Control Method and Control Device for a Neutral-Point-Clamped Inverter

Abstract

The present application discloses a control method and a control device for a neutral point clamped inverter. The control method includes: constructing a current prediction model to predict the output current based on the input voltage vector; constructing a first cost function, minimizing the first cost function, selecting a voltage vector and determining the type of the selected voltage vector; constructing a second cost function, which is configured to: based on the type of the selected voltage vector, select and output the switching state that is beneficial to promoting the neutral point potential balance; in response to the type of the voltage vector being a large vector or a medium vector, if it is determined that the current neutral point potential deviation exceeds a preset neutral point potential offset threshold, and the currently selected voltage vector is not conducive to capacitor voltage balance, then re-select the voltage vector based on the first cost function, otherwise, output the switching state corresponding to the selected voltage vector. The present application can realize the prediction of the inverter output current and the balance of the inverter neutral point voltage.

Description

Technical Field

[0001] The present application relates to the technical field of inverters, and in particular, to a control method and a control device for a neutral-point clamped inverter. Background Art

[0002] As the topology with the longest development time and the widest application, the neutral-point clamped three-level inverter is favored by many researchers due to its simple structure, low control difficulty, low loss, and relatively balanced loss distribution of each power module. In terms of control strategies, the finite set model predictive control is widely used in the control of power converters because of its fast dynamic response, simple implementation method, and multi-variable processing ability. However, the existing control methods need to predict through inductance parameters such as weight factors and capacitance values, resulting in a strong dependence of the existing control methods on inductance parameters such as weight factors and capacitance values. Summary of the Invention

[0003] In order to solve the deficiencies of the prior art, the present application adopts the following technical solutions:

[0004] In a first aspect, the present application provides a control method for a neutral-point clamped inverter. The control method includes the following steps:

[0005] Construct a current prediction model, which predicts the output current based on the input voltage vector to obtain a current prediction value;

[0006] Construct a first cost function, which is composed of a current reference value and a current prediction value;

[0007] With the goal of minimizing the first cost function, select a voltage vector and determine the type of the selected voltage vector;

[0008] Construct a second cost function, which is configured to: based on the type of the selected voltage vector, select and output a switching state that is conducive to promoting the neutral-point potential balance;

[0009] Wherein, in response to the type of the selected voltage vector being a large vector or a medium vector, if it is determined that the current neutral-point potential deviation exceeds a preset neutral-point potential offset threshold, and the currently selected voltage vector is not conducive to capacitor voltage balance, then reselect the voltage vector based on the first cost function, otherwise, output the switching state corresponding to the selected voltage vector.

[0010] In summary, a control method for a neutral point clamped inverter provided by the present application predicts the output current of the inverter by constructing a current prediction model. Without the capacitance information of the capacitor and the weight factor, the optimal voltage vector is selected by minimizing the cost function to achieve the control of the neutral point clamped inverter. Furthermore, the switching state of the optimal voltage vector is further selected, thereby promoting the neutral point voltage balance of the neutral point clamped inverter without increasing the complexity of the control method.

[0011] Furthermore, the method further includes:

[0012] Represent the continuous-time model of the neutral point voltage of the neutral point clamped inverter as a discrete prediction model of the neutral point voltage, and the discrete prediction model of the neutral point voltage is represented by the following formula:

[0013] ;

[0014] In the formula, v n ( k+ (1) represents the neutral point voltage at the next moment, v n ( k ) represents the neutral point voltage at this moment, T s represents the sampling frequency or control frequency of the controller, C dc represents the bus capacitor on the DC side, i n represents the current flowing through the neutral point;

[0015] Design the second cost function based on the discrete prediction model of the neutral point voltage, evaluate the switching state of the selected voltage vector using the second cost function, and select the switching state that can promote the neutral point voltage v n ( k+ (1)) tending to 0 to achieve neutral point voltage balance.

[0016] Furthermore, the second cost function includes a cost function for evaluating large vectors and medium vectors, and the cost function for evaluating large vectors and medium vectors is represented by the following formula:

[0017] ;

[0018] In the formula, sign(*) represents the sign function, v n ( k ) represents the neutral point voltage under the influence of the selected voltage vector, i n ( k) represents the current flowing through the midpoint under the influence of the selected voltage vector, b v represents the preset midpoint potential offset threshold;

[0019] wherein, in response to the cost function J b = 2, then reselect the voltage vector based on the first cost function, otherwise, output the switching state corresponding to the selected voltage vector.

[0020] Further, the second cost function includes a cost function for evaluating small vectors, and the cost function for evaluating small vectors is represented by the following formula:

[0021] ;

[0022] In the formula, sign(*) represents the sign function, v n ( k ) represents the midpoint voltage under the influence of the selected voltage vector, i n ( k ) represents the current flowing through the midpoint under the influence of the selected voltage vector;

[0023] wherein, in response to the type of the selected voltage vector being a small vector, based on the cost function J v , among the two switching states corresponding to the small vector, select and output the switching state that makes the cost function J v the smallest.

[0024] Further, the second cost function includes a cost function for evaluating zero vectors, and the cost function for evaluating zero vectors is represented by the following formula:

[0025] ;

[0026] In the formula, S abc ( k ) represents the switching state at time k, represents the two-norm, The role of is to minimize the change in the switching state between two control periods;

[0027] wherein, in response to the type of the selected voltage vector being a zero vector, based on the cost function J s , among the three switching states corresponding to the zero vector, select and output the switching state that makes the cost function J s the smallest.

[0028] Further, the first cost function is expressed by the following formula:

[0029] ;

[0030] wherein, represents the current reference value, represents the predicted current value.

[0031] Further, the predicted current value is obtained by the following formula:

[0032] ;

[0033] wherein, R o represents the load resistance value of the inverter, L o represents the load inductance value of the inverter, T s represents the sampling / control frequency of the controller, represents the inverter output current in the stationary orthogonal coordinate system, represents the output voltage of the inverter in the stationary orthogonal coordinate system, where the inverter output current is determined by the selected voltage vector aiming at minimizing the first cost function.

[0034] Further, the current reference value is obtained by the following formula:

[0035] ;

[0036] wherein, i * ( k ) represents k the current reference value at time

[0037] Further, any one of the following methods is adopted for current prediction: a current prediction structure based on an extended Kalman filter, or, a data-driven neural network predictor, or, a current prediction structure based on adaptive predictive control.

[0038] In a second aspect, the present application further provides a control device for a neutral point clamped inverter, and the control device controls the neutral point clamped inverter by applying the above control method. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flowchart of the steps of a control method for a neutral point clamped inverter provided by an embodiment of the present application;

[0040] Figure 2Schematic diagram of the circuit topology of a neutral-point clamped inverter to which a control method according to an embodiment of the present application is applied;

[0041] Figure 3 Schematic diagram of the space voltage vector generated by the control method of the neutral-point clamped inverter provided by an embodiment of the present application;

[0042] Figure 4 Flowchart of the algorithm for achieving neutral-point voltage balance in the control method of the neutral-point clamped inverter provided by an embodiment of the present application;

[0043] Figure 5 Effect display diagram of the capacitor voltage balance of the control method of the neutral-point clamped inverter provided by an embodiment of the present application. Detailed implementation manners

[0044] The following will describe the present application in detail in conjunction with the specific implementation manners shown in the accompanying drawings. However, these implementation manners do not limit the present application, and any structural, method, or functional transformation made by those of ordinary skill in the art based on these implementation manners is included in the protection scope of the present application.

[0045] To solve the deficiencies of the prior art, in a first aspect, the present application provides a control method for a neutral-point clamped inverter, as Figure 1 shown, the control method includes the following steps:

[0046] Step S11, constructing a current prediction model, which predicts the output current based on the input voltage vector to obtain a current prediction value;

[0047] Step S12, constructing a first cost function, which is composed of a current reference value and a current prediction value;

[0048] Step S13, with the goal of minimizing the first cost function, selecting a voltage vector and determining the type of the selected voltage vector;

[0049] Step S14, constructing a second cost function, which is configured to: based on the type of the voltage vector, select and output a switching state that is conducive to promoting the neutral-point potential balance;

[0050] Among them, in response to the type of the selected voltage vector being a large vector or a medium vector, if it is determined that the current neutral-point potential deviation exceeds a preset neutral-point potential offset threshold, and the currently selected voltage vector is not conducive to capacitor voltage balance, then reselect the voltage vector based on the first cost function, otherwise, output the switching state corresponding to the selected voltage vector.

[0051] The circuit topology of the three-level neutral-point clamped inverter is as Figure 2As shown, a resistor-inductor load is configured in the inverter, where R in the figure o represents the resistance value of the load, and L o represents the inductance value of the load. C1 represents the upper capacitor on the DC side, and C2 represents the lower capacitor on the DC side. The inverter includes three bridge arms, and each bridge arm can generate three switching states , among which . The output voltage of each phase with respect to the midpoint O can be expressed as:

[0052] (1);

[0053] In the formula, represents the DC bus voltage, represents the bridge arm switching state.

[0054] The space voltage vectors generated by the three-level neutral-point clamped inverter are as Figure 3 shown. The voltage levels "P", "O", and "N" represent , 0, and respectively. The converter can generate a total of kinds of different switching states, thus generating 19 unique voltage vectors. Classified according to their magnitudes, these voltage vectors can be divided into the following four categories: zero vectors, small vectors, medium vectors, and large vectors.

[0055] Among them, the zero vector corresponds to three switching states: (OOO, PPP, NNN); the inner circle has six small vectors, and each small vector corresponds to the following switching states respectively: (OOP, NNO), (OPP, NOO), (OPO, NON), (PPO, OON), (POO, ONN), (ONO, POP); the outer circle has 12 medium vectors / large vectors, and each medium vector corresponds to one of the following switching states: (OPN), (PON), (PNO), (ONP), (NOP), (NPO); each large vector corresponds to one of the following switching states: (NPN), (PPN), (PNN), (PNP), (NNP), (NPP).

[0056] The DC bus voltage can be expressed as the sum of the voltage of the upper capacitor and the voltage of the lower capacitor on the DC side, that is , where represents the voltage of the upper capacitor on the DC side, represents the voltage of the lower capacitor on the DC side. Further, the voltage of the midpoint o of the three-level neutral-point clamped inverter can be expressed as follows:

[0057] (2);

[0058] In the formula, represents the midpoint voltage of a three-level neutral-point clamped inverter, represents the voltage of the upper capacitor on the DC side, represents the voltage of the lower capacitor on the DC side.

[0059] Through the current prediction model, the output current at the next moment is predicted based on the input voltage vector, so as to obtain the current prediction value. After obtaining the current prediction model of the inverter control system, a first cost function is constructed to evaluate all possible voltage vectors in the three-level neutral-point clamped inverter. The first cost function includes the current reference value and the current prediction value.

[0060] Input the voltage vector into the current prediction model, traverse all possible voltage vectors, minimize the first cost function, identify the voltage vector that makes the function value of the first cost function the smallest, and judge the vector type of this voltage vector. Based on the vector type of this voltage vector, through the second cost function, the balance of the midpoint voltage of the three-level neutral-point clamped inverter is realized.

[0061] In a three-level neutral-point clamped inverter, the output voltage vectors can be divided into six large vectors, six medium vectors, six small vectors and one zero vector according to their amplitudes, a total of 19 voltage vectors. There are 27 switching states in the three-level neutral-point clamped inverter, and some of the switching states will generate the same voltage vector. The above-mentioned same voltage vectors are called redundant vectors, and the redundant vectors are mainly small vectors and zero vectors. The amplitudes and phases of each group of redundant vectors are equal, but the influence of each group of redundant vectors on the midpoint voltage is different. When the ideal output voltage vector of the three-level neutral-point clamped inverter belongs to a small vector or a zero vector, the balance of the capacitor voltage can be promoted by selecting the switching state.

[0062] Based on the optimal voltage vector obtained from the first cost function, judge the type of the optimal voltage vector, and the second cost function selects the switching state that is beneficial to the balance of the midpoint potential. Among them, in response to the type of the voltage vector being a large vector or a medium vector, if the current midpoint potential deviation exceeds the preset midpoint potential offset threshold, and the currently selected voltage vector is not conducive to the balance of the capacitor voltage, then return to the first cost function to re-select the voltage vector (for example, the sub-optimal voltage vector with the second smallest value of the first cost function); if the above conditions are not met, then directly output the switching state corresponding to the selected voltage vector.

[0063] According to the above description, a control method for a neutral-point clamped inverter provided by the present application predicts the output current of the inverter by constructing a current prediction model. Under the condition of not requiring the capacitance value information and weight factors of the capacitor, the optimal voltage vector is obtained by minimizing the first cost function, the type of the optimal voltage vector is judged, and further the switching state corresponding to the optimal voltage vector is selected through the second cost function to promote the balance of the midpoint voltage of the neutral-point clamped inverter.

[0064] As an implementation, the first cost function is represented by the following formula:

[0065] (3);

[0066] In the formula, represents the current reference value, represents the predicted current value.

[0067] Through the first cost function, all possible voltage vectors of the three-level neutral-point clamped inverter are traversed, the function values corresponding to each voltage vector are calculated, and the voltage vector with the smallest function value is selected as the preliminary optimal solution to ensure that the output current of the three-level neutral-point clamped inverter is as close as possible to the reference value. Through the first cost function, the optimal voltage vector is screened to provide a high-precision candidate solution for the subsequent neutral-point potential balance of the inverter.

[0068] Furthermore, as an implementation, the predicted current value of the first cost function is obtained by the following formula:

[0069] (4);

[0070] In the formula, R o represents the load resistance value of the inverter, L o represents the load inductance value of the inverter, T s represents the sampling / control frequency of the controller, represents the output current of the inverter in the stationary orthogonal coordinate system, represents the output voltage of the inverter in the stationary orthogonal coordinate system.

[0071] Among them, the inverter output current is determined by the selected voltage vector aiming to minimize the first cost function. The inverter output current and the output voltage under the influence of 19 voltage vectors are substituted into formula (4) for calculation to obtain different predicted current values. The different predicted current values are substituted into the first cost function for calculation, and the voltage vector that makes the first cost function the smallest is selected as the optimal voltage vector.

[0072] Furthermore, as an implementation, the current reference value of the first cost function is obtained by the following formula:

[0073] (5);

[0074] In the formula, i * ( k ) representsk The current reference value at a moment.

[0075] The reference current at a future moment is obtained by Lagrangian extrapolation. The continuous-time models of the output current and the midpoint voltage are discretized by a discretization method to realize the prediction of the driving current corresponding to the voltage vector, which is combined with the first cost function to realize the global optimal selection of the voltage vector.

[0076] After bringing 19 different voltage vectors into the prediction model and calculating through the first cost function, a voltage vector that minimizes the first cost function is obtained. Then, through the second cost function, the switching state of this voltage vector is carefully selected to promote the balance of the midpoint voltage of the inverter.

[0077] As an alternative implementation, using first-order Euler discretization, the continuous-time model of the midpoint voltage of the inverter can be expressed as a discrete prediction model, and the discrete prediction model of the midpoint voltage can be expressed as:

[0078] (6);

[0079] In the formula, v n ([ k+ 1) represents the midpoint voltage at the next moment, v n ([ k ) represents the midpoint voltage at this moment, T s represents the sampling frequency or control frequency of the controller, C dc represents the bus capacitor on the DC side, i n represents the current flowing through the midpoint.

[0080] According to the parameter description of formula (6), the current coefficient is positive, making the midpoint voltage v n ([ k ) and the current i n have opposite signs, so that the midpoint voltage v n ([ k+ 1) tends to 0 to achieve the balance of the midpoint voltage of the inverter.

[0081] Based on the above discrete prediction model of the midpoint voltage, a second cost function can be designed to evaluate the switching state of the optimal voltage vector, and select the switching state that can promote the midpoint voltage v n ([ k+ 1) tending to 0 to achieve the balance of the midpoint voltage.

[0082] As an implementation, the second cost function includes a cost function for evaluating the large vector and the medium vector. The cost function for evaluating the large vector and the medium vector is represented by the following formula:

[0083] (7);

[0084] In the formula, sign(*) represents the sign function, v n ( k ) represents the midpoint voltage under the influence of the selected voltage vector, i n ( k ) represents the current flowing through the midpoint under the influence of the selected voltage vector, b v represents the preset midpoint potential offset threshold.

[0085] Among them, in response to the cost function J b = 2, the voltage vector is reselected based on the first cost function; otherwise, the switching state corresponding to the selected voltage vector is output.

[0086] Specifically, if the cost function J b is equal to 2, it indicates that the sign function sign (| v n ( k )|- b v ) = 1 (indicating that the current voltage deviation of the inverter midpoint exceeds the preset midpoint potential offset threshold b v ), and the sign function sign ( v n ( k ))* sign ( i n ( k )) = 1 (indicating that the signs of the midpoint voltage and current under the influence of the currently selected voltage vector are the same). In this case, the midpoint voltage offset tends to be serious, and the currently selected voltage vector is not conducive to the capacitor voltage balance of the inverter. At this time, the voltage vector is reselected based on the first cost function, and the type of the newly selected voltage vector is judged to avoid further offset of the inverter midpoint voltage.

[0087] If the cost function J bNot equal to 2, indicating that the neutral point potential offset of the inverter at this time does not exceed the preset neutral point potential offset threshold (i.e., within the allowable range of neutral point potential offset), or the signs of the neutral point voltage and current under the influence of the currently selected voltage vector are different (i.e., the neutral point potential offset can be suppressed). At this time, directly output the switching state corresponding to the selected voltage vector.

[0088] As an implementation, the second cost function includes a cost function for evaluating small vectors, and the cost function for evaluating small vectors is expressed by the following formula:

[0089] (8);

[0090] In the formula, sign(*) represents the sign function, v n ( k ) represents the neutral point voltage under the influence of the selected voltage vector, i n ( k ) represents the current flowing through the neutral point under the influence of the selected voltage vector.

[0091] Among them, in response to the type of the selected voltage vector being a small vector, based on the cost function J v , among the two switching states corresponding to the small vector, select and output the switching state that makes the cost function J v the smallest.

[0092] From Figure 3 it can be seen that each small vector corresponds to two switching states. Through the cost function J v multiply the signs of the neutral point voltage and current under the influence of the two switching states corresponding to the voltage vector, and combine formula (6) and formula (8). Among the two switching states corresponding to the small vector, when the calculation result of the cost function J v is negative, it indicates that the signs of the neutral point voltage and current under the influence of the voltage vector are opposite. Then output the switching state that makes the calculation result of the cost function J v negative to achieve the balance of the neutral point voltage of the inverter.

[0093] As an implementation, the second cost function includes a cost function for evaluating zero vectors, and the cost function for evaluating zero vectors is expressed by the following formula:

[0094] (9);

[0095] In the formula, S abc ( krepresents the switching state at time k, represents the two - norm, The role is to minimize the change in the switching state between two control periods;

[0096] Among them, in response to the selected voltage vector being a zero vector, based on the cost function J s , among the three switching states corresponding to the zero vector, select the output that makes the cost function J s the smallest switching state, so as to achieve the minimum switching frequency of the inverter.

[0097] In summary, as Figure 4 shown, the mid - point voltage balance strategy described above does not require the capacitance value information of the inverter DC side during implementation, and does not require the intervention of a weight factor during the balancing process, reducing the dependence of the control method of a mid - point clamped inverter provided by this application on inductance parameters and improving the applicability of the control method of this application.

[0098] As an implementation method, any of the following methods is used for current prediction: the current prediction structure based on the extended Kalman filter, or, the data - driven neural network predictor, or, the current prediction structure based on adaptive predictive control.

[0099] Specifically, the extended Kalman filter is a state estimation algorithm suitable for nonlinear systems. By combining the system dynamic model and real - time measurement data, the state estimate value is recursively updated. The current prediction structure based on the extended Kalman filter can use the state estimate and input voltage vector at the previous moment to predict the current value at the current moment, and can adjust the estimate value by combining the Kalman gain according to the difference between the real - time measured current and the predicted value, suppressing the influence of noise and improving the prediction accuracy.

[0100] Or, use the data - driven neural network predictor for current prediction. By collecting the historical data of the inverter under different working conditions, the neural network predictor learns the input - output mapping relationship of the inverter through offline training, without relying on an accurate physical model and without accurate circuit equations or parameter calibration, and is suitable for processing predicted currents under complex working conditions.

[0101] Or, use the current prediction structure based on adaptive predictive control for current prediction. Adaptive predictive control dynamically adjusts the prediction model by online identifying system parameters to cope with load or environmental changes. The current prediction structure based on adaptive predictive control uses the recursive least - squares method or the gradient - descent method to estimate the load parameters in real time, substitutes the identification result into the discrete state equation, updates the current prediction model, calculates the predicted current based on the updated model, and selects the optimal voltage vector through the cost function, further improving the robustness and real - time performance of the control system.

[0102] As another implementation, the output voltage value can also be directly calculated so that the inverter can reach the desired voltage reference value within one sampling period, thus eliminating the need for step-by-step approximation. By using the discrete state equation of the system, the output voltage at the next moment is predicted, and the switching state is adjusted immediately to achieve the target voltage.

[0103] To further illustrate a control method for a neutral-point clamped inverter provided by the present application, an experimental platform is established below to verify the effectiveness of the method proposed by the present application. The experimental object is a three-level neutral-point clamped inverter, and the detailed parameters of the experimental platform are shown in Table 1:

[0104]

[0105] The experimental results are as Figure 5 shown. A control method for a neutral-point clamped inverter provided by the present application can achieve accurate and effective balance of the DC-side capacitor voltage. In the experiment, the output current waveform of the three-level neutral-point clamped inverter is also tested, and the amplitude, frequency, and waveform quality of the output current are not disturbed or affected. A control method for a neutral-point clamped inverter provided by the present application has high adaptability to the current control target. In the case of load changes or the introduction of external disturbances, a control method for a neutral-point clamped inverter provided by the present application can maintain a good voltage balance effect and at the same time maintain the stability of the output current.

[0106] According to the above description, a control method for a neutral-point clamped inverter provided by the present application predicts the output current of the inverter by constructing a current prediction model. Without the capacitance value information and weight factor of the capacitor, the optimal voltage vector is obtained by minimizing the first cost function, the type of the optimal voltage vector is judged, and the switching state corresponding to the optimal voltage vector is selected through the second cost function, realizing the neutral-point voltage balance of the neutral-point clamped inverter without increasing the complexity of the control method; and through the above verification experiment, a control method for a neutral-point clamped inverter provided by the present application can adapt to inverters of different capacitor types and maintain stable performance.

[0107] In a second aspect, the present application also provides a control device for a neutral-point clamped inverter. The control device controls the neutral-point clamped inverter by applying the control method described above, realizes the prediction of the output current of the inverter, and realizes the neutral-point voltage balance of the inverter.

[0108] It will be understood that the term "exemplary" as used herein means "serving as an example, instance, or illustration". Any embodiment described as "exemplary" is not necessarily preferred or superior to other embodiments and / or does not preclude the combination of features with other embodiments. It should be understood that certain features of the present application described in the context of separate embodiments for clarity may also be provided in combination in a single embodiment. Conversely, the various features of the present application described in the context of a single embodiment for clarity may also be provided separately or in any suitable combination or as any other described embodiment of the present application.

[0109] The foregoing disclosure is only a preferred embodiment of the present application, but it is not intended to limit the scope of the rights of the present application. Those of ordinary skill in the art can understand that: within the spirit and scope of the present application and the appended claims, changes, modifications, substitutions, combinations, and simplifications should all be equivalent replacement methods and still fall within the scope covered by the invention.

Claims

1. A control method for a midpoint clamped inverter, characterized in that: The control method comprises the following steps: Constructing a current prediction model, wherein the current prediction model predicts the output current based on the input voltage vector to obtain a current prediction value; Constructing a first cost function, wherein the first cost function is composed of a current reference value and a current prediction value; With the goal of minimizing the first cost function, selecting a voltage vector and determining a type of the selected voltage vector; constructing a second cost function, wherein the second cost function is configured to: based on the type of the selected voltage vector, select and output a switch state that is conducive to promoting midpoint potential balance; wherein, in response to the type of the selected voltage vector being a large vector or a medium vector, if it is determined that the current midpoint potential deviation exceeds a preset midpoint potential offset threshold, and the currently selected voltage vector is not conducive to capacitor voltage balance, then the voltage vector is reselected based on the first cost function, otherwise, the switch state corresponding to the selected voltage vector is output; The second cost function includes a cost function for evaluating a large vector and a medium vector, and the cost function for evaluating a large vector and a medium vector is expressed by the following formula: In the formula, sign(*) represents the sign function, v n ( k ) represents the midpoint voltage under the influence of the selected voltage vector, i n ( k ) represents the current flowing through the midpoint under the influence of the selected voltage vector, b v represents the preset midpoint potential offset threshold; In which, in response to the cost function J b =2, then reselect the voltage vector based on the first cost function, otherwise, output the switch state corresponding to the selected voltage vector; The first cost function is expressed by the following formula: ; In the formula, represents the current reference value, represents the predicted current value; The current prediction value is obtained by the following formula: In the formula, R o Indicates the load resistance value of the inverter, L o Indicates the load inductance value of the inverter, T s represents the sampling / control frequency of the controller, represents the inverter output current in a stationary orthogonal coordinate system, represents the output voltage of the inverter in a stationary orthogonal coordinate system, where the inverter output current determined by a selected voltage vector with the objective of minimizing said first cost function; The current reference value is obtained by the following formula: ; In the formula, i * ( k )represent k The current reference value at the moment.

2. The control method of the neutral point clamped inverter according to claim 1, characterized in that: The method further comprises: The midpoint voltage continuous time model of the midpoint clamped inverter is expressed as a midpoint voltage discrete prediction model, and the midpoint voltage discrete prediction model is expressed by the following formula: ； In the formula, v n ( k+ 1) indicates the midpoint voltage at the next moment, v n ( k ) represents the midpoint voltage at this moment, T s Indicates the sampling frequency or control frequency of the controller, C dc represents the bus capacitance on the DC side, i n Indicates the current flowing through the midpoint; The second cost function is designed based on the midpoint voltage discrete prediction model, and the switching state of the selected voltage vector is evaluated using the second cost function, and the switching state that can promote the midpoint voltage is selected. v n ( k+ 1) The switching state tends to 0 to achieve mid-point voltage balance.

3. The control method of the neutral point clamped inverter according to claim 2, characterized in that: The second cost function includes a cost function for evaluating a small vector, and the cost function for evaluating a small vector is expressed by the following formula: ; In the formula, sign(*) represents the sign function, v n ( k ) represents the midpoint voltage under the influence of the selected voltage vector, i n ( k ) represents the current flowing through the midpoint under the influence of the selected voltage vector; In response to the type of the selected voltage vector being a small vector, based on the cost function J v , in the two switch states corresponding to the small vector, the output is selected so that the cost function J v Minimum switching state.

4. The control method of the neutral point clamped inverter according to claim 2, characterized in that: The second cost function includes a cost function for evaluating a zero vector, and the cost function for evaluating a zero vector is expressed by the following formula: ; In the formula, S abc ( k ) represents the switch state at time k, represents the two-norm, The role of is to minimize the amount of switch state change between two control cycles; In response to the type of the selected voltage vector being a zero vector, based on the cost function J s , in the three switching states corresponding to the zero vector, the output is selected so that the cost function J s Minimum switching state.

5. The control method of the neutral point clamped inverter according to claim 1, characterized in that: The current prediction is performed by any of the following methods: a current prediction structure based on an extended Kalman filter, or a current prediction structure based on a data-driven neural network predictor, or a current prediction structure based on an adaptive predictive control.

6. A control device for a midpoint clamped inverter, characterized in that: The control device controls the neutral point clamped inverter using the control method according to any one of claims 1 to 5.

Images