Adaptive prediction control method for midpoint potential of LCL grid-connected inverter

Abstract

The invention discloses an LCL grid-connected inverter midpoint potential adaptive prediction control method, and the method comprises the steps: carrying out the discretization of a three-level NPC type LCL grid-connected inverter based on a forward Euler method, and building a prediction model; secondly, constructing a value function including grid-connected current tracking, filter capacitor voltage, inverter side current and a neutral-point potential balance term, wherein the neutral-point potential balance term is weighted by a variable weight coefficient; according to the method, a model-free adaptive control theory is innovatively introduced, a dynamic data model between midpoint potential deviation and a weight coefficient is established online by using a tight-format dynamic linearization method, and an adaptive updating law of the weight coefficient is designed through real-time estimation of a pseudo-partial derivative. According to the method, complex working conditions such as current step change can be effectively handled, neutral-point potential fluctuation can be strictly limited within 1V without manually setting parameters, cooperative control of grid-connected current high-precision tracking and neutral-point potential balance is realized, and the robustness and dynamic and steady-state performance of a system are remarkably improved.

Description

Technical Field

[0001] This invention relates to a predictive control method, specifically to an adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter. Background Technology

[0002] With the rapid development of distributed generation technologies such as photovoltaics and wind power, grid-connected inverters, as the core interface for energy conversion, directly affect the power quality of the power grid. Three-level NPC topologies are widely used in medium- and high-voltage, high-power applications due to their advantages such as high output voltage waveform quality and low voltage stress on switching transistors. To further suppress high-frequency switching harmonics, LCL filters are widely adopted. However, NPC inverters inherently suffer from DC-side midpoint potential imbalance. Excessive midpoint potential fluctuations not only lead to increased low-order harmonics and output voltage waveform distortion but also increase voltage stress on the switching transistors, potentially damaging components or causing DC-side capacitor overvoltage breakdown in severe cases.

[0003] Model Predictive Control (MPC) is a highly promising control strategy that uses a mathematical model of the system to predict future states and selects the optimal control action by minimizing the value function. In existing NPC inverter MPC strategies, a "midpoint potential balance term" is typically introduced into the value function and weighted by a fixed coefficient. This is to adjust its relative importance with the "current tracking term". The main drawback of the existing technology is: 1. Difficulty in setting weighting coefficients: Fixed weighting coefficients usually rely on repeated trial and error based on human experience, lacking a systematic design theory; 2. Insufficient robustness: In actual operation, when operating conditions change drastically, a fixed weighting coefficient often cannot simultaneously ensure current quality and midpoint balance. Too small a weighting coefficient leads to midpoint potential divergence, while too large a weighting coefficient results in increased total harmonic distortion (THD) of the grid-connected current. Therefore, an adaptive control method that can automatically adjust the weighting coefficient under changing operating conditions is urgently needed. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides an adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter. This method eliminates the need for tedious manual adjustments of weight coefficients under complex operating conditions to achieve optimal results, and solves the problem of performance degradation of existing fixed-weight model predictive control under complex operating conditions of LCL grid-connected inverters. It also exhibits strong robustness.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an adaptive predictive control method for the neutral point potential of an LCL grid-connected inverter, comprising the following steps: Step 1: Collect real-time operating data of the LCL grid-connected inverter, including the DC side upper and lower capacitor voltages. , Inverter-side current Grid-connected current Filter capacitor voltage and grid voltage Establish a mathematical model for the LCL grid-connected inverter; Step 2: Based on the discretized mathematical model of the LCL filter and combined with the given value of the grid-connected current, calculate the reference predicted value of the grid-connected current. Filter capacitor voltage reference predicted value Inverter-side current reference prediction value Predict all possible switching states in Predicted grid-connected current at time [time] Predicted value of filter capacitor voltage Inverter-side current prediction value Midpoint potential prediction value ; Step 3: Define the value function for the discretized model predictive control of the LCL grid-connected inverter. This value function includes the inverter-side current prediction error term, the filter capacitor voltage prediction error term, the grid-connected current prediction error term, and the midpoint potential balance term, where the weighting coefficient of the midpoint potential balance term is specified. These are time-varying parameters; Step 4: Based on the compact scheme dynamic linearization theory, establish a dynamic data relationship model between the change in midpoint potential deviation and the change in weighting coefficients, and use the projection algorithm to estimate the pseudo-partial derivatives of the model online. ; Step 5: Based on the estimated pseudo-partial derivatives The optimal weight coefficients at the current time are calculated using the adaptive update law of weight coefficients. ; Step Six: Substitute the value function described in step three, iterate through all valid switching states, and select the switching state that minimizes the value function as the control signal for the grid-connected inverter at the next moment.

[0006] Furthermore, the dynamic data relationship model in step four is based on expressing the nonlinear relationship between the midpoint potential deviation and the weighting coefficient as follows: ; in, They represent in The midpoint potential deviation input and weighting coefficient output at time t. and It is an unknown positive integer. It is an unknown nonlinear function that describes the mapping relationship between the midpoint potential deviation input and the weighting coefficient output.

[0007] Furthermore, the dynamic data relationship model in step four is represented as follows: ; in, Output the change amount for the dynamic data relationship model system, i.e., the CFDL model system; This represents the change in the weighting coefficients; It is a pseudo-partial derivative, representing the dynamic influence rate of changes in the weighting coefficients on changes in the midpoint potential deviation.

[0008] Furthermore, the characteristic is that the pseudo-partial derivatives in step four... The estimation algorithm is based on the following pseudo-partial derivative (PDD) estimation of the standard function to solve for the extrema: ; in, A weighting penalty factor greater than 0 is used to limit the rate of change of pseudo-partial derivatives.

[0009] Furthermore, the pseudo-partial derivatives The online estimation algorithm is as follows: ; in, This is an estimate of the pseudo-partial derivative. This is the step size factor.

[0010] Furthermore, the pseudo-partial derivatives The online estimation process also includes initialization judgment and range limitation procedures, and the specific algorithm is as follows: ; in, The initial value of the pseudo-partial derivative is given. It is a very small positive integer.

[0011] Furthermore, the derivation of the adaptive update law for the weight coefficients in step five is based on the following control input index function: ; in, This serves as a reference value for the midpoint potential deviation at the next sampling time. As another weighting coefficient; by applying this index function to By taking the derivative and setting it to zero, the adaptive update law of the weight coefficients is derived.

[0012] Furthermore, the adaptive update law for the weight coefficients in step five is as follows: ; in, The reference value for the midpoint potential deviation is set to 0; To control the step size factor.

[0013] Compared with the prior art, the present invention has the following advantages: 1. Strong adaptive robustness: The algorithm does not rely on manually set weight parameters, but is driven solely by input and output data. When operating conditions change, it can automatically sense the fluctuation trend of the midpoint potential and dynamically adjust the weights without the need for manual parameter readjustment; 2. High control precision: This method resolves the contradiction between "steady-state error" and "dynamic response" under fixed weights. Experiments show that it can strictly limit the midpoint potential fluctuation to within 1V while ensuring that the total harmonic distortion of the grid-connected current remains at a low level. 3. Good compatibility and easy to implement: The adaptive algorithm is embedded as an independent module in the MPC framework, with low computational load, making it suitable for real-time operation in digital controllers such as DSPs. Attached Figure Description

[0014] Figure 1 This is a flowchart of the control method of the present invention; Figure 2 This is a schematic diagram of the control system structure of the control method of the present invention; Figure 3 This is a diagram showing the basic voltage vector distribution of the three-level LCL inverter output in an embodiment of the present invention. Figure 4 The graph shows the grid-connected current waveform and the midpoint potential effect of the traditional model under the condition that the grid-connected current reference value is 50A. Figure 5 The graph shows the grid-connected current waveform and the effect of the midpoint potential for predictive control under the condition that the grid-connected current reference value is 60A, provided by the traditional model. Figure 6 The diagram shows the grid current waveform and midpoint potential effect under the condition that the grid current reference value is 50A, according to an embodiment of the present invention. Figure 7 The diagram shows the grid current waveform and midpoint potential effect under the condition that the grid current reference value is 60A, according to an embodiment of the present invention. Detailed Implementation

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, the present invention provides an adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter.

[0018] Step 1: Collect real-time operating data of the LCL grid-connected inverter, including the DC side upper and lower capacitor voltages. , Inverter-side current Grid-connected current Filter capacitor voltage and grid voltage Establish a mathematical model for the LCL grid-connected inverter; Step 2: Based on the discretized mathematical model of the LCL filter and combined with the given value of the grid-connected current, calculate the reference predicted value of the grid-connected current. Filter capacitor voltage reference predicted value Inverter-side current reference prediction value Predict all possible switching states in Predicted grid-connected current at time [time] Predicted value of filter capacitor voltage Inverter-side current prediction value Midpoint potential prediction value ; Step 3: Define the value function for the discretized model predictive control of the LCL grid-connected inverter. This value function includes the inverter-side current prediction error term, the filter capacitor voltage prediction error term, the grid-connected current prediction error term, and the midpoint potential balance term, where the weighting coefficient of the midpoint potential balance term is specified. These are time-varying parameters; Step 4: Based on the compact scheme dynamic linearization theory, establish a dynamic data relationship model between the change in midpoint potential deviation and the change in weighting coefficients, and use the projection algorithm to estimate the pseudo-partial derivatives of the model online. ; The dynamic data relationship model is represented as: ; in, Output the change amount for the CFDL model system; This represents the change in the weighting coefficients; It is a pseudo-partial derivative, representing the dynamic influence rate of changes in the weighting coefficients on changes in the midpoint potential deviation.

[0019] pseudo-partial derivatives The online estimation algorithm is as follows: ; in, This is an estimate of the pseudo-partial derivative. Step size factor This is the weighting penalty factor.

[0020] Step 5: Based on the estimated pseudo-partial derivatives The optimal weight coefficients at the current time are calculated using the adaptive update law of weight coefficients. ; The adaptive update law for weight coefficients is: ; in, The reference value for the midpoint potential deviation is set to 0; To control the step size factor, This is another weighting coefficient.

[0021] Step Six: Substitute the value function described in step three, iterate through all valid switching states, and select the switching state that minimizes the value function as the control signal for the grid-connected inverter at the next moment.

[0022] Figure 2 This is a schematic diagram of the control system structure of an adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter disclosed in an embodiment of the present invention. The control system includes core functional modules such as a coordinate transformation module, a reference value calculation module, a data-driven calculation module, and a model predictive control module. These modules work together to achieve adaptive control of the midpoint potential of the LCL grid-connected inverter.

[0023] In the data acquisition and coordinate transformation section, the coordinate transformation module serves as the link between the physical system and the control algorithm, acquiring the grid voltage of the LCL grid-connected inverter in real time. Grid-connected current Filter capacitor voltage and inverter-side current This module performs Clark transformation and inverse Park transformation, converting the AC components in the three-phase stationary coordinate system into components in the two-phase stationary coordinate system, and the components in the rotating coordinate system into components in the two-phase stationary coordinate system, and then sending them as feedback signals to the reference value calculation module, the model predictive control module, and the data-driven calculation module, respectively.

[0024] The reference value calculation module is responsible for generating the reference signal for system tracking. This module receives an externally provided grid-connected current reference value. In addition to grid voltage information provided by the coordinate transformation module, the grid-connected current reference vector, filter capacitor voltage reference vector, and inverter-side current reference vector are calculated based on the mathematical model of the LCL grid-connected inverter, for the current and predicted times. The calculation results are then transformed and output. Reference values ​​in coordinate system , and To the model prediction and control module.

[0025] The data-driven calculation module is the adaptive core of this invention, used for online adjustment of the weighting coefficients of the midpoint potential in the value function. This module receives the midpoint potential deviation signal monitored in real time and establishes a dynamic data relationship model between the change in midpoint potential deviation and the change in weighting coefficients using compact scheme dynamic linearization (CFDL) theory. The pseudo-partial derivatives (PPDs) of this model are estimated online using a projection algorithm, and combined with the adaptive update law of the weighting coefficients, the optimal real-time weighting coefficients of the midpoint potential are calculated in real time. This coefficient is then output to the model predictive control module to dynamically balance current tracking performance with midpoint potential stability.

[0026] The model predictive control module is responsible for determining the optimal switching state. This module receives real-time sampled data from the coordinate transformation module, reference predicted values ​​from the reference value calculation module, and adaptive weighting coefficients from the data-driven calculation module. Internally, the module integrates a discretized predictive model of the LCL inverter, calculating the value of the value function by traversing all valid voltage vector switching states. Finally, the switching state that minimizes the value function is selected as the control signal for the next time step. The power devices that act on the grid-connected inverter drive the system to run along a predetermined trajectory.

[0027] This embodiment takes a three-phase three-level LCL grid-connected inverter as the object and adopts a data-driven method based on tight-format dynamic linearization to achieve adaptive adjustment of the neutral point potential weighting coefficient. The specific strategy is as follows: Step 1: Collect real-time operating data of the LCL grid-connected inverter and establish an LCL grid-connected inverter topology model. According to Kirchhoff's voltage and current laws, in the three-phase stationary abc coordinate system, the voltage and current equations of the LCL grid-connected inverter are as follows: ; in, For the grid-side inductance and its equivalent resistance, This refers to the inverter-side inductance and its equivalent resistance. This is a filter capacitor. This is the grid voltage. For grid-connected current, This is the voltage across the filter capacitor. Inverter-side current, This is the inverter output voltage.

[0028] Define the midpoint potential deviation as The fluctuation of the midpoint potential depends on the current flowing into the midpoint of the DC side. ; In the formula, For DC side capacitors, , The current is the midpoint current, with the direction of flow into the midpoint of the DC side taken as the positive direction.

[0029] Based on the working principle and topology of the LCL grid-connected inverter, we can conclude that: ; In the formula, The switching states of the three-phase bridge arms are 1, 0, and -1.

[0030] To simplify calculations and achieve decoupled control, the above model is transformed to a two-phase stationary coordinate system using the Clark transformation. Down.

[0031] Step 2: Discretize the data according to the forward Euler formula, calculate the reference values, and establish the prediction model. This method uses a multivariate model for prediction, incorporating the filter capacitor voltage into the control, without needing to consider additional damping strategies or damping components. At this point, reference value prediction is required for three variables: inverter-side current, filter capacitor voltage, and grid-connected current. Based on the given grid-connected current reference value, the current time-varying value is constructed. Grid-connected current reference vector Based on the mathematical model of the LCL grid-connected inverter in the dq coordinate system, we obtain: ; In the formula, for Reference value of the filter capacitor voltage in the dq coordinate system at time t. for The value of the grid voltage in the dq coordinate system at time 1. For grid voltage angular velocity, for Reference value of grid-connected current in the dq coordinate system at time t. for Reference value of inverter-side current in the dq coordinate system at time t.

[0032] By performing an inverse Park transformation on the reference variable dq component obtained from the above equation, the reference value of the grid-connected inverter side current in the αβ coordinate system can be obtained. Reference value of filter capacitor voltage and grid-connected current reference value Second-order Lagrange interpolation method is used for prediction. The reference predicted value at the sampling time is given by... The reference value for time is ,but: ; In the formula, It is the current Reference value at any time, and Given the reference values ​​from the previous and previous time points respectively, the reference predicted value of the grid-connected inverter side current can be obtained. Filter capacitor voltage reference predicted value and grid-connected current reference prediction value .

[0033] Using the forward Euler method, according to Sample value at time and the optimal voltage vector calculated and applied in the previous control cycle. ,predict The state at time t is such that, after discretizing the control equation of the inverter-side current in the αβ coordinate system, its prediction model is obtained as follows: ; in, For inverter-side current at The predicted value at the sampling time; For the inverter output voltage at The sampled value at time; For the filter capacitor voltage at The sampled value at time; For inverter-side current at The sampled value at time; This is the system sampling period.

[0034] Considering the linear average value of the inverter current prediction trajectory replace ,in By discretizing the control equation of the filter capacitor voltage in the αβ coordinate system using the forward Euler formula, a prediction model for the filter capacitor voltage is obtained: ; in, For the voltage of the filter capacitor at The predicted value at the sampling time; For the filter capacitor voltage at The sampled value at time; For inverter-side current at Predicted value at sampling time; For inverter-side current at The sampled value at time; For grid-connected current in The sampled value at time.

[0035] Utilizing the linear discretization characteristic of the forward Euler formula, the linear average value of the filter capacitor voltage is used. replace ,in At this time, the prediction model for the grid-connected current is: ; in, For grid-connected current in The predicted value at the sampling time; For grid-connected current in The sampled value at time; For the grid voltage at The sampled value at time.

[0036] Based on the current switch state Calculate the midpoint current ,get Midpoint potential value at time: ; Step 3: Define the value function of the discretized model predictive control for the LCL grid-connected inverter. This is used for subsequent scrolling optimization. ; in, For the fixed weight of the filter capacitor voltage, is the adaptive weighting coefficient for the midpoint potential, and is the time-varying coefficient.

[0037] Step 4: Based on the CFDL model, data-driven calculations are performed to derive the nonlinear relationship between the midpoint potential deviation and the weighting coefficients. A data-driven calculation module is designed, and the weighting coefficients of the adaptive midpoint potential balance term are used for model predictive control. The nonlinear relationship between the midpoint potential deviation and the weighting coefficients in an NPC-type three-level LCL grid-connected inverter can be expressed as: ; in, They represent in The midpoint potential deviation input and weighting coefficient output at time t. and They are two unknown positive integers. It is a nonlinear function.

[0038] Simplify the equation, let Let the output change value be between two adjacent sampling times. Input the change values ​​for two adjacent time points. For a nonlinear system that meets the conditions, there must be one value at any given time. Bounded partial derivatives , making ; To calculate the pseudo-partial derivative value at the current time The standard function for PPD (pseudo-partial derivative) estimation is set as follows: ; in, As a weighting penalty factor, The larger the value, the better. The more stable the change, the better. The input change value from the previous time step. This is the pseudo-partial derivative value from the previous time step.

[0039] Solve the above equation regarding The extreme value is obtained. The estimation algorithm is as follows: ; in It is a step size factor that makes the algorithm more flexible and versatile. The larger, The larger the search range, the better.

[0040] To limit the variation range of PPD by incorporating an initialization judgment process into the proposed compact scheme dynamic linearization algorithm, the overall calculation strategy for the weight coefficients of the midpoint potential balance term based on the PPD estimation algorithm is as follows: ; in, These are the initial values ​​of the pseudo-partial derivatives. It is a very small positive integer value.

[0041] Step 5: Based on the real-time estimated pseudo-partial derivatives (PPD), further calculate the adaptive weighting coefficients for the midpoint potential. To ensure the midpoint potential value of the control is close to the reference value, a control input index function is used to design the control law: ; in, This serves as the reference value for the midpoint potential at the next sampling time. This is another weighting coefficient.

[0042] Control laws Differentiating and setting it to zero yields the equation for calculating the weighting coefficient of the midpoint potential equilibrium term: ; in, The step size factor is used to adjust the control algorithm to make it more versatile. This is the pseudo-partial derivative value calculated at the current sampling time. The reference value for the midpoint potential deviation is set to 0.

[0043] The value of the weighting coefficient of the midpoint potential balance term is calculated based on the relationship between the weighting coefficient of the midpoint potential balance term and the midpoint potential.

[0044] Step Six: Substitute the calculated weight coefficients of the midpoint potential balance term into the model predictive control to calculate the predicted current value corresponding to each voltage vector and perform cost function calculation. Based on the sector where the LCL inverter is located at the current sampling time, select the switching state of adjacent voltage vectors, substitute the weight coefficients of the midpoint potential balance term calculated in real time into the model current prediction calculation formula, calculate the current value of each vector at the next moment, and output the control signal for rolling optimization.

[0045] The optimal switching state of the voltage vector is selected through rolling optimization, and a control signal is output. The value function of the switching states of each adjacent vector in the sector is calculated, and the switching state with the smallest value function is selected as the control signal output to control the operation of the LCL grid-connected inverter. The voltage vector sequence is as follows: Figure 3 As shown.

[0046] To better verify the feasibility and effectiveness of the above method, real-time simulation verification was performed in MATLAB / Simulink. The parameter diagram of the simulation system of the three-level LCL grid-connected inverter of this invention is shown in the table below:

[0047] Figure 4 This is a graph obtained after multiple manual adjustments of the weighting coefficients, based on the traditional model prediction with a grid-connected current d-axis reference value of 50A. Figure 4 a is the grid-connected current waveform. Figure 4 b is the THD diagram of the grid-connected current. Figure 4 c is the effect diagram of controlling the midpoint potential, from Figure 4 As can be seen from the effect diagram, the grid-connected current follows the given value well, the THD is 2.78%, and the midpoint potential is around ±5V, indicating good performance.

[0048] Figure 5 This is the result of using the above weighting coefficients under the traditional model prediction after the grid-connected current d-axis reference value is changed to 60A. Figure 5 a is the grid-connected current waveform. Figure 5 b is the THD diagram of the grid-connected current. Figure 5 c is the effect diagram of controlling the midpoint potential. It can be seen from the effect diagram that the grid-connected current follows the given value well. However, after the operating conditions change, the THD rises to 2.91%, the midpoint potential rises, and the performance deteriorates to around ±8V.

[0049] Figure 6 and Figure 7 The diagram illustrates the effects of this invention when the grid-connected current d-axis reference values ​​are 50A and 60A, respectively. Figure 6 a, Figure 7 a) are all grid-connected current waveforms. Figure 6 b、 Figure 7 b represents the THD diagram of the grid-connected current. Figure 6 c. Figure 7 c represents the effect diagram of controlling the midpoint potential. As can be seen from the effect diagram, the grid-connected current can follow the given value well, and the THD has decreased, from 2.78% to 1.96% and from 2.91% to 1.67%. The midpoint potential is controlled around ±1V. When the operating conditions change, there is no need to adjust the midpoint potential weighting coefficient to achieve good performance. It can realize the coordinated adaptive control of grid-connected current and midpoint potential.

[0050] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any minor modifications, equivalent substitutions, and improvements made to the above embodiments based on the technical essence of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for adaptive predictive control of the midpoint potential in an LCL grid-connected inverter, characterized in that, Includes the following steps: Step 1: Collect real-time operating data of the LCL grid-connected inverter, including the DC side upper and lower capacitor voltages. , Inverter-side current Grid-connected current Filter capacitor voltage and grid voltage Establish a mathematical model for the LCL grid-connected inverter; Step 2: Based on the discretized mathematical model of the LCL filter and combined with the given value of the grid-connected current, calculate the reference predicted value of the grid-connected current. Filter capacitor voltage reference predicted value Inverter-side current reference prediction value Predict all possible switching states in Predicted grid-connected current at time [time] Predicted value of filter capacitor voltage Inverter-side current prediction value Midpoint potential prediction value ; Step 3: Define the value function for the discretized model predictive control of the LCL grid-connected inverter. This value function includes the inverter-side current prediction error term, the filter capacitor voltage prediction error term, the grid-connected current prediction error term, and the midpoint potential balance term, where the weighting coefficient of the midpoint potential balance term is specified. These are time-varying parameters; Step 4: Based on the compact scheme dynamic linearization theory, establish a dynamic data relationship model between the change in midpoint potential deviation and the change in weighting coefficients, and use the projection algorithm to estimate the pseudo-partial derivatives of the model online. ; Step 5: Based on the estimated pseudo-partial derivatives The optimal weight coefficients at the current time are calculated using the adaptive update law of weight coefficients. ; Step Six: Substitute the value function described in step three, iterate through all valid switching states, and select the switching state that minimizes the value function as the control signal for the grid-connected inverter at the next moment.

2. The adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter according to claim 1, characterized in that, The dynamic data relationship model in step four is based on expressing the nonlinear relationship between the midpoint potential deviation and the weighting coefficient as follows: ； in, They represent in The midpoint potential deviation input and weighting coefficient output at time t. and It is an unknown positive integer. It is an unknown nonlinear function that describes the mapping relationship between the midpoint potential deviation input and the weighting coefficient output.

3. The adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter according to claim 2, characterized in that, The dynamic data relationship model in step four is represented as follows: ； in, Output the change amount for the dynamic data relationship model system; This represents the change in the weighting coefficients; It is a pseudo-partial derivative, representing the dynamic influence rate of changes in the weighting coefficients on changes in the midpoint potential deviation.

4. The pseudo-partial derivative of the rate of dynamic influence of changes in weighting coefficients on changes in midpoint potential deviation as described in claim 3, characterized in that, The pseudo-partial derivative in step four The estimation algorithm is based on solving for the extrema of the standard function using the following pseudo-partial derivative estimation: ； in, A weighting penalty factor greater than 0 is used to limit the rate of change of pseudo-partial derivatives.

5. The adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter according to claim 4, characterized in that, The pseudo-partial derivative The online estimation algorithm is as follows: ； in, This is an estimate of the pseudo-partial derivative. This is the step size factor.

6. The adaptive predictive control method for the neutral point potential of an LCL grid-connected inverter according to claim 5, characterized in that, The pseudo-partial derivative The online estimation process also includes initialization judgment and range limitation procedures, and the specific algorithm is as follows: ； in, The initial value of the pseudo-partial derivative is given. It is a very small positive integer.

7. The adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter according to claim 1, characterized in that, The derivation of the adaptive update law for the weight coefficients in step five is based on the following control input index function: ； in, This serves as a reference value for the midpoint potential deviation at the next sampling time. As another weighting coefficient; by applying this index function to By taking the derivative and setting it to zero, the adaptive update law of the weight coefficients is derived.

8. The adaptive predictive control method for the midpoint potential of an LCL grid-connected inverter according to claim 7, characterized in that, The adaptive update law for the weight coefficients in step five is as follows: ； in, The reference value for the midpoint potential deviation is set to 0; To control the step size factor.

Images