groups between them, the finite control set should be updated as follows: 。 9. The adaptive optimization model predictive control method according to claim 7, characterized in that, The computational cost of global optimization is reduced to (2 G +1)~(2 G +max( M j To minimize computational cost, the number of groups... G and the number of submodules within each group M j The value of should follow the following: 1) Order and ; 2) Calculate separately and ; 3) Select The corresponding number of groups is used as the group number. G Number of submodules M j According to the number of groups G Calculated.

Images