Improved data-driven energy storage inverter current predictive control method and system

Abstract

The application discloses an improved data-driven energy storage inverter current prediction control method and system, comprising: establishing a mathematical model in a static coordinate system of an energy storage inverter, analyzing to obtain a plurality of basic voltage vectors, and synthesizing a plurality of virtual vectors; estimating and compensating measurement noise; establishing a current gradient relationship corresponding to the plurality of basic voltage vectors, and updating the current gradient of the remaining vectors; performing current prediction, and analyzing to obtain a value function value corresponding to the plurality of basic voltage vectors and an action time of the plurality of virtual vectors; analyzing to obtain a current prediction value corresponding to the plurality of virtual vectors and a virtual voltage vector with the minimum value function value, and taking the virtual voltage vector with the minimum value function value as an optimal voltage vector to act on a next control period. The current gradient update stagnation problem is solved, current spikes are avoided, the prediction result is affected by measurement noise, the robustness of parameter changes is improved, and the ripple of output current is reduced.

Description

Technical Field

[0001] This invention relates to the field of energy storage inverter technology, and in particular to an improved data-driven current prediction and control method and system for energy storage inverters. Background Technology

[0002] In the field of power electronics, for energy storage inverters, data-driven predictive current control is generally used to eliminate the influence of system parameters on predictive control. However, when performing data-driven predictive current control on energy storage inverters, the gradient updated by current measurement is only applicable to the applied vector, and the current gradients corresponding to other remaining vectors remain at their old values. Therefore, data-driven predictive current control of energy storage inverters suffers from the problem of current gradient update stagnation. Stagnation in current gradient update can lead to large current spikes in the output current of the energy storage inverter.

[0003] In existing technologies, when addressing the problem of current gradient update stagnation in data-driven predictive current control methods for energy storage inverters, the impact of measurement noise on the current gradient is not considered. This results in the current prediction results being susceptible to measurement noise and having large errors. Furthermore, the limited number of vectors used leads to large output current ripple.

[0004] Therefore, there is a need for a multi-vector energy storage inverter data-driven predictive current control method that can take into account measurement noise and solve the problem of stagnant current gradient updates in data-driven prediction of energy storage inverters. Summary of the Invention

[0005] The purpose of this invention is to provide an improved data-driven current prediction control method and system for energy storage inverters, which can realize real-time and rapid updates of all current gradients in the lookup table. By using an extended state observer, measurement noise is estimated and compensated, thus solving the problems of large current ripple and susceptibility to measurement noise in traditional data-driven predictive current control methods.

[0006] To address the aforementioned technical problems, this invention provides an improved data-driven current prediction control method for energy storage inverters, comprising the following steps:

[0007] S1. Establish a mathematical model of the energy storage inverter in the static coordinate system, and obtain multiple basic voltage vectors based on the state of the three-phase switching transistors of the energy storage inverter. Use the current gradient as the prediction model of the energy storage inverter and synthesize multiple virtual vectors.

[0008] S2, to estimate and compensate for the measurement noise of the grid current measured by the energy storage inverter;

[0009] S3, establish the current gradient relationship with the multiple basic voltage vectors, and update the current gradient corresponding to the remaining vectors;

[0010] S4, perform initial current prediction, and obtain multiple value function values ​​corresponding to the multiple basic voltage vectors and the action time of the multiple virtual vectors through value function equation analysis;

[0011] S5. Based on the updated current gradient and the application time of the multiple virtual vectors, the predicted current values ​​corresponding to the multiple virtual vectors are analyzed and obtained. The virtual voltage vector with the smallest value function value is obtained through the value function and applied as the optimal voltage vector to the next control cycle.

[0012] Optionally, the plurality of basic voltage vectors are eight basic voltage vectors, namely u0(0,0,0), u1(1,0,0), u2(1,1,0), u3(0,1,0), u4(0,1,1), u5(0,0,1), u6(1,0,1), and u7(1,1,1).

[0013] The mathematical model of the energy storage inverter in the static coordinate system is as follows:

[0014]

[0015] Among them, u x i is the output voltage of the energy storage inverter. g e is the output current of the energy storage inverter. g t is the grid-side voltage of the energy storage inverter, L is the filter inductance, R is the filter resistance, and t is time.

[0016] The plurality of virtual vectors consists of six virtual vectors.

[0017] Optionally, in S2, an extended state observer is established to estimate and compensate for the noise in the measured grid current. This extended state observer is capable of measuring current error feedback. The extended state observer is:

[0018]

[0019] Analysis of the extended state observer yields:

[0020]

[0021] Among them, i e The estimated value of the output current of the energy storage inverter is Δi. e The measured current gradient is the estimated value, δ1 is the feedback error gain of the sampled current, δ2 is the feedback error gain of the sampled grid current, and T is the estimated value of the actual measured current gradient. s To control the cycle, i g i is the output current of the energy storage inverter. e – ig Let t be the current error and t be the time.

[0022] Optionally, the current gradient update formula is expressed as:

[0023]

[0024]

[0025]

[0026] ,

[0027] Where, △i m For the current gradient of the fundamental voltage vector in the first application, t m The duration of the basic voltage vector in the first application is given by t, where A is the scaling factor of the basic voltage vector in the first application. n Let B be the duration of the basic voltage vector for the second application, and let Δi be the scaling factor for the basic voltage vector for the second application. n For the second application, the current gradient of the fundamental voltage vector, Δi z The current gradient is the zero voltage vector, Δi x Let Δi be the current gradient of the sampled current. y Let Δi1 be the current gradient of the remaining voltage vector, Δi4 be the current gradient of the first basic voltage vector among the plurality of basic voltage vectors, k-1 be the current gradient of the fourth basic voltage vector among the plurality of basic voltage vectors, and T be the current gradient of the last control time. s For the control period, t1 is the duration of the first basic voltage vector, k-2 is the previous control time, and Δi 2,6 t represents the current gradient of the second and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z t4 represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, t4 represents the duration of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, and Δi represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors. 3,5 The current gradient Δi is the current gradient of the third and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. 1,4 The current gradient is the second and fourth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0028] Optionally, in S4, the plurality of basic voltage vectors are added to the sampled current to perform current prediction, and the current prediction formula is:

[0029]

[0030] Among them, i iαβ Let i be the grid current in the stationary coordinate system, k+1 be the next control time, and i αβ Let Δi be the sampling current in the stationary coordinate system. iαβ Let be the current gradient corresponding to the i-th basic voltage vector, and k be the corresponding time.

[0031] The value function is:

[0032]

[0033] Where G is the value function value, i refα Let i be the reference current of the energy storage inverter in the α stationary coordinate system. gα Let i be the grid current of the energy storage inverter in the α stationary coordinate system, k+1 be the next control time, and i refβ Let i be the reference current of the energy storage inverter in the β stationary coordinate system. gβ The grid current of the energy storage inverter in the β stationary coordinate system is given.

[0034] Optionally, in S5, the value obtained from the current prediction is substituted into the value function to obtain the corresponding value function value;

[0035] The duration of action of the multiple virtual vectors is:

[0036]

[0037] Among them, t m For the duration of action of the fundamental voltage vector in the first application, t n For the duration of action of the fundamental voltage vector in the second application, t z G represents the duration of action of the first and eighth fundamental voltage vectors of the fundamental voltage vectors. n For the value function of the basic voltage vector in the second application, G z G is the value function value of the zero voltage vector. m For the value function of the basic voltage vector in the first application, T s To control the cycle.

[0038] Optionally, when three basic voltage vectors are applied in each control cycle, the predicted current value at time k+1 is:

[0039]

[0040] Among them, i g t is the output current of the energy storage inverter.m The duration of the first application of the basic voltage vector is given by T, where k+1 is the next control time, k is the current control time, and T is the duration of the first application of the basic voltage vector. s To control the period, △i m Let t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n For the second application of the basic voltage vector, the duration of action of the basic voltage vector is Δi. n Let be the current gradient of the nth basic voltage vector among the plurality of basic voltage vectors, k be the current control time, and t be the current gradient of the nth basic voltage vector. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors;

[0041] The predicted current value at time k+2 is:

[0042]

[0043] Among them, i g The output current of the energy storage inverter is k, k+2 is the next control time, k+1 is the next control time, and t is the next control time. m The duration of the first application of the fundamental voltage vector, T. s To control the period, △i m Let t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n For the second application of the basic voltage vector, the duration of action of the basic voltage vector is Δi. n Let t be the current gradient of the nth fundamental voltage vector among the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0044] The present invention also provides an improved data-driven current prediction control system for energy storage inverters, comprising:

[0045] The parameter setting module is used to establish a mathematical model of the energy storage inverter in a static coordinate system, obtain multiple basic voltage vectors based on the state of the three-phase switching transistors of the energy storage inverter, and use the current gradient as a prediction model of the energy storage inverter to synthesize multiple virtual vectors.

[0046] The noise estimation module is used to estimate and compensate for the measurement noise of the grid current measured by the energy storage inverter.

[0047] The current gradient update module is used to establish the current gradient relationship with the multiple basic voltage vectors and update the current gradient corresponding to the other remaining vectors.

[0048] The current prediction module is used to obtain multiple value function values ​​corresponding to the multiple basic voltage vectors through value function equation analysis, and to obtain the action time of the multiple virtual vectors. Based on the updated current gradient and the action time of the multiple virtual vectors, it analyzes and obtains the current prediction value corresponding to the multiple virtual vectors, and obtains the virtual voltage vector with the minimum value function value through the value function, and applies the virtual voltage vector as the optimal voltage vector to the next control cycle.

[0049] Compared with the prior art, the present invention has at least the following beneficial effects:

[0050] This invention provides an improved data-driven current prediction control method for energy storage inverters. Based on the coordinate relationship of the voltage vector, a model is established, and the current gradient is updated in real time and quickly using the current gradient update formula. This solves the problem of current gradient update stagnation in data-driven predictive current control of energy storage inverters, eliminates current spikes in the output current of energy storage inverters, and reduces prediction errors.

[0051] Taking into account the measurement noise of the grid current in the data-driven predictive current control of the energy storage inverter, the measurement noise of the grid current was estimated and compensated. Under the condition of fast dynamic response speed, the problem of the data-driven predictive current control of the energy storage inverter being easily affected by measurement noise was solved, and the problem of large current prediction error was further solved.

[0052] Furthermore, by applying three vectors in each control cycle, the robustness to parameter changes is improved. Without increasing the lookup table, the voltage vector applied in each control cycle is increased, reducing the impact of sampling disturbances and the ripple of the output current, and improving the quality of the output current.

[0053] Furthermore, the improved data-driven energy storage inverter current prediction control method predicts current based on the current gradient stored in a lookup table, achieving real-time updates of all current gradients in the lookup table without requiring any system parameters, and maintaining good current performance even when parameters are mismatched. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the control flow according to an embodiment of the present invention;

[0055] Figure 2 This is a topology diagram of an energy storage inverter according to an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of a three-vector synthesis according to an embodiment of the present invention;

[0057] Figure 4 This is a control block diagram of an extended state observer according to an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of a current gradient synthesis method according to an embodiment of the present invention;

[0059] Figure 6 This is a schematic diagram illustrating real-time current gradient updates according to an embodiment of the present invention;

[0060] Figure 7a Current gradient simulation for existing current gradient update methods Figure 1 ;

[0061] Figure 7b Current gradient simulation for existing current gradient update methods Figure 2 ;

[0062] Figure 7c Current gradient simulation for existing current gradient update methods Figure 3 ;

[0063] Figure 7d Current gradient simulation for existing current gradient update methods Figure 4 ;

[0064] Figure 7e Simulation of a current gradient update method according to an embodiment of the present invention Figure 1 ;

[0065] Figure 7f Simulation of a current gradient update method according to an embodiment of the present invention Figure 2 ;

[0066] Figure 8a This is the current gradient map of the vector data-driven predictive current control with injected noise in Embodiment 3 of the present invention;

[0067] Figure 8b This is the current gradient map of the three-vector data-driven predictive current control based on the extended state observer of the present invention;

[0068] Figure 9a Experimental update of current gradient at 10A using existing technology Figure 1 ;

[0069] Figure 9b Experimental update of current gradient at 10A using existing technology Figure 2 ;

[0070] Figure 9cThis is an experimental diagram of current gradient update for vector data-driven predictive current control according to Embodiment 3 of the present invention.

[0071] Figure 9d This is an experimental diagram of current gradient update for three-vector data-driven predictive current control based on an extended state observer, according to an embodiment of the present invention.

[0072] Figure 10a This is an experimental diagram of three-vector data-driven predictive current control when the parameters of the filter inductor controller are not matched with the actual parameters in the existing technology.

[0073] Figure 10b This is an experimental diagram of three-vector data-driven predictive current control based on an extended state observer when the parameters of the filter inductor controller are not matched with the actual parameters according to an embodiment of the present invention.

[0074] Figure 11a This is an experimental diagram of three-vector data-driven predictive current control in the prior art when the reference current changes from 5A to 10A;

[0075] Figure 11b This is an experimental diagram of three-vector data-driven predictive current control when the reference current changes from 5A to 10A according to an embodiment of the present invention.

[0076] Figure 11c This is an experimental diagram of three-vector data-driven predictive current control based on an extended state observer when the reference current changes from 5A to 10A according to an embodiment of the present invention. Detailed Implementation

[0077] The improved data-driven energy storage inverter current prediction control method and system of the present invention will be described in more detail below with reference to the schematic diagrams, which illustrate preferred embodiments of the invention. It should be understood that those skilled in the art can modify the invention described herein while still achieving the advantageous effects of the invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the invention.

[0078] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description and claims. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0079] This invention discloses an improved data-driven current prediction control method for energy storage inverters, comprising the following steps, please refer to... Figure 1 As shown:

[0080] S1. Establish a mathematical model of the energy storage inverter in a static coordinate system, and obtain multiple basic voltage vectors based on the three-phase switching states of the energy storage inverter. Use the current gradient as a prediction model for the energy storage inverter and synthesize multiple virtual vectors. Please refer to [reference needed for more information]. Figure 3 As shown.

[0081] S2, to estimate and compensate for the measurement noise of the energy storage inverter when measuring the grid current.

[0082] S3. Establish the current gradient relationship corresponding to the multiple basic voltage vectors, and update the current gradients corresponding to the other remaining vectors. Please refer to the schematic diagram of the current gradient synthesis method. Figure 5 As shown in the diagram, please refer to the real-time update illustration of the current gradient. Figure 6 As shown.

[0083] S4, perform initial current prediction, and obtain multiple value function values ​​corresponding to the multiple basic voltage vectors and the action time of the multiple virtual vectors through value function equation analysis.

[0084] S5. Based on the updated current gradient and the application time of the multiple virtual vectors, the predicted current values ​​corresponding to the multiple virtual vectors are analyzed and obtained. The virtual voltage vector with the smallest value function value is obtained through the value function and applied as the optimal voltage vector to the next control cycle.

[0085] In a specific example, the plurality of basic voltage vectors are eight basic voltage vectors, namely u0(0,0,0), u1(1,0,0), u2(1,1,0), u3(0,1,0), u4(0,1,1), u5(0,0,1), u6(1,0,1), and u7(1,1,1). Please refer to the relevant schematic diagram. Figure 3 As shown.

[0086] The mathematical model of the energy storage inverter in the static coordinate system is as follows:

[0087]

[0088] Among them, u x i is the output voltage of the energy storage inverter. g e is the output current of the energy storage inverter. g t represents the grid-side voltage of the energy storage inverter, L represents the filter inductance, R represents the filter resistance, and t represents time.

[0089] The multiple virtual vectors refer to six virtual vectors. Please refer to [the relevant documentation]. Figure 3 As shown.

[0090] Further, please refer to Figure 4As shown, in S2, an extended state observer is established to estimate and compensate for the noise in the measured grid current. The extended state observer can measure current error feedback. The extended state observer is:

[0091]

[0092] Analysis of the extended state observer yields:

[0093]

[0094] Among them, i e The estimated value of the output current of the energy storage inverter is Δi. e The measured current gradient is the estimated value, δ1 is the feedback error gain of the sampled current, δ2 is the feedback error gain of the sampled grid current, and T is the estimated value of the actual measured current gradient. s To control the cycle, i g i is the output current of the energy storage inverter. e – i g Let t be the current error and t be the time.

[0095] For example, please refer to Figure 5 and Figure 6 As shown, the current gradient update formula can be expressed as:

[0096]

[0097]

[0098]

[0099]

[0100] Where, △i m For the current gradient of the fundamental voltage vector in the first application, t m The duration of the basic voltage vector in the first application is given by t, where A is the scaling factor of the basic voltage vector in the first application. n Let B be the duration of the basic voltage vector for the second application, and let Δi be the scaling factor for the basic voltage vector for the second application. n For the second application, the current gradient of the fundamental voltage vector, Δi z The current gradient is the zero voltage vector, Δi x Let Δi be the current gradient of the sampled current. y Let Δi1 be the current gradient of the remaining voltage vector, Δi4 be the current gradient of the first basic voltage vector among the plurality of basic voltage vectors, k-1 be the current gradient of the fourth basic voltage vector among the plurality of basic voltage vectors, and T be the current gradient of the last control time.s For the control period, t1 is the duration of the first basic voltage vector, k-2 is the previous control time, and Δi 2,6 t represents the current gradient of the second and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z t4 represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, t4 represents the duration of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, and Δi represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors. 3,5 The current gradient Δi is the current gradient of the third and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. 1,4 The current gradient is the second and fourth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0101] In one embodiment of the present invention, in S4, the plurality of basic voltage vectors are added to the sampled current to perform current prediction, and the current prediction formula is:

[0102]

[0103] Among them, i iαβ Let i be the grid current in the stationary coordinate system, k+1 be the next control time, and i αβ Let Δi be the sampling current in the stationary coordinate system. iαβ Let be the current gradient corresponding to the i-th basic voltage vector, and k be the corresponding time.

[0104] The value function is:

[0105]

[0106] Where G is the value function value, i refα Let i be the reference current of the energy storage inverter in the α stationary coordinate system. gα Let i be the grid current of the energy storage inverter in the α stationary coordinate system, k+1 be the next control time, and i refβ Let i be the reference current of the energy storage inverter in the β stationary coordinate system. gβ The grid current of the energy storage inverter in the β stationary coordinate system is given.

[0107] Furthermore, in S5, the value obtained from the current prediction is substituted into the value function to obtain the corresponding value function value.

[0108] The duration of action of the multiple virtual vectors is:

[0109]

[0110] Among them, t m The fundamental voltage vector u m The duration of action, t n The fundamental voltage vector u n The duration of action, t z G represents the duration of action of the first and eighth fundamental voltage vectors of the fundamental voltage vectors. n For u n The value function value, G z For u z The value function value, G m For u m The value function value, T s To control the cycle.

[0111] Specifically, when three basic voltage vectors are applied in each control cycle, the predicted current value at time k+1 is:

[0112]

[0113] Among them, i g t is the output current of the energy storage inverter. m The fundamental voltage vector u m The duration of action, k+1 is the next control moment, k is the current control moment, T s To control the period, △i m Let t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n The fundamental voltage vector u m The duration of action, △i n Let be the current gradient of the nth basic voltage vector among the plurality of basic voltage vectors, k be the current control time, and t be the current gradient of the nth basic voltage vector. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0114] The predicted current value at time k+2 is:

[0115]

[0116] Among them, i g The output current of the energy storage inverter is k, k+2 is the next control time, k+1 is the next control time, and t is the next control time. m The duration of the first application of the fundamental voltage vector, T. s To control the period, △i mLet t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n For the second application of the basic voltage vector, the duration of action of the basic voltage vector is Δi. n Let t be the current gradient of the nth fundamental voltage vector among the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0117] The virtual voltage vector with the minimum value function is obtained by using the value function and the current prediction value, and the virtual voltage vector is used as the optimal voltage vector in the next control cycle.

[0118] In another aspect, the invention also includes an improved data-driven energy storage inverter current prediction control system, the data-driven energy storage inverter current prediction control system comprising:

[0119] The parameter setting module is used to establish a mathematical model of the energy storage inverter in a static coordinate system, obtain multiple basic voltage vectors based on the state of the three-phase switching transistors of the energy storage inverter, and use the current gradient as a prediction model of the energy storage inverter to synthesize multiple virtual vectors.

[0120] The noise estimation module is used to estimate and compensate for the measurement noise of the grid current measured by the energy storage inverter.

[0121] The current gradient update module is used to establish the current gradient relationship with the multiple basic voltage vectors and update the current gradient corresponding to the other remaining vectors.

[0122] The current prediction module is used to obtain multiple value function values ​​corresponding to the multiple basic voltage vectors through value function equation analysis, and to obtain the action time of the multiple virtual vectors. Based on the updated current gradient and the action time of the multiple virtual vectors, it analyzes and obtains the current prediction value corresponding to the multiple virtual vectors, and obtains the virtual voltage vector with the minimum value function value through the value function, and applies the virtual voltage vector as the optimal voltage vector to the next control cycle.

[0123] In a specific embodiment of the present invention, the data-driven energy storage inverter current prediction control method is performed according to the following steps, please refer to [reference needed]. Figure 1 As shown:

[0124] Step 1, according to... Figure 2The topology of the energy storage inverter shown is used to establish a mathematical model of the inverter in a stationary coordinate system. Based on the states of the three-phase switches of the inverter, eight basic voltage vectors are obtained: u0(0,0,0), u1(1,0,0), u2(1,1,0), u3(0,1,0), u4(0,1,1), u5(0,0,1), u6(1,0,1), and u7(1,1,1). Using the current gradient as a prediction model for the inverter, six virtual vectors are synthesized. Please refer to [reference needed]. Figure 3 As shown.

[0125] Specifically, while existing data-driven predictive current control methods can use the measured current gradient as a predictive model for the energy storage inverter, thus eliminating the influence of model parameters, the output current performance of these methods is affected by the number of applied vectors, current gradient update stagnation, and sampling disturbances. To address these issues, this embodiment synthesizes a predictive current control method based on the eight basic voltage vectors of the energy storage inverter, as shown below. Figure 3 The virtual voltage vector shown.

[0126] The sampled three-phase grid current i abc Perform the Clarke transform to obtain i αβ :

[0127]

[0128] Among them, i α Let i be the grid current in the α stationary coordinate system. β Let i be the grid current in the β stationary coordinate system. a Let i be the current in phase a of the power grid. b Let i be the phase b grid current. c This represents the current in phase c of the power grid.

[0129] According to such Figure 2 A mathematical model of the energy storage inverter topology shown is established, and the mathematical model of the two-level energy storage inverter in the stationary coordinate system is obtained as follows:

[0130]

[0131] Among them, u x = [u xα u xβ ] T i is the output voltage of the energy storage inverter. g = [i gα i gβ ] T e is the output current of the energy storage inverter. g = [e gα e gβ ]T R is the grid-side voltage of the energy storage inverter, L is the filter inductance, and R is the parasitic resistance.

[0132] According to the forward Euler method, the prediction model for the energy storage inverter at time k+1 is as follows:

[0133]

[0134] Among them, i g Let T be the output current of the energy storage inverter, k+1 be the next control time, k be the current control time, and T be the current control time. s To control the cycle, L is the filter inductance, R is the parasitic resistance, and Δi x (k) is the vector u x The current gradient under the action of (k), u x e is the output voltage of the energy storage inverter. g This refers to the grid-side voltage of the energy storage inverter.

[0135] To eliminate the influence of model parameters on the predicted current, the current gradient can be expressed as follows, based on the current measurement table:

[0136]

[0137] Where, △i x (k-1) represents the current gradient at the previous time step, k-1 represents the previous control time step, and i g is the output current of the energy storage inverter, and k is the current control time.

[0138] As can be seen from the above formula, the current gradient obtained at time k should correspond to the vector applied from time k-1 to time k. The obtained result is stored in a lookup table for current prediction in subsequent steps.

[0139] In summary, the predicted current based on the measured current gradient can be expressed as:

[0140]

[0141] Among them, i g Let $k$ be the output current of the energy storage inverter, $k+1$ be the current control time, and $Δi$ be the next control time. x (k) represents the current gradient at the current moment.

[0142] Step 2: Apply the extended state observer. Please refer to [reference needed for the extended state observer]. Figure 4 As shown, the measurement noise of the grid current is estimated and compensated to reduce the impact of measurement disturbances on the prediction results and ensure the stability of the output current performance; the extended state observer with measurement current error feedback is represented as follows:

[0143]

[0144] Among them, i e For i g The estimated value, △i e For △i s The estimated value, i e – i g δ1 is the feedback error gain of the sampled current, and δ2 is the feedback error gain of the sampled grid current.

[0145] The above formula can be discretized as:

[0146]

[0147] Among them, i e (k-1) represents the estimated value of the grid current at the previous moment, i g (k-1) represents the sampled value of the grid current at the previous moment, and err(k-1) represents the error between the estimated grid current and the sampled value at the previous moment; i e (k) is the estimated value of the grid current at the current moment, Δi e (k-1) is the estimated value of the current gradient at the previous moment; Δi e (k) represents the estimated value of the current gradient at the current moment; δ1 is the feedback error gain of the sampled current, and δ2 is the feedback error gain of the sampled grid current; T s To control the cycle.

[0148] To obtain the error feedback gain, the matrix form of the extended state observer can be expressed as:

[0149]

[0150] Among them, i e (k-1) represents the estimated value of the grid current at the previous moment, i g (k-1) represents the sampled value of the grid current at the previous moment, and err(k-1) represents the error between the estimated grid current and the sampled value at the previous moment; i e (k) is the estimated value of the grid current at the current moment, Δi e (k-1) is the estimated value of the current gradient at the previous moment; Δi e (k) represents the estimated value of the current gradient at the current moment; δ1 is the feedback error gain of the sampled current, and δ2 is the feedback error gain of the sampled grid current; T s To control the cycle.

[0151] The characteristic polynomial of the above matrix can be expressed as:

[0152]

[0153] Where z is the observer state vector and I is the second-order identity matrix.

[0154] To obtain the fastest dynamic performance, both eigenvalues ​​of the above characteristic polynomial are chosen to be zero to achieve the beatless characteristic, i.e.:

[0155]

[0156] Analyzing and calculating the above characteristic polynomial, we obtain:

[0157]

[0158] Where z is the observer state vector, G is the value function, δ1 is the feedback error gain of the sampled current, and δ2 is the feedback error gain of the sampled grid current; T s To control the cycle.

[0159] Step 3, please refer to Figure 5 and Figure 6 As shown, the current gradient relationship corresponding to different vectors is established, and the current gradient corresponding to the remaining vectors is updated according to the current gradient relationship corresponding to different vectors.

[0160] The virtual vector applied between time k-1 and time k consists of three basic vectors u m u n and u z During the configuration, the measured current gradient Δi x (k-1) by △i m (k-1), △i n (k-1) and △i z Composed of (k-1).

[0161] The current gradient can then be expressed as:

[0162]

[0163] Where, △i x (k-1) represents the current gradient at the previous time step, i g Let t be the output current of the energy storage inverter, k be the current control time, k-1 be the previous control time, and t be the current output current of the energy storage inverter. m The duration of the first application of the fundamental voltage vector, T. s To control the period, △i m Let t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n For the second application of the basic voltage vector, the duration of action of the basic voltage vector is Δi. n Let t be the current gradient of the nth fundamental voltage vector among the plurality of fundamental voltage vectors. zFor the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0164] t z / 2 represents the duration of action of u0 or u7, u m and u n As the fundamental vector, t m and t n u m and u n The duration of action.

[0165] Current gradient Δi m (k-1), △i n (k-1) and △i z (k-1) are all unknowns, where the current gradient Δi x (k-1) can be considered as a sum of [u] x (k-1) - Ri g (k-1) - e g (k-1)]T s / L are equal.

[0166] Select voltage vector u m u n and u z The current gradient equation at time k-1 is obtained as follows:

[0167]

[0168] Similarly, at time k-2, Δi m (k-2), △i n (k-2) and △i z The current gradient equation between (k-2) can be expressed as:

[0169]

[0170] Through calculation, the new circuit gradient equation is obtained:

[0171]

[0172] in,

[0173]

[0174] Furthermore, through analysis and calculation, the current gradient Δi is obtained. m (k-1), △i n (k-1) and △i z (k-1) is:

[0175]

[0176] Where L is the filter inductance, Δi m For the current gradient of the fundamental voltage vector in the first application, t m The duration of the basic voltage vector in the first application is given by t, where A is the scaling factor of the basic voltage vector in the first application. n Let B be the duration of the basic voltage vector for the second application, and let Δi be the scaling factor for the basic voltage vector for the second application. n For the second application, the current gradient of the fundamental voltage vector, Δi x Let Δi be the current gradient of the sampled current. y Let Δi1 be the current gradient of the remaining voltage vector, Δi4 be the current gradient of the first basic voltage vector among the plurality of basic voltage vectors, k-1 be the current gradient of the fourth basic voltage vector among the plurality of basic voltage vectors, and T be the current gradient of the last control time. s For the control period, t1 is the duration of the first basic voltage vector, k-2 is the previous control time, and Δi 2,6 t represents the current gradient of the second and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z t4 represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, t4 represents the duration of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, and Δi represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors. 3,5 The current gradient Δi is the current gradient of the third and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. 1,4 The current gradient is the second and fourth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0177] Through the above steps, the current gradient values ​​under the action of the three virtual voltage vectors applied in the previous control cycle are updated.

[0178] Similarly, the current gradient equations for the residual vector and the zero vector are established to update the remaining four residual vectors u. y The current gradient value Δi of (k-1) y (k-1). The update equation for the residual current gradient is derived as follows:

[0179]

[0180] Where C is the scaling factor of the residual vector, Δi y The current gradient of the residual voltage vector, Δiz The current gradients of the first and eighth basic voltage vectors of the plurality of basic voltage vectors are given, where k-1 is the previous control time and k-2 is the control time before that; and

[0181] .

[0182] Where C is the scaling factor of the residual vector, u y (k-1) represents the coordinate components of the remaining vector in the previous control cycle, u y (k-2) represents the coordinate components of the remaining vector in the previous control cycle, u z (k-1) represents the coordinate components of the zero vector in the previous control cycle, u z (k-2) represents the coordinate component of the zero vector in the previous control cycle.

[0183] When the denominators of A, B, and C in the above current gradient formula are not zero, since the DC voltage remains constant within one control cycle, the values ​​of A, B, and C are approximately equal to 1.

[0184] Therefore, △i m (k-1), △i n (k-1), △i z (k-1) and △i y The value of (k-1) can be expressed as:

[0185]

[0186]

[0187] Where, △i m For the first application, the current gradient of the fundamental voltage vector, Δi n For the second application, the current gradient of the fundamental voltage vector, Δi z The current gradient Δi is the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors. x Let Δi be the current gradient of the sampled current. y The current gradient of the residual voltage vector, t m For the duration of action of the fundamental voltage vector in the first application, t n For the duration of the fundamental voltage vector in the second application, T s The control period is defined as k-1, where k-1 is the previous control time and k-2 is the control time before that.

[0188] However, when the denominators A, B, and C in the above current gradient formula approach 0, the above update equation becomes invalid.

[0189] At this point, the equation relationship between △i4(k-1) and △i3(k-1) and △i5(k-1) is established, thereby realizing the update of △i4(k-1).

[0190] When Δi1(k-1) and Δi4(k-1) are updated as the current gradients applied as vectors, the update formula can be expressed as:

[0191]

[0192] When Δi1(k-1) and Δi4(k-1) are updated as the residual current gradients of the residual vector, the update formula can be expressed as:

[0193] .

[0194] Where, △i m For the current gradient of the fundamental voltage vector in the first application, t m The duration of the basic voltage vector in the first application is given by t, where A is the scaling factor of the basic voltage vector in the first application. n Let B be the duration of the basic voltage vector for the second application, and let Δi be the scaling factor for the basic voltage vector for the second application. n For the second application, the current gradient of the fundamental voltage vector, Δi z The current gradient is the zero voltage vector, Δi x Let Δi be the current gradient of the sampled current. y Let Δi1 be the current gradient of the remaining voltage vector, Δi4 be the current gradient of the first basic voltage vector among the plurality of basic voltage vectors, k-1 be the current gradient of the fourth basic voltage vector among the plurality of basic voltage vectors, and T be the current gradient of the last control time. s For the control period, t1 is the duration of the first basic voltage vector, k-2 is the previous control time, and Δi 2,6 t represents the current gradient of the second and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z t4 represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, t4 represents the duration of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, and Δi represents the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors. 3,5 The current gradient Δi is the current gradient of the third and sixth fundamental voltage vectors of the plurality of fundamental voltage vectors. 1,4 The current gradient is the second and fourth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0195] In summary, the above method enables real-time updating of all current gradients within each control cycle, completely eliminating the stagnation phenomenon in current gradient updates.

[0196] Step 4: Add the current gradients corresponding to the eight basic voltage vectors to the sampled currents to predict the current, and substitute the eight predicted currents into the value function equation to obtain the eight value function values. Since the vector action time is inversely proportional to the value function value, the action time of each vector in the virtual vector can be calculated.

[0197] After updating all current gradients, the vector action time for the next control cycle is calculated.

[0198] First, current prediction is performed based on the updated current gradient for each voltage vector:

[0199]

[0200] Among them, i iαβ Let i be the grid current in the stationary coordinate system, k+1 be the next control time, and i αβ Let Δi be the sampling current in the stationary coordinate system. iαβ Let Δi be the current gradient corresponding to the i-th fundamental voltage vector, and k be the corresponding time. iαβ (k) represents the current gradient corresponding to the i-th fundamental voltage vector, i = 1, 2, ..., 8, i iαβ (k+1) represents the predicted current corresponding to the i-th basic voltage vector.

[0201] The duration of action of each vector can be expressed as:

[0202]

[0203] Among them, t m For the duration of action of the fundamental voltage vector in the first application, t n For the duration of action of the fundamental voltage vector in the second application, t z G represents the duration of action of the first and eighth fundamental voltage vectors of the fundamental voltage vectors. n For the value function of the basic voltage vector in the second application, G z G is the value function value of the zero voltage vector. m For the value function of the basic voltage vector in the first application, T s For control period; G m G n and G z u m u n and u z The value function value; u zThis represents u0 and u7.

[0204] The value function is:

[0205]

[0206] Where G is the value function value, i refα Let i be the reference current of the energy storage inverter in the α stationary coordinate system. gα Let i be the grid current of the energy storage inverter in the α stationary coordinate system, k+1 be the next control time, and i refβ Let i be the reference current of the energy storage inverter in the β stationary coordinate system. gβ The grid current of the energy storage inverter in the β stationary coordinate system is given.

[0207] Step 5: Based on the updated current gradient and time obtained in the above steps, the predicted current values ​​corresponding to the six virtual vectors are obtained through analysis and calculation. The predicted current values ​​are substituted into the value function for evaluation. The virtual voltage vector with the smallest value function value is taken as the optimal voltage vector, and the optimal voltage vector is applied to the next control cycle.

[0208] When three basic voltage vectors are applied in each control cycle, the predicted current at time k+1 is expressed as:

[0209]

[0210] To compensate for the control delay of one cycle, the predicted current at time k+2 is expressed as:

[0211]

[0212] Among them, i g t is the output current of the energy storage inverter. m The basic voltage vector is the first applied basic voltage vector, where k+2 is the next control time, k+1 is the next control time, and k is the current control time. T s To control the period, △i m Let t be the current gradient of the m-th fundamental voltage vector among the plurality of fundamental voltage vectors. n For the second application of the basic voltage vector, the duration of action of the basic voltage vector is Δi. n Let be the current gradient of the nth basic voltage vector among the plurality of basic voltage vectors, k be the current control time, and t be the current gradient of the nth basic voltage vector. z For the duration of action of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, Δi z The current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

[0213] Substitute the six predicted currents obtained from the above formula into the value function for optimization. The virtual vector with the lowest value function value is taken as the optimal vector and applied to the next control cycle. The control block diagram is as follows: Figure 6 As shown.

[0214] In a specific experiment, in order to verify the effectiveness of the improved data-driven current prediction control method for energy storage inverters of the present invention, the steady-state performance of the improved data-driven current prediction control method for energy storage inverters of the present invention under different update methods was tested and compared with the existing data-driven predictive current control method obtained from the test.

[0215] In Experiment 1, please refer to... Figures 7a to 7f As shown, an embodiment of the present invention provides a three-vector data-driven predictive current control method for energy storage inverters that outperforms existing technologies.

[0216] In Experiment 2, please refer to... Figures 8a to 8b As shown, an embodiment of the present invention provides a three-vector data-driven predictive current control method for energy storage inverters that outperforms existing technologies.

[0217] In Experiments 3 through 5, Matlab / Simulink simulations were used to verify the effectiveness of the proposed strategy. The relevant parameters used are shown in the table below:

[0218]

[0219] In Experiment 3, please refer to... Figure 9a As shown, the existing update method results in a prolonged stagnation, leading to current spikes in the output current ripple. Please refer to... Figure 9b As shown, while the updated method in the prior art effectively improved the stagnation phenomenon, reducing the total harmonic distortion from 2.77% to 2.42%, it still resulted in current spikes in the output current ripple because it could not completely eliminate the stagnation. Please refer to... Figure 9c As shown, in one embodiment of the three-vector data-driven predictive current control method of the present invention, the total harmonic distortion is reduced to 2.20%; further, please refer to... Figure 9d As shown, in one embodiment of the three-vector data-driven predictive current control method based on an extended state observer, the total harmonic distortion is further reduced to 1.71%, eliminating stagnation and suppressing prediction errors caused by sampling disturbances. In summary, the update method and sampling disturbance suppression method of the three-vector data-driven predictive current control method for energy storage inverters of this invention have superior effects compared to existing technologies.

[0220] In Experiment 4, please refer to... Figures 10a-10bAs shown, the reference current is set to 10A. The filter inductor controller parameters are not matched with the actual parameters. For details, please refer to [link / reference]. Figure 10a As shown, when applying the existing three-vector data-driven predictive current control method, the current ripple and prediction error increase with the increase of inductor parameter error. Specifically, refer to... Figure 10a It can be seen that the current prediction error is largest when L0 is close to 0.2L, and the current ripple becomes distorted at this point. Please refer to [reference needed]. Figure 10b As shown, compared with the existing three-vector data-driven predictive current control method, the current quality of the three-vector data-driven predictive current control method based on the extended state sensor of the present invention is not affected by parameter mismatch. In summary, compared with the existing data-driven predictive current control methods, the three-vector data-driven predictive current control method for energy storage inverters of the present invention has better parameter robustness.

[0221] In Experiment 5, the reference current was changed from 5A to 10A. Please refer to [the relevant documentation / reference]. Figures 11a-11c As shown, Figure 11a This paper presents the three-phase current waveform, dynamic response, and prediction error of the existing three-vector data-driven predictive current control method. Figure 11b The three-phase current waveform, dynamic response, and prediction error of the three-vector data-driven predictive current control method of the present invention are shown below. Figure 11cThe three-phase current waveform, dynamic response, and prediction error of the three-vector data-driven predictive current control method based on an extended state observer of the present invention are shown. It can be seen that the total harmonic distortion (THD) of the a-phase current in the prior art three-vector data-driven predictive current control method is reduced from 3.85% to 1.91%. The current performance of the three-vector data-driven predictive current control method of the present invention is affected by sampling disturbances, and the THD is slightly higher than that of the prior art three-vector data-driven predictive current control method, specifically, it is reduced from 4.42% to 2.20%. The current ripple of the three-vector data-driven predictive current control method based on an extended state observer of the present invention is improved, and the THD is reduced from 3.31% to 1.71%. In summary, the three-vector data-driven predictive current control method based on an extended state observer of the present invention can effectively suppress sampling disturbances and does not suffer from the disadvantage of being affected by system nonlinear factors in the prior art three-vector data-driven predictive current control method. Both the current ripple and prediction error are smaller, making it superior to the prior art three-vector data-driven predictive current control method. Furthermore, the error gain feedback in the extended state observer of the three-vector data-driven predictive current control method based on the extended state observer in this invention aims for the fastest dynamic response, with a response speed of 1.32 ms. In summary, the three-vector data-driven predictive current control method based on the extended state observer in this invention is superior to the existing three-vector model-free predictive current control method.

[0222] In the actual experiment, the experimental parameters were consistent with the simulation parameters.

[0223] In summary, the improved data-driven current prediction control method for energy storage inverters of this invention outperforms existing data-driven predictive current control methods for energy storage inverters in both simulation and actual experiments.

[0224] In summary, this invention provides an improved data-driven current prediction control method for energy storage inverters. Based on the coordinate relationship of the voltage vector, a model is established, and the current gradient is updated in real time and rapidly using the current gradient update formula. This solves the problem of stagnant current gradient updates in model-less predictive current control of energy storage inverters, eliminates current spikes in the output current of energy storage inverters, and reduces prediction errors.

[0225] Taking into account the measurement noise of the grid current in the model-free predictive current control of the energy storage inverter, the measurement noise of the grid current was estimated and compensated. Under the condition of fast dynamic response speed, the problem of the model-free predictive current control of the energy storage inverter being easily affected by measurement noise was solved, and the problem of large current prediction error was further solved.

[0226] Applying three vectors in each control cycle improves the robustness to parameter changes. Without increasing the lookup table, the voltage vector applied in each control cycle is increased, reducing the impact of sampling disturbances and output current ripple, and improving the quality of the output current.

[0227] Furthermore, the current prediction of the data-driven energy storage inverter current prediction control method is based on the current gradient stored in the lookup table. It realizes the real-time update of all current gradients in the lookup table without any system parameters, and can still maintain good current performance when parameters are mismatched.

[0228] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An improved data-driven energy storage inverter current predictive control method, characterized in that, The method comprises the following steps: S1, establishing a mathematical model of the energy storage inverter in a static coordinate system, and obtaining a plurality of basic voltage vectors according to a three-phase switch tube state of the energy storage inverter, using a current gradient as a prediction model of the energy storage inverter and synthesizing a plurality of virtual vectors; S2, estimating and compensating for measurement noise of a grid current measured by the energy storage inverter; S3, establishing a current gradient relationship corresponding to the plurality of basic voltage vectors, and updating a current gradient corresponding to a remaining vector; S4, performing initial current prediction, and analyzing to obtain a plurality of value function values corresponding to the plurality of basic voltage vectors and an action time of the plurality of virtual vectors through a value function equation; S5, according to the updated current gradient and the action time of the plurality of virtual vectors, analyzing to obtain a current prediction value corresponding to the plurality of virtual vectors, and obtaining a virtual voltage vector with a minimum value function value through the value function, and taking the virtual voltage vector as an optimal voltage vector acting on a next control period.

2. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, The plurality of basic voltage vectors are eight basic voltage vectors, respectively u0(0, 0, 0), u1(1, 0, 0), u2(1, 1, 0), u3(0, 1, 0), u4(0, 1, 1), u5(0, 0, 1), u6(1, 0, 1), and u7(1, 1, 1); The mathematical model of the energy storage inverter in the static coordinate system is: wherein u x is the output voltage of the energy storage inverter, i g is the output current of the energy storage inverter, e g is the grid-side voltage of the energy storage inverter, L is a filter inductance, R is a filter resistance, and t is time. The plurality of virtual vectors are six virtual vectors.

3. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, In S2, an extended state observer is established to estimate and compensate for the noise of the measured grid current, the extended state observer can measure current error feedback, and the extended state observer is: The extended state observer is analyzed to obtain: where i e is an estimate of the output current of the energy storage inverter, Δi e is an estimate of the actual measured current gradient, δ1 is a feedback error gain of the sampled current, δ2 is a feedback error gain of the sampled grid current, T s is a control period, i g is the output current of the energy storage inverter, i e – i g is a current error, t is time.

4. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, The current gradient update formula is expressed as: ， wherein, △i m is the current gradient of the first applied fundamental voltage vector, t m is the action time of the first applied fundamental voltage vector, A is the proportional coefficient of the first applied fundamental voltage vector, t n is the action time of the second applied fundamental voltage vector, B is the proportional coefficient of the second applied fundamental voltage vector, △i n is the current gradient of the second applied fundamental voltage vector, △i z is the current gradient of the zero voltage vector, △i x is the current gradient of the sampling current, △i y is the current gradient of the residual voltage vector, △i1 is the current gradient of the first fundamental voltage vector of the plurality of fundamental voltage vectors, △i4 is the current gradient of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, k-1 is the last control time, T s is the control period, t1 is the action time of the first fundamental voltage vector of the fundamental voltage vectors, k-2 is the second last control time, △i 2,6 is the current gradient of the second fundamental voltage vector and the sixth fundamental voltage vector of the plurality of fundamental voltage vectors, t z is the action time of the first fundamental voltage vector and the eighth fundamental voltage vector of the plurality of fundamental voltage vectors, △i z is the current gradient of the first fundamental voltage vector and the eighth fundamental voltage vector of the plurality of fundamental voltage vectors, △i x is the current gradient of the sampling measurement, t4 is the action time of the fourth fundamental voltage vector of the plurality of fundamental voltage vectors, △i 3,5 is the current gradient of the third fundamental voltage vector and the sixth fundamental voltage vector of the plurality of fundamental voltage vectors, △i 1,4 is the current gradient of the second fundamental voltage vector and the fourth fundamental voltage vector of the plurality of fundamental voltage vectors.

5. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, In S4, the plurality of basic voltage vectors are added to the sampling current for current prediction, and the current prediction formula is: wherein, i iαβ is the grid current in the stationary coordinate system, k+1 is the next control time, i αβ is the sampled current in the stationary coordinate system, △i iαβ is the current gradient corresponding to the i-th basic voltage vector, k is the corresponding time; The value function is: wherein G is a value function value, i refα is the reference current of the energy storage inverter in the α stationary coordinate system, i gα is the grid current of the energy storage inverter in the α stationary coordinate system, k+1 is the next control time, i refβ is the reference current of the energy storage inverter in the β stationary coordinate system, i gβ is the grid current of the energy storage inverter in the β stationary coordinate system.

6. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, In S5, the value obtained by the current prediction is substituted into the value function to obtain a corresponding value function value; The action time of the plurality of virtual vectors is: wherein t m is the action time of the first applied fundamental voltage vector, t n is the action time of the second applied fundamental voltage vector, t z is the action time of the first and eighth fundamental voltage vectors of the fundamental voltage vector, G n is the value function value of the second applied fundamental voltage vector, G z is the value function value of the zero voltage vector, G m is the value function value of the first applied fundamental voltage vector, T s is the control period.

7. The improved data-driven energy storage inverter current predictive control method of claim 1, wherein, When three basic voltage vectors are applied in each control period, the current prediction value at k+1 time is: wherein, i g is the output current of the energy storage inverter, t m is the action time of the fundamental voltage vector first applied, k+1 is the next control time, k is the current control time, T s is the control period, △i m is the current gradient of the mth fundamental voltage vector of the plurality of fundamental voltage vectors, t n is the action time of the fundamental voltage vector second applied, △i n is the current gradient of the nth fundamental voltage vector of the plurality of fundamental voltage vectors, k is the current control time, t z is the action time of the first fundamental voltage vector and the eighth fundamental voltage vector of the plurality of fundamental voltage vectors, △i z is the current gradient of the first fundamental voltage vector and the eighth fundamental voltage vector of the plurality of fundamental voltage vectors; The current prediction value at k+2 time is: wherein, i g is the output current of the energy storage inverter, k+2 is the next control time, k+1 is the next control time, t m is the action time of the first applied fundamental voltage vector, T s is the control period, △i m is the current gradient of the mth fundamental voltage vector of the plurality of fundamental voltage vectors, t n is the action time of the second applied fundamental voltage vector, △i n is the current gradient of the nth fundamental voltage vector of the plurality of fundamental voltage vectors, t z is the action time of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors, △i z is the current gradient of the first and eighth fundamental voltage vectors of the plurality of fundamental voltage vectors.

8. An improved data-driven energy storage inverter current predictive control system, characterized by, The method comprises: A parameter setting module is configured to establish a mathematical model of the energy storage inverter in a static coordinate system, obtain a plurality of basic voltage vectors according to a three-phase switch tube state of the energy storage inverter, and synthesize a plurality of virtual vectors by taking a current gradient as a prediction model of the energy storage inverter; A noise estimation module is configured to estimate and compensate for measurement noise of a grid current measured by the energy storage inverter; A current gradient update module is configured to establish a current gradient relationship corresponding to the plurality of basic voltage vectors, and update a current gradient corresponding to a remaining vector; The current prediction module is configured to analyze the plurality of value function values corresponding to the plurality of basic voltage vectors by a value function equation, and obtain the action time of the plurality of virtual vectors; analyze the current prediction values corresponding to the plurality of virtual vectors according to the updated current gradient and the action time of the plurality of virtual vectors, and obtain the virtual voltage vector with the minimum value function value by the value function; and the virtual voltage vector is used as the optimal voltage vector to act on the next control period.

Images