A multi-dimensional self-adaptive dynamic adjustment inverter modulation method and storage medium

Abstract

The application discloses a kind of multi-dimension adaptive dynamic adjustment's inverter modulation method and storage medium, the method includes: offline stage: to complete the setting of initial optimal modulation strategy;That is, under the premise of not being subject to real-time calculation constraint, optimal initial modulation strategy and weight are obtained by large-scale sampling and high-dimensional search, to provide reliable initial point for online adaptive modulation;Online stage: to carry out adaptive weight update and real-time modulation;That is, modulation strategy is adjusted according to real-time operating condition, and multi-objective joint optimization is realized.The storage medium is realized based on the above method.The application has the advantages of simple principle, wide application range, can effectively improve the efficiency and stability of ANPC inverter and the like.

Description

Technical Field

[0001] This invention mainly relates to the field of control technology for new energy power electronic equipment, specifically a multi-dimensional adaptive dynamic adjustment inverter modulation method and storage medium. Background Technology

[0002] With global renewable energy installations exceeding 4,000 GW, inverters are gradually evolving towards ultra-high power density, high efficiency, high reliability, active support and grid-friendliness, and integrated photovoltaic and energy storage technologies. Among these, three-level and higher-order topologies have become the mainstream solutions. The ANPC topology, with its lower device voltage stress, multi-path current distribution capability, midpoint voltage adjustment freedom, and lower loss potential, is widely used in 1500V DC-side photovoltaic inverters and energy storage converters.

[0003] However, in practical applications, the aforementioned traditional technologies are prone to problems such as large fluctuations in midpoint current, rapid changes in load current direction, and uneven conduction losses in devices, leading to increased capacitor voltage ripple, reduced efficiency, and accelerated device aging. In other words, when the load changes dynamically, it is difficult to balance the switching action with the fluctuations in the DC-side capacitor voltage of the ANPC, as well as operating efficiency and power quality. Summary of the Invention

[0004] The technical problem to be solved by this invention is: in view of the technical problems existing in the prior art, this invention provides a multi-dimensional adaptive dynamic adjustment inverter modulation method and storage medium that is simple in principle, has a wide range of applications, and can effectively improve the efficiency and stability of ANPC inverters.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A multi-dimensional adaptive dynamic adjustment inverter modulation method, comprising: Offline phase: Used to set the initial optimal modulation strategy; that is, without being constrained by real-time calculation, the optimal initial modulation strategy and weights are obtained through large-scale sampling and high-dimensional search, providing a reliable initial point for online adaptive modulation; Online phase: used for adaptive weight updates and real-time modulation; that is, adjusting the modulation strategy according to the real-time operating conditions to achieve multi-objective joint optimization. As a further improvement of the present invention: in the offline phase, four normalized performance metrics are constructed, including: (1) DC capacitor voltage balance:

[0006] in For the first i capacitor voltage M This represents the total number of capacitors. This is the average value. Rated voltage; (2) Switching losses:

[0007] in and They represent the first k The conduction and turn-off losses of individual devices N Indicates the total number of switches; (3) Power quality:

[0008] THD V ac This indicates the total harmonic distortion (THD) of the AC side voltage. (4) Dynamic response:

[0009] i load ( t )and i ref ( t ) represent the load current and the inverter reference current, respectively.

[0010] As a further improvement of the present invention: the offline stage includes a multi-objective optimization process; in the multi-objective optimization process, the NSGA-II algorithm is used for high-dimensional search; the initial population is randomly generated, and then the performance vector of each individual is calculated. The non-dominated sorting and crowding comparison are performed to ensure the diversity of Pareto solutions; each solution represents an implementable ANPC modulation strategy, laying the foundation for selecting engineering-feasible compromise solutions in the offline stage.

[0011] As a further improvement of the present invention: in the offline stage, after obtaining the Pareto solution set, an index matrix is ​​constructed:

[0012] The objective weights of each indicator are calculated using the entropy weight method, and its standardized indicators are... p ij :

[0013] Where |P| represents the total number of individuals, and the entropy value is calculated. e j :

[0014] in, c =1 / ln|P|, thus obtaining the information entropy difference.d j :

[0015] Finally, the objective weight is obtained. : .

[0016] As a further improvement of the present invention: in the offline phase, a slight subjective weight is given by AHP. Its value is determined using a typical 9-scale method and is given by the field engineer, and is finally fused together with objective weights:

[0017] in, As a weighting factor; For each Pareto solution, a comprehensive weighted average is performed:

[0018] choose The minimum solution is used as the initial modulation strategy, and a modulation table and initial weight vector are generated. ω i ( 0 This provides initialization parameters for the online phase.

[0019] As a further improvement of the present invention: in the online phase, a two-layer adaptive structure is constructed, including: The first layer is a lightweight exponentially weighted adaptive mechanism; the weights are updated in each control cycle.

[0020] in This is the sensitivity coefficient. The allowable error band or ideal value for the indicator. t and t +1 represents the current time and the next time; if the indicator exceeds the threshold, the weight increases. The second layer is for MPC modulation action selection; the candidate action set includes equivalent vector combinations, different bridge arm clamping paths, and allowed commutation sequences.

[0021] As a further improvement of the present invention: in the MPC state space, the system state vector is defined as:

[0022] in and These are the upper and lower capacitor voltages of the ANPC, respectively. It is a three-phase AC output current. k Indicates the firstk One sampling period; The control input vector is defined as a combination of switching actions:

[0023] in Indicates the corresponding bridge arm in the 1st... k Switching state for each sampling period: 1 indicates forward conduction, 0 indicates midpoint vector conduction, -1 indicates reverse conduction; The dynamic response of the ANPC capacitor is obtained by integrating the current:

[0024] in i c,i ( t ) is the first i The current in the capacitor C i For the first i The capacitance value of each capacitor is used to perform forward Euler discretization on the continuous-time system, resulting in the discrete-time state update equation:

[0025] in T s To control the sampling period, V c,i ()and i c,i () indicates the first i The voltage and current of each capacitor; the capacitor current is calculated using switching action and an AC current load model:

[0026]

[0027] coefficient Determined by the bridge arm clamping strategy and switch combination, and pre-calculated through table lookup. i ac,ph Indicates alternating phase current; Alternating current is approximated by a resistive-inductive load:

[0028] in R , L Representing resistance and inductance, V ph and V load,ph Representing the AC phase voltage and the load phase voltage, discretization yields:

[0029] in Action of switch u ( k This is obtained by mapping the DC-side capacitor voltage; By combining the dynamics of capacitor voltage and AC current, a linear or quasi-linear state-space model can be constructed: x ( k +1)= Ax ( k )+ Bu ( k ) The state matrix A With input matrix B These represent the influence of the system's inherent dynamics and control inputs on the state, respectively, and are obtained through small-signal linearization or table lookup methods. The output vector is defined as:

[0030] Where iac, a / b / c represent the AC currents of phases a, b, and c, respectively, and MPC is directly referenced in the cost function. y ( k This is used to calculate capacitor voltage deviation, dynamic response, and power quality indicators.

[0031] As a further improvement of the present invention: a candidate action prediction model is included in the online phase. In MPC control, the short-term impact of candidate switching actions on the system needs to be predicted in each sampling period, including capacitor voltage changes, switching losses and current paths.

[0032] As a further improvement of the present invention: in the online phase, the candidate action set u It includes three types of vector combinations: Small vector V S It has minimal impact on capacitors and responds quickly to load disturbances. Large vector V l High amplitude output, rapid adjustment of output voltage; Midpoint vector V m Adjusting the DC capacitor voltage balance.

[0033] The present invention also provides a storage medium that can be read by a computer or processor, wherein the storage medium stores a computer program for executing any of the above methods.

[0034] Compared with the prior art, the advantages of the present invention are as follows: 1. The multi-dimensional adaptive dynamic adjustment inverter modulation method and storage medium of this invention are multi-objective adaptive dynamic modulation methods for ANPC three-level inverters. This method is simple in principle, widely applicable, and can effectively improve the efficiency and stability of ANPC inverters. It achieves capacitor voltage balance, reduced switching losses, improved power quality, and enhanced dynamic response through offline optimization and online adaptive control. In the offline stage, the Pareto optimal set of modulation and control parameters is searched using NSGA-II, and the entropy weight method and Analytic Hierarchy Process (AHP) are combined to determine the comprehensive weights and generate the initial modulation strategy. In the online stage, an exponentially weighted adaptive mechanism is used to adjust the index weights in real time, and Model Predictive Control (MPC) is used to predict the capacitor current, current dynamics, and losses of candidate vectors, selecting the optimal switching action for the current cycle. This method requires no hardware modification and can effectively improve the efficiency, stability, and adaptability of ANPC inverters.

[0035] 2. The multi-dimensional adaptive dynamic adjustment inverter modulation method and storage medium of the present invention utilize offline and online hybrid optimization and predicts the capacitor current, current dynamics and losses of candidate vectors based on MPC, selects the optimal switching action for the current cycle, and takes into account capacitor voltage, operating efficiency, power quality and dynamic response capability, thereby improving the efficiency, stability and adaptability of ANPC inverters. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the method of the present invention.

[0037] Figure 2 This is a schematic diagram of a specific process in a specific embodiment of the present invention. Detailed Implementation

[0038] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] In the description of this application, the following technical terms need to be understood: ANPC: Active Neutral Point Clamped; MPC: Model Predictive Control; AHP: Analytic Hierarchy Process.

[0040] like Figure 1 and Figure 2As shown, this invention discloses a multi-dimensional adaptive dynamic adjustment inverter modulation method for an ANPC three-level inverter, comprising: Offline phase: Used to set the initial optimal modulation strategy; that is, without being constrained by real-time calculation, the optimal initial modulation strategy and weights are obtained through large-scale sampling and high-dimensional search, providing a reliable initial point for online adaptive modulation; Online phase: used for adaptive weight updates and real-time modulation; that is, adjusting the modulation strategy according to the real-time operating conditions to achieve multi-objective joint optimization.

[0041] In a specific application example, during the offline phase of this invention, in order to comprehensively evaluate the performance of the modulation strategy, this invention first constructs four normalized performance indicators, which may include, depending on actual needs: (1) DC capacitor voltage balance:

[0042] in For the first i capacitor voltage M This represents the total number of capacitors. This is the average value. This is the rated voltage.

[0043] (2) Switching losses:

[0044] in and They represent the first k The conduction and turn-off losses of individual devices N Indicates the total number of switches.

[0045] (3) Power quality:

[0046] THD V ac ) indicates total harmonic distortion of the AC side voltage.

[0047] (4) Dynamic response:

[0048] i load ( t )and i ref ( t ) represent the load current and the inverter reference current, respectively.

[0049] In a specific application example, the offline phase of this invention includes a multi-objective optimization process; furthermore, in this example, the NSGA-II algorithm is used for high-dimensional search during the multi-objective optimization process. That is, the initial population is randomly generated, and then the performance vector of each individual is calculated. The algorithm performs non-dominated sorting and crowding comparison to ensure the diversity of the Pareto solution set (Pareto front solution set). Each solution represents an implementable ANPC modulation strategy, laying the foundation for selecting engineering-feasible compromise solutions in the offline phase. The NSGA-II algorithm generates an initial population, which consists of multiple individuals, referring to the individuals within the population.

[0050] In a specific application example, during the offline phase of this invention, after obtaining the Pareto solution set, an index matrix is ​​constructed:

[0051] The objective weights of each indicator are calculated using the entropy weight method, and its standardized indicators are... p ij :

[0052] Where |P| represents the total number of individuals, and the entropy value is calculated. e j :

[0053] in, c =1 / ln|P|, thus obtaining the information entropy difference. d j :

[0054] Finally, the objective weight is obtained. :

[0055] As a preferred embodiment, in order to incorporate a weak subjective bias based on engineering experience, this example further uses AHP to provide a slight subjective weight. Its value is determined using a typical 9-scale method and is given by the field engineer, and is finally fused together with objective weights:

[0056] in, This is the weighting factor, which can be 0.3.

[0057] For each Pareto solution, a comprehensive weighted average is performed:

[0058] choose The minimum solution is used as the initial modulation strategy, and a modulation table and initial weight vector are generated. ω i ( 0 This provides initialization parameters for the online phase.

[0059] In a specific application example, during the online phase of this invention, to balance real-time performance and stability, this invention further designs a two-layer adaptive structure, which includes: The first layer is a lightweight exponentially weighted adaptive mechanism.

[0060] Weights are updated every control cycle:

[0061] in This is the sensitivity coefficient. The allowable error band or ideal value for the indicator. t and t +1 represents the current time and the next time step. If the indicator exceeds the threshold, the weight increases; for example, when the capacitor voltage deviation increases, the system prioritizes the midpoint vector path to quickly restore balance. This method has low computational complexity and is suitable for real-time operation on DSPs or FPGAs.

[0062] The second layer is for selecting the MPC modulation action.

[0063] The candidate action set includes equivalent vector combinations, different arm clamping paths, and allowed commutation sequences. Therefore, this invention performs discrete-time state-space modeling of the ANPC topology inverter, enabling the prediction and optimization of key states such as DC-side capacitor voltage, current, and output power.

[0064] Specifically, this example includes the MPC state space, where: The system state vector is defined as:

[0065] in and These are the upper and lower capacitor voltages of the ANPC, respectively. It is a three-phase AC output current. k Indicates the first k One sampling period.

[0066] The control input vector is defined as a combination of switching actions:

[0067] in Indicates the corresponding bridge arm in the 1st... k The switching state for each sampling period: 1 indicates forward conduction, 0 indicates midpoint vector conduction, and -1 indicates reverse conduction.

[0068] The dynamic response of the ANPC capacitor can be obtained by integrating the current:

[0069] in i c,i ( t ) is the first i The current in the capacitor C i For the first i The capacitance value of each capacitor is used to perform forward Euler discretization on the continuous-time system, resulting in the discrete-time state update equation:

[0070] in T s To control the sampling period, V c,i ()and i c,i () indicates the first i The voltage and current of each capacitor. The capacitor current can be calculated using switching action and an AC current load model:

[0071]

[0072] coefficient Determined by the bridge arm clamping strategy and switch combination, it can be pre-calculated by looking up a table. i ac,ph It represents the alternating phase current.

[0073] Alternating current can be approximated by an inductive load:

[0074] in R , L Representing resistance and inductance, V ph and V load,ph Representing the AC phase voltage and the load phase voltage, discretization yields:

[0075] in Can be activated by a switch u ( k This is obtained by mapping the DC-side capacitor voltage.

[0076] By combining the dynamics of capacitor voltage and AC current, a linear or quasi-linear state-space model can be constructed: x ( k+1)= Ax ( k )+ Bu ( k ) The state matrix A With input matrix B These represent the influence of the system's inherent dynamics and control inputs on the state, respectively, and can be obtained through small-signal linearization or table lookup.

[0077] The output vector is defined as:

[0078] Where iac, a / b / c represent the AC currents of phases a, b, and c, respectively, and MPC can be directly referenced in the cost function. y ( k This is used to calculate capacitor voltage deviation, dynamic response, and power quality indicators.

[0079] Specifically, this example includes a candidate action prediction model. In MPC control, each sampling period needs to predict the short-term impact of candidate switching actions on the system, including capacitor voltage changes, switching losses, and current paths.

[0080] Furthermore, as a preferred embodiment, in this example, the candidate action set u It includes three types of vector combinations: 1. Small vector V S It has minimal impact on capacitors and responds quickly to load disturbances. 2. Large vectors V l High amplitude output, rapid adjustment of output voltage; 3. Midpoint vector V m Adjusting the DC capacitor voltage balance.

[0081] Each action vector can be represented as: u k =[ u 1, u 2, u 3] Corresponds to the three-bridge arm state.

[0082] For the i A capacitor, action u k The voltage prediction is as follows:

[0083] Capacitor current prediction It can be represented by a linear combination:

[0084] in It is based on action vectors u k Calculate the midpoint split coefficient.

[0085] Switching losses can be predicted based on semi-empirical formulas:

[0086] in E on,i and E off,i Indicates the first i Turn-on and turn-off losses of individual power devices.

[0087] The predicted AC current value for each action is:

[0088] Output voltage prediction can be further used for THD assessment and power quality constraints:

[0089] In the prediction window N p Within this framework, the cost function for MPC (Multi-Objective Computation) is:

[0090] in ω 1, ω 2, ω 3 and ω 4 represents the weighting coefficient. l Let the step size be 1 or 2. Ultimately, the optimization problem can be expressed as:

[0091] Output optimal action Used for the current control cycle, the window scrolls forward to form a closed-loop real-time modulation. The overall process is as follows: Figure 2 As shown.

[0092] As shown above, this invention achieves capacitor voltage balance, reduced switching losses, improved power quality, and enhanced dynamic response through offline optimization and online adaptive control. In the offline stage, the NSGA-II algorithm searches for the Pareto optimal set of modulation and control parameters, and combines entropy weighting and AHP to determine the comprehensive weights and generate the initial modulation strategy. In the online stage, an exponentially weighted adaptive mechanism is used to adjust the index weights in real time, and based on MPC, the capacitor current, current dynamics, and losses of candidate vectors are predicted to select the optimal switching action for the current cycle. Using this method, the efficiency, stability, and adaptability of ANPC inverters can be effectively improved without hardware modifications. This invention is particularly suitable for medium- and high-power photovoltaic inverters, energy storage converters, flexible DC converters, grid-connected inverters, and microgrid converters.

[0093] The present invention also provides a storage medium that can be read by a computer or processor, wherein the storage medium stores a computer program for executing any of the above methods.

[0094] Those skilled in the art will understand that the above embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0095] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A multi-dimensional adaptive dynamic adjustment inverter modulation method for an ANPC three-level inverter, characterized in that, include: Offline phase: Used to set the initial optimal modulation strategy; that is, without being constrained by real-time calculation, the optimal initial modulation strategy and weights are obtained through large-scale sampling and high-dimensional search, providing a reliable initial point for online adaptive modulation; Online phase: Used for adaptive weight updates and real-time modulation; That is, to adjust the modulation strategy according to the real-time operating conditions and achieve multi-objective joint optimization; During the offline phase, four normalized performance metrics are constructed, including: (1) DC capacitor voltage balance: in For the first i capacitor voltage M This represents the total number of capacitors. This is the average value. Rated voltage; (2) Switching losses: in and They represent the first k The conduction and turn-off losses of individual devices N Indicates the total number of switches; (3) Power quality: THD V ac This indicates the total harmonic distortion (THD) of the AC side voltage. (4) Dynamic response: i load ( t )and i ref ( t These represent the load current and the inverter reference current, respectively. The offline phase includes a multi-objective optimization process; during the multi-objective optimization process, the NSGA-II algorithm is used for high-dimensional search; the initial population is randomly generated, and then the performance vector of each individual is calculated. The non-dominated sorting and crowding comparison are performed to ensure the diversity of Pareto solutions; each solution represents an implementable ANPC modulation strategy, laying the foundation for selecting engineering-feasible compromise solutions in the offline stage. In the online phase, a two-layer adaptive structure is constructed, including: The first layer is a lightweight exponentially weighted adaptive mechanism; the weights are updated in each control cycle. in This is the sensitivity coefficient. The allowable error band or ideal value for the indicator. t and t +1 represents the current time and the next time; if the indicator exceeds the threshold, the weight increases. The second layer is for MPC modulation action selection; the candidate action set includes equivalent vector combinations, different bridge arm clamping paths, and allowed commutation sequences.

2. The inverter modulation method with multi-dimensional adaptive dynamic adjustment according to claim 1, characterized in that, In the offline phase, after obtaining the Pareto solution set, an index matrix is ​​constructed: The objective weights of each indicator are calculated using the entropy weight method, and its standardized indicators are... p ij : Where |P| represents the total number of individuals, and the entropy value is calculated. e j : in, c =1 / ln|P|, thus obtaining the information entropy difference. d j : Finally, the objective weight is obtained. : 。 3. The inverter modulation method with multi-dimensional adaptive dynamic adjustment according to claim 2, characterized in that, In the offline phase, a slight subjective weight is given by AHP. Its value is determined using a typical 9-scale method and is given by the field engineer, and is finally fused together with objective weights: in, As a weighting factor; For each Pareto solution, a comprehensive weighted average is performed: choose The minimum solution is used as the initial modulation strategy, and a modulation table and initial weight vector are generated. ω i ( 0 This provides initialization parameters for the online phase.

4. The inverter modulation method with multi-dimensional adaptive dynamic adjustment according to claim 1, characterized in that, In the MPC state space, the system state vector is defined as: in and These are the upper and lower capacitor voltages of the ANPC, respectively. It is a three-phase AC output current. k Indicates the first k One sampling period; The control input vector is defined as a combination of switching actions: in Indicates the corresponding bridge arm in the 1st... k Switching state for each sampling period: 1 indicates forward conduction, 0 indicates midpoint vector conduction, -1 indicates reverse conduction; The dynamic response of the ANPC capacitor is obtained by integrating the current: in i c,i ( t ) is the first i The current in the capacitor C i For the first i The capacitance value of each capacitor is used to perform forward Euler discretization on the continuous-time system, resulting in the discrete-time state update equation: in T s To control the sampling period, V c,i ()and i c,i () indicates the first i The voltage and current of each capacitor; the capacitor current is calculated using switching action and an AC current load model: coefficient Determined by the bridge arm clamping strategy and switch combination, and pre-calculated through table lookup. i ac,ph Indicates alternating phase current; Alternating current is approximated by a resistive-inductive load: in R , L Representing resistance and inductance, V ph and V load,ph Representing the AC phase voltage and the load phase voltage, discretization yields: in Action of switch u ( k This is obtained by mapping the DC-side capacitor voltage; By combining the dynamics of capacitor voltage and AC current, a linear or quasi-linear state-space model can be constructed: x ( k +1)= Ax ( k )+ Bu ( k ) The state matrix A With input matrix B These represent the influence of the system's inherent dynamics and control inputs on the state, respectively, and are obtained through small-signal linearization or table lookup methods. The output vector is defined as: Where iac, a / b / c represent the AC currents of phases a, b, and c, respectively, and MPC is directly referenced in the cost function. y ( k This is used to calculate capacitor voltage deviation, dynamic response, and power quality indicators.

5. The inverter modulation method with multi-dimensional adaptive dynamic adjustment according to claim 1, characterized in that, The online phase includes a candidate action prediction model. In MPC control, each sampling period needs to predict the short-term impact of candidate switching actions on the system, including capacitor voltage changes, switching losses, and current paths.

6. The inverter modulation method with multi-dimensional adaptive dynamic adjustment according to claim 1, characterized in that, Candidate action set in the online phase u It includes three types of vector combinations: Small vector V S It has minimal impact on capacitors and responds quickly to load disturbances. Large vector V l High amplitude output, rapid adjustment of output voltage; Midpoint vector V m Adjusting the DC capacitor voltage balance.

7. A storage medium capable of being read by a computer or processor, characterized in that, The storage medium stores a computer program for performing the method described in any one of claims 1 to 6.

Images