A Model Predictive Control Method for MMC Based on Level Modulation Method

Abstract

The present invention discloses an MMC model predictive control method based on a level modulation method, and the steps are as follows: (1) Reduce the control objective of the cost function, and perform delay compensation by predicting the discrete model one step forward to optimize the cost function; (2) Calculate the number of sub-modules to be put into operation, and adjust the number of sub-modules put into operation by controlling the difference in the output voltage levels at the previous and current moments; (3) Design a capacitor voltage balancing control strategy, sort the sub-modules in the put-in and cut-out states separately, and on this basis, add the fluctuation threshold of the capacitor voltage. The present invention can reduce the current harmonic content while reducing the number of rollings, and reduce the switching frequency of the MMC; the present invention can reduce the number of times of switching the sub-module switch state while ensuring the balance of the sub-module capacitor voltage, reduce the excessive change of the sub-module, and enhance the robustness of the system.

Description

Technical Field

[0001] The present invention relates to model predictive control, and in particular to an MMC model predictive control method based on a level modulation method. Background Art

[0002] The main control objectives of MMC are AC-side current tracking, circulating current suppression, and balanced control of sub-module capacitor voltages. Currently, the traditional control strategy of MMC requires a PI linear controller to perform closed-loop design for a single objective. For multi-objective control, a cascade structure is used to achieve the control effect. However, this method has many control links, and the PI controller requires parameter tuning. Many control links will cause system oscillation when a certain link fails. The mutual tuning and cooperation between controllers are also very complex and cumbersome. Poor coordination in the parameter tuning process will affect the control effect of MMC.

[0003] Model Predictive Control (MPC) does not require parameter tuning. It is a non-linear control strategy obtained through the discretized model of MMC. The design process is simple, has good robustness, and has good dynamic response characteristics. It is suitable for dealing with multi-objective control problems, so it is suitable for the control requirements of MMC. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide an MMC model predictive control method based on a level modulation method, so as to reduce the current harmonic content while reducing the number of rollings, reduce the switching frequency of MMC, reduce the number of times of switching the sub-module state while ensuring the balance of the sub-module capacitor voltages, reduce the excessive change of the sub-module, and enhance the robustness of the system.

[0005] Technical Solution: An MMC model predictive control method based on a level modulation method according to the present invention includes the following steps:

[0006] (1) Reduce the control objectives of the cost function, and perform delay compensation by predicting the discrete model one step forward to optimize the cost function.

[0007] (2) Calculate the number of sub-modules to be put in, and adjust the number of sub-modules to be put in by controlling the difference in the output voltage levels at the previous and current moments.

[0008] (3) Design a balanced control strategy for capacitor voltages, sort the sub-modules in the on and off states separately, and add a fluctuation threshold of the capacitor voltage on this basis.

[0009] The specific content of step (1) is as follows:

[0010] Optimize the control objectives. The control objectives of the MMC include the AC-side current, the circulating current, and the sub-module capacitor voltage. Among them, there are 2N sub-module capacitor voltages that need to be predicted, with the most prediction times and being proportional to the number of sub-modules. Therefore, this method only performs model predictive control on the AC-side current and the circulating current, and performs independent balancing control on the sub-module capacitor voltage. So the MMC value function is expressed as:

[0011] J = λ 1 |i xref (k + 1) - i x (k + 1)| + λ 2 |i cirx (k + 1)| (3)

[0012] Among them, λ 1 and λ 2 are the weight factors of the AC-side current and the circulating current respectively. The value of the weight factor determines the dominance of the control objective, and different system control requirements can be met by adjusting the weight factors of different magnitudes; i cirx (k + 1) is the AC-side circulating current at the (k + 1)-th sampling moment; i x (k + 1) is the grid-side current at the (k + 1)-th sampling moment; i xref (k + 1) is the reference value of the grid-side current at the (k + 1)-th sampling moment.

[0013] The discretized models of the upper and lower bridge arm voltages u px and u nx are expressed as:

[0014]

[0015] In the formula, n px and n nx are the number of sub-modules in the upper bridge arm and the lower bridge arm respectively; u Cpxi (k + 1) and u Cnxi (k + 1) are the capacitor voltages in a sub-module in the upper bridge arm and the capacitor voltages in a sub-module in the lower bridge arm at the (k + 1)-th sampling moment respectively; N is the number of sub-module capacitors to be predicted in a single bridge arm.

[0016] The discretized models of the AC-side current and the circulating current are expressed as:

[0017]

[0018]

[0019] In the formula, T s is the sampling time; L is the AC-side inductor; U dc is the DC voltage output by the MMC.

[0020] According to the above formulas (3) and (4), the values of the AC - side current and the circulating current at the (k + 1) - th moment are related to the number of sub - modules inserted in the upper and lower bridge arms. After obtaining the optimal combination through the rolling optimization of MPC, it is input into the voltage sorting algorithm for sub - module voltage balance control; this strategy converts the rolling optimization of the switch state in the traditional FCS - MPC strategy into optimizing the number of inserted sub - modules. When N sub - modules are inserted and operating normally, a voltage of N + 1 levels will be output, indicating that there is a corresponding relationship between the number of inserted sub - modules and the output voltage levels during normal operation. The output voltage level can be adjusted by controlling the number of inserted sub - modules, and by using the capacitor voltage sorting strategy, the number of rolling optimization times can be reduced from to N + 1, greatly reducing the number of calculations and calculation time.

[0021] The specific content of step (2) is as follows:

[0022] Predict the formulas (3) and (4) one step forward for delay compensation. The AC - side current is as follows:

[0023]

[0024]

[0025] In the formula, L arm is the arm inductor; R is the AC - side resistance.

[0026] By replacing i x (k + 2) and i cirx (k + 2) with their respective reference values and Therefore, the optimal voltages of the upper arm and the lower arm are expressed as:

[0027]

[0028]

[0029] In the formula, and are the optimal voltages of the upper arm and the lower arm respectively.

[0030] The expressions of A' and B' are as follows:

[0031]

[0032]

[0033] So the number of sub - modules temporarily inserted in the upper and lower bridge arms is expressed as:

[0034]

[0035]

[0036] The number of sub - modules inserted into the bridge arm obtained from the above formula is not the final number and needs to be modified. Without affecting the controllability of the circulating current, the output current and voltage quality should be improved. To output a more stable voltage level, it is necessary to adjust the temporarily inserted sub - modules in the upper and lower bridge arms according to the difference between the previously output voltage level level old and the temporarily output voltage level level at the current sampling moment tem The difference between them is used to adjust the temporarily inserted sub - modules in the upper and lower bridge arms and The previously output voltage level is expressed as follows:

[0037]

[0038] In the formula, and are the final numbers of sub - modules inserted at the previous moment;

[0039]

[0040] Δlevel = level tem - level old (15)

[0041] To reduce the harmonic content of the output current, it is necessary to ensure a smooth transition between adjacent output levels and reduce the difference between the output level at the previous moment and the current output level. Therefore, three conditions regarding Δlevel are generated. If Δlevel > 1, the number of sub - modules inserted is not directly the temporarily obtained sub - module numbers inserted into the upper and lower arms through equations (11) and (12), but is obtained by adding 1 to or subtracting 1 from Therefore, two possible combinations of sub - modules inserted into the upper and lower arms are generated: and These two combinations will cause changes in the output level, thereby reducing the difference between the output levels at the previous moment and the current moment.

[0042] Among them and only one value needs to be changed, without changing both variables simultaneously, which can ensure the controllability of the circulating current and will not increase the switching loss. It is necessary to find the optimal number of inserted sub - modules when the cost function reaches the minimum value from the possible combinations M u(l)1 and M u(l)2 The cost function is expressed as follows:

[0043] J = λ 1 |i xref (k + 2)-ix (k + 2)| + λ 2 |i cirx (k + 2)| (16)

[0044] Reduce the output level conversion difference to eliminate redundant levels to achieve better output performance; select the optimal number of sub - modules in the upper and lower arms from two possible combinations, find the control variable corresponding to the (k + 2) - th moment, and then find the number of sub - modules put into operation when the value function reaches the minimum. For each sampling moment, the cost function only needs to be calculated twice, so the number of rolling optimizations is reduced and the computational burden is reduced; the NLM - OMPC control strategy mainly includes two parts: the model prediction part and the capacitor voltage sorting and equalizing control. First, calculate the electrical quantities required by the NLM - OMPC control strategy, then perform delay compensation to obtain the value at the (k + 2) - th moment, calculate the optimal number of sub - modules put into operation when the value function reaches the minimum through two possible combinations, and finally input the optimal number of sub - modules put into operation into the capacitor voltage sorting control strategy to complete the control of three control objectives: the AC - side current, the circulating current, and the sub - module capacitor voltage.

[0045] The specific content of step (3) is as follows:

[0046] Sort the sub - modules to be put into and removed separately. Then, when the capacitor voltage fluctuates, by taking the difference from the number of sub - modules switched in and out at the previous moment, obtain the number of newly switched - in and out sub - modules. When , input sub - modules with corresponding voltages from the removed sub - modules according to the direction of the arm current; when input sub - modules with corresponding voltages from the sub - modules put into operation according to the direction of the arm current; the equalizing control of the sub - module capacitor voltages does not mean that all capacitor voltages need to be exactly the same at all times. Slight fluctuations near the reference value of the sub - module capacitor voltages are allowed, but the upper and lower limits of the fluctuations need to be set, which can reduce the switching frequency; after reconstructing the capacitor voltages, the upper and lower limits can be expressed as:

[0047]

[0048] When the arm current ixm > 0, input sub - modules with small capacitor voltages to charge the sub - module capacitors. To reduce the switching frequency, it is necessary to reduce the upper limit value of the capacitor voltage of the input sub - modules and increase the lower limit value of the capacitor voltage of the removed sub - modules, that is, ΔU ck > 0, Ucx1 is the upper limit value, and Ucx2 is the lower limit value.

[0049] When the arm current \(i_{xm}<0\), the sub-module with a larger capacitor voltage among the inserted sub-modules discharges the capacitor of the sub-module. To reduce the switching frequency, it is necessary to increase the upper limit value of the capacitor voltage of the inserted sub-module and decrease the lower limit value of the capacitor voltage of the removed sub-module, that is, \(\Delta U\). ck \(<0\), \(U_{cx1}\) is the lower limit value, and \(U_{cx2}\) is the upper limit value.

[0050] In the first-round sorting, the sub-module with the smallest capacitor voltage value among the inserted sub-modules is inserted, and the remaining sub-modules are removed. Then, the initial optimal switching state is sent to the upper and lower arms; after that, when there is a capacitor voltage fluctuation, the number of newly inserted and removed sub-modules is controlled to perform voltage sorting; When the arm current \(i_{xm}>0\), the capacitor voltage fluctuation range is \([U_{ck2}, U_{ck1}]\), and the sub-module with the smallest voltage is selected from the removed sub-modules and inserted; When the arm current \(i_{xm}<0\), the capacitor voltage fluctuation range is \([U_{ck1}, U_{ck2}]\), and the sub-module with the largest voltage is selected from the removed sub-modules and inserted; When the arm current \(i_{xm}<0\), the capacitor voltage fluctuation range is \([U_{ck1}, U_{ck2}]\), and the sub-module with the smallest voltage is selected from the inserted sub-modules and removed; When the arm current \(i_{xm}>0\), the capacitor voltage fluctuation range is \([U_{ck2}, U_{ck1}]\), and the sub-module with the smallest voltage is selected from the inserted sub-modules and removed; By this method, the switching frequency is reduced, the voltage fluctuation of the sub-module exceeding the limit is avoided, the constraint on the deviation of the capacitor voltage fluctuation is strengthened, and finally the optimal switching state is allocated to each sub-module.

[0051] A computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned MMC model predictive control method based on the level modulation method.

[0052] A computer device, including a storage, a processor, and a computer program stored on the storage and executable on the processor. When the processor executes the computer program, it implements the above-mentioned MMC model predictive control method based on the level modulation method.

[0053] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0054] 1. Aiming at the problems of large computational complexity, time delay, and large output current harmonic content in the traditional FCS-MPC control strategy, the present invention proposes an optimized model predictive control strategy based on NLM. First, the control objective of the value function is reduced, and the discrete model is predicted one step forward for delay compensation to optimize the value function. Then, the idea of NLM is introduced to calculate the number of input sub-modules, and the input sub-modules are adjusted by controlling the difference in the output voltage levels at the previous and current moments to ensure the smooth transition of the levels, reducing the number of rollings and the current harmonic content at the same time;

[0055] 2. The present invention separately designs a capacitor voltage equalization control strategy, sorts the sub-modules in the input and cut-off states separately, reduces the excessive changes of the sub-modules, thereby reducing the switching frequency of the MMC. On this basis, a fluctuation threshold of the capacitor voltage is added to ensure the equalization of the capacitor voltages of the sub-modules while reducing the number of times the sub-modules switch the switch state. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is the flowchart of the steps of the present invention;

[0057] Figure 2 is the schematic diagram of the optimized model predictive control strategy process;

[0058] Figure 3 is the schematic diagram of the improved capacitor voltage sorting strategy. DETAILED DESCRIPTION OF THE INVENTION

[0059] The technical solution of the present invention will be further described below in conjunction with the drawings.

[0060] As Figure 1 shown, a model predictive control method for MMC based on the level modulation method is as follows:

[0061] (1) Reduce the control objective of the value function, and predict the discrete model one step forward for delay compensation to optimize the value function.

[0062] (2) Calculate the number of input sub-modules, and adjust the input sub-modules by controlling the difference in the output voltage levels at the previous and current moments to ensure the smooth transition of the levels.

[0063] (3) Design a capacitor voltage equalization control strategy, sort the sub-modules in the input and cut-off states separately, and add a fluctuation threshold of the capacitor voltage on this basis.

[0064] As Figure 2 shown is the flowchart of the optimized model control strategy. It is necessary to find the optimal number of input sub-modules when the cost function is minimized from the possible combinations M u(l)1 and M u(l)2 . The cost function is expressed as follows:

[0065] J = λ 1 |i xref (k + 2) - i x (k + 2)| + λ 2 |i cirx (k + 2)| (1)

[0066] Reduce the output level conversion difference to eliminate redundant levels to achieve better output performance. Select the optimal number of sub - modules in the upper and lower arms from two possible combinations, find the control variable corresponding to the (k + 2) - th moment, and then find the number of sub - modules put into operation when the value function reaches the minimum. For each sampling moment, only need to calculate the cost function twice, so the number of rolling optimizations is reduced and the calculation burden is reduced.

[0067] As Figure 3 shown in the improved capacitor voltage strategy flowchart, by sorting the sub - modules put into and removed separately, and then when the capacitor voltage fluctuates, by taking the difference from the number of sub - modules switched in the previous moment, the number of newly switched sub - modules is obtained. When it is time, put in sub - modules corresponding to the voltage according to the direction of the arm current from the removed sub - modules; when it is time, remove sub - modules corresponding to the voltage according to the direction of the arm current from the put - in sub - modules. The balanced control of the sub - module capacitor voltage does not mean that all capacitor voltages need to be exactly the same at all times. Slight fluctuations near the reference value of the sub - module capacitor voltage are allowed, but the upper and lower limits of the fluctuations need to be set, which can reduce the switching frequency. Then after reconstructing the capacitor voltage, the upper and lower limits can be expressed as:

[0068]

[0069] When the arm current ixm > 0, put in sub - modules with small capacitor voltages to charge the sub - module capacitors. To reduce the switching frequency, it is necessary to reduce the upper limit value of the capacitor voltage of the put - in sub - modules and increase the lower limit value of the capacitor voltage of the removed sub - modules, that is, ΔU ck > 0, Ucx1 is the upper limit value, and Ucx2 is the lower limit value.

[0070] When the arm current ixm < 0, put in sub - modules with large capacitor voltages to discharge the sub - module capacitors. To reduce the switching frequency, it is necessary to increase the upper limit value of the capacitor voltage of the put - in sub - modules and reduce the lower limit value of the capacitor voltage of the removed sub - modules, that is, ΔU ck < 0, Ucx1 is the lower limit value, and Ucx2 is the upper limit value.

[0071] In the first round of sorting, the sub-module with the smallest capacitor voltage value is put in, and the remaining sub-modules are cut off. Then, the initial optimal switching state is sent to the upper and lower bridge arms. After that, when there is a capacitor voltage fluctuation, the number of newly switched sub-modules is controlled. to perform voltage sorting. When the arm current ixm > 0, the capacitor voltage fluctuation range is [Uck2, Uck1]. Select sub-modules with the smallest voltage from the cut-off sub-modules and put them in; When the arm current ixm < 0, the capacitor voltage fluctuation range is [Uck1, Uck2]. Select sub-modules with the largest voltage from the cut-off sub-modules and put them in. When the arm current ixm < 0, the capacitor voltage fluctuation range is [Uck1, Uck2]. Select sub-modules with the smallest voltage from the put-in sub-modules and cut them off; When the arm current ixm > 0, the capacitor voltage fluctuation range is [Uck2, Uck1]. Select sub-modules with the smallest voltage from the put-in sub-modules and cut them off. By this method, the switching frequency is reduced, the voltage fluctuation of the sub-modules exceeding the limit is avoided, the constraint on the deviation of the capacitor voltage fluctuation is strengthened, and finally the optimal switching state is assigned to each sub-module.

Claims

1. A model predictive control method for MMC based on level modulation method, characterized in that, it includes the following steps: (1) Reduce the control objective of the cost function, and predict the discrete model one step forward for delay compensation to optimize the cost function; specifically: Optimize the control objective, and the MMC cost function is expressed as: J = λ 1 |i xref (k + 1) - i x (k + 1)| + λ 2 |i cirx (k + 1)| (1) Among them, λ 1 and λ 2 are the weight factors of the AC-side current and the circulating current respectively. The value of the weight factor determines the dominance of the control objective, and different system control requirements can be met by adjusting the weight factors of different magnitudes; i cirx (k + 1) is the AC-side circulating current at the (k + 1)-th sampling instant; i x (k + 1) is the grid-side current at the (k + 1)-th sampling instant; i xref (k + 1) is the reference value of the grid-side current at the (k + 1)-th sampling instant; Upper and lower bridge arm voltages u px and u nx The discretized model representation is as follows: where n px and n nx are the numbers of sub - modules in the upper bridge arm and the lower bridge arm respectively; u Cpxi (k + 1) and u Cnxi (k + 1) are the capacitor voltages in a sub - module in the upper bridge arm and the capacitor voltage in a sub - module in the lower bridge arm at the (k + 1)-th sampling moment respectively; N is the number of sub - module capacitors to be predicted in a single bridge arm; The discretized models of the AC-side current and circulating current are expressed as: where T s is the sampling time; L is the inductor on the AC side; U dc is the DC voltage output by the MMC; According to equations (3) and (4), the values of the AC-side current and the circulating current at the k+1 moment are related to the number of sub-modules inserted in the upper and lower bridge arms. After obtaining the optimal combination through the rolling optimization of MPC, it is input into the voltage sorting algorithm for sub-module voltage balancing control. This strategy converts the rolling optimization of the switching state in the traditional FCS-MPC strategy into optimizing the number of inserted sub-modules. When N sub-modules are inserted and operating normally, a voltage of N+1 levels will be output, indicating that there is a corresponding relationship between the number of inserted sub-modules and the output voltage levels during normal operation. The output voltage levels can be adjusted by controlling the number of inserted sub-modules. Moreover, by using the capacitor voltage sorting strategy, the number of rolling optimization times can be reduced from to N+1, significantly reducing the number of calculations and calculation time. (2) Calculate the number of sub-modules to be inserted, and adjust the number of inserted sub-modules by controlling the difference in the output voltage levels at the previous and current moments; (3) Design a capacitor voltage balancing control strategy, sort the sub-modules in the inserted and removed states separately, and on this basis, add the fluctuation threshold of the capacitor voltage.

2. A model predictive control method for MMC based on level modulation method according to claim 1, characterized in that, the specific content of step (2) is: (2.1) Predict the formulas (3) and (4) one step forward for delay compensation, and the AC-side current is as follows: Where L arm is the arm inductance; R is the AC-side resistance; By substituting i x (k + 2) and i cirx (k + 2) with their respective reference values and Therefore, the optimal voltages of the upper and lower arms are expressed as: Wherein, and are the optimal voltages of the upper arm and the lower arm respectively; where the expressions of A' and B' are as follows: Therefore, the number of temporarily inserted sub-modules in the upper and lower bridge arms is expressed as: The number of sub-modules inserted into the bridge arm obtained from the above formula is not the final number, and it needs to be modified to improve the output current and voltage quality without affecting the controllability of the circulating current; in order to output a more stable voltage level, it is necessary to adjust the temporarily inserted sub-modules in the upper and lower bridge arms according to the difference between the previously output voltage level level old and the temporarily output voltage level level at the current sampling moment tem The sub-modules temporarily inserted in the upper and lower bridge arms are adjusted according to the difference between and The previously output voltage level is represented as follows: Wherein, and are the final quantities of the input sub-modules at the previous moment; △level = level tem -level old (15) To reduce the harmonic content of the output current, it is necessary to ensure a smooth transition between adjacent output levels and reduce the difference between the output level at the previous moment and the current output level. Therefore, three conditions regarding △level are generated. If △level > 1, the number of sub-modules inserted is not directly the temporary number of sub-modules obtained by inserting through equations (11) and (12) in the upper arm and the lower arm, but is obtained by adding 1 to or subtracting 1 from . Therefore, two possible combinations of sub-modules inserted in the upper arm and the lower arm are generated: and These two combinations will cause the output level to change, thereby reducing the difference between the output levels at the previous moment and the current moment. Among them and Only one value needs to be changed, without changing both variables simultaneously, which can ensure the controllability of the circulating current and will not increase the switching loss; (2.2) From the possible combinations M u(l)1 and M u(l)2 Find the optimal number of input sub - modules when the cost function reaches its minimum. The cost function is expressed as follows: Reduce the output level conversion difference to eliminate redundant levels to achieve better output performance; select the optimal number of sub-modules in the upper arm and lower arm from two possible combinations, find the control variables corresponding to the k+2 moment, and then find the number of inserted sub-modules when the cost function reaches the minimum value. For each sampling moment, only need to calculate the cost function twice, so the rolling optimization times are reduced and the calculation burden is reduced; the NLM-OMPC control strategy mainly includes two parts: the model prediction part and the capacitor voltage sorting and equalizing control. First, calculate the electrical quantities required by the NLM-OMPC control strategy, then perform delay compensation to obtain the value at the k+2 moment, calculate the optimal number of inserted sub-modules when the cost function reaches the minimum value through two possible combinations, and finally input the optimal number of inserted sub-modules into the capacitor voltage sorting control strategy to complete the control of the three control objectives of the AC-side current, circulating current and sub-module capacitor voltage.

3. A model predictive control method for MMC based on level modulation method according to claim 1, characterized in that, the specific content of step (3) is: By sorting the inserted and removed sub-modules separately, and then when the capacitor voltage fluctuates, by taking the difference from the number of inserted and removed sub-modules at the previous moment, the number of newly inserted and removed sub-modules is obtained. When , insert sub-modules with corresponding voltages from the removed sub-modules according to the direction of the arm current; when , remove sub-modules with corresponding voltages from the inserted sub-modules according to the direction of the arm current; the equalization control of the sub-module capacitor voltage does not mean that all capacitor voltages need to be exactly the same at all times. Slight fluctuations near the reference value of the sub-module capacitor voltage are also allowed, but the upper and lower limits of the fluctuations need to be set, which can reduce the switching frequency; then after reconstructing the capacitor voltage, the upper and lower limits can be expressed as: When the arm current ixm > 0, the sub-module with a smaller capacitor voltage is inserted to charge the sub-module capacitor. To reduce the switching frequency, it is necessary to lower the upper limit value of the capacitor voltage of the inserted sub-module and increase the lower limit value of the capacitor voltage of the removed sub-module, that is, △U ck > 0, Ucx1 is the upper limit value, and Ucx2 is the lower limit value; When the arm current ixm < 0, the sub-module with a larger capacitor voltage is turned on to discharge the sub-module capacitor. To reduce the switching frequency, it is necessary to increase the upper limit value of the capacitor voltage of the turned-on sub-module and decrease the lower limit value of the capacitor voltage of the turned-off sub-module, that is, △U ck < 0, Ucx1 is the lower limit value, and Ucx2 is the upper limit value; In the first round of sorting, the sub-module with the smallest input capacitor voltage value is selected, and the remaining sub-modules are removed. Then, the initial optimal switching state is sent to the upper and lower bridge arms. After that, when there is a capacitor voltage fluctuation, the number of newly switched sub-modules is controlled. to perform voltage sorting; When the arm current ixm > 0, the capacitor voltage fluctuation range is [Uck2, Uck1]. Select sub-modules with the smallest voltage from the removed sub-modules and put them in; When the arm current ixm < 0, the capacitor voltage fluctuation range is [Uck1, Uck2]. Select sub-modules with the largest voltage from the removed sub-modules and put them in; When the arm current ixm < 0, the capacitor voltage fluctuation range is [Uck1, Uck2]. Select sub-modules with the smallest voltage from the inserted sub-modules and remove them; When the arm current ixm > 0, the capacitor voltage fluctuation range is [Uck2, Uck1]. Select sub-modules with the smallest voltage from the inserted sub-modules and remove them. By this method, the switching frequency is reduced, the voltage fluctuation of the sub-module exceeding the limit is avoided, the constraint on the deviation of the capacitor voltage fluctuation is strengthened, and finally the optimal switching state is allocated to each sub-module.

4. A computer storage medium, on which a computer program is stored, characterized in that, when the computer program is executed by a processor, it implements a model predictive control method for MMC based on level modulation method as described in any one of claims 1-3.

5. A computer device, including a storage, a processor, and a computer program stored on the storage and executable on the processor, characterized in that, when the processor executes the computer program, it implements a model predictive control method for MMC based on level modulation method as described in any one of claims 1-3.

Images