Data-driven optimization method for grid-connected inverter grid-connected passive control parameters

Abstract

The present invention discloses a data-driven method for optimizing the design of grid-connected inverter grid-type passive control parameters, and belongs to the field of electrical engineering. In view of the fact that the passive controller has a complex structure and it is difficult to find a suitable damping gain in the grid-connected control of the converter based on the passivity theory, the present invention proposes a data-driven method for optimizing the grid-connected inverter grid-type passive control parameters. The method improves the grid adaptability and stability of the grid-connected inverter by online adjusting the passive controller parameters using a particle swarm algorithm. The present invention is not only simple to implement, but also realizes real-time observation of the impact of different parameters on system stability, ensuring that the grid-connected inverter system under complex grid conditions can avoid falling into local optimality and realizes the robustness of passive control parameter design.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering, and in particular to a data-driven method for optimizing grid-connected inverter grid-type passive control parameters. Background Art

[0002] The proportion of grid-connected inverters installed in power systems is increasing. Their control characteristics have a profound impact on high-penetration renewable energy power generation systems, bringing huge challenges to the stable operation of grid-connected inverters. Grid-connected inverters themselves are nonlinear, especially when they are used in nonlinear application scenarios where the grid structure / parameters change with plug-and-play inverters and random switching. The adaptability of grid-connected inverters is often insufficient, which may cause stability problems such as broadband oscillations. For this reason, some studies have proposed to directly use nonlinear control methods to study the control of grid-connected inverters. For example, sliding mode control, adaptive control, model predictive control, passivity-based control and other nonlinear control technologies for grid-connected inverters.

[0003] Passivity control directly designs nonlinear controllers based on energy dissipation characteristics. This method has an important characteristic: the new system obtained by connecting multiple passive subsystems in parallel still satisfies passivity. According to the passivity theory, the obtained system can be proved to satisfy asymptotic stability based on Lyapunov theory. Therefore, the controller designed based on passivity theory provides an effective solution for the grid-connected control technology of high-penetration renewable energy power generation systems. For example:

[0004] 1) "Shipra, K., Maurya, R. and Sharma, SN (2020), Port-controlled Hamiltonian-based controller for an interleaved boost PFC converter", published in IET Transactions on Power Electronics 2020, proposes to apply the interconnection and damping distribution passive control method to the control of three-phase VSI.

[0005] 2) "M. Li, H. Geng and X. Zhang, Robust Passivity-Based Control for Grid-Forming Converter, CIEEC", published in the 2023 IEEE 6th International Electrical and Energy Conference (CIEEC). This paper uses the D-partition method to introduce nonlinear control delays, providing a robust passive control method for grid-forming converters that meets multiple performance indicators, obtains the stable region of control parameters, and is robust to system parameter changes and external disturbances.

[0006] However, passive controllers have complex structures and it is difficult to find suitable damping gains. Artificial intelligence technology has made great progress in recent years. Among them, particle swarm optimization has gradually been used to design controller parameters due to its flexibility, simplicity and ease of use. For example:

[0007] 1) "B. Ufnalski, A. Kaszewski and L. M. Grzesiak, Particle Swarm Optimization of the Multioscillatory LQR for a Three-Phase Four-Wire Voltage-Source Inverter With an LC Output Filter", published in IEEE Transactions on Industrial Electronics in 2015, uses the PSO algorithm to optimize the system and searches for the quadratic cost function of the optimal weighting factor based on the PSO algorithm, making the whole process less complicated.

[0008] 2) Published in the 2022 IEEE 6th International Conference on Information Technology and Mechatronics Engineering (ITOEC), (“J. Guo, Y. Lu and Z. Li, PID parameter tuning algorithm of rotor UAV Based on Improved Particle Swarm Optimization”, 2022, pp. 1251-1255) (“PID parameter tuning algorithm of rotor UAV based on improved particle swarm optimization”) This paper uses the improved particle swarm optimization method to adjust the parameters so that the system can achieve better control performance indicators, and the adjustment time is short, so that the UAV has better dynamic performance. At the same time, it also saves manual time and reduces the time for parameter tuning, which has certain guiding significance for practical engineering applications.

[0009] In summary, the prior art has the following problems:

[0010] (1) Although the damping gain designed by the methods involved in the existing literature can provide good stability performance and ensure robustness to parameter changes, the design process is very cumbersome and requires repeated adjustment, and there are many limitations during operation.

[0011] (2) The existing literature does not involve online adjustment of the damping gain of the passive controller in the grid-connected inverter to achieve real-time acquisition of stability parameters under scenarios of impedance and grid nonlinear changes. Summary of the invention

[0012] In order to overcome the limitations of the above-mentioned technical solutions, the present invention proposes a data-driven grid-connected inverter grid-type passive control parameter optimization method, which solves the parameter setting problem of the passive controller damping gain in the grid-connected inverter. At the same time, it can obtain the stability parameters in the scenario of impedance and grid nonlinear changes in real time, and proposes an effective method for analyzing the impact of different parameters on system stability.

[0013] The object of the present invention is achieved in this way. The present invention provides a data-driven grid-connected inverter grid-type passive control parameter optimization method, wherein the grid-connected inverter is an LCL-type grid-connected inverter, comprising a DC source, an inverter, an LCL-type filter, a line equivalent inductance and a three-phase power grid connected in series in sequence, and the LCL-type filter comprises a bridge arm side inductance, a bridge arm side inductance equivalent resistance, a grid side inductance and a filter capacitor;

[0014] The optimization method designs a passive feedback controller by establishing a mathematical model of an LCL-type grid-connected inverter in a grid-connected operation mode and a Hamiltonian model with the LCL-type grid-connected inverter as a control object, and adopts a data-driven method to optimize two control parameters in the passive control law, wherein the two control parameters are a first injection damping gain r1 and a second injection damping gain r2; the specific steps are as follows:

[0015] Step 1: Establish a mathematical model of the LCL grid-connected inverter in the grid-connected operation mode, and its expression is:

[0016]

[0017] In the formula, L1 is the inductance of the bridge arm side, C is the capacitance of the filter capacitor, r f is the resistance value of the inductor equivalent resistance on the bridge arm side, i 1d ,i 1q is the dq axis component of the inductor current on the bridge arm side, ω is the rated angular frequency of the power grid, u d ,u q is the dq axis component of the inverter output voltage, u Cd ,u Cqis the dq axis component of the filter capacitor voltage, i gd ,i gq is the dq axis component of the grid-side current;

[0018] Step 2: Based on the dissipative passivity theory, a Hamiltonian model with the LCL grid-connected inverter as the control object is established, and its expression is:

[0019]

[0020] Where u is the passive feedback controller, u=[u d u q ] T ; x is the state variable of the Hamiltonian model, denoted as state variable x, x = [i 1d i 1q u Cd u Cq ] T , is the derivative of the state variable x; J is the antisymmetric interconnection matrix, R is the semi-positive definite dissipation matrix, G is the input matrix, ξ is the interference matrix; H(x) is the initial Hamiltonian energy storage function;

[0021] Substitute the mathematical model of the LCL grid-connected inverter in the grid-connected operation mode in step 1 into the Hamiltonian model to obtain the antisymmetric interconnection matrix J, the semi-positive definite dissipation matrix R, the input matrix G, and the interference matrix ξ;

[0022] Step 3, according to the Hamiltonian model established in step 2, a passive control law of a passive feedback controller is obtained based on an interconnection and damping distribution passive control method, wherein the expression of the passive control law includes a first injection damping gain r1 and a second injection damping gain r2;

[0023] Step 4: Sample the dq-axis component u of the filter capacitor voltage Cd ,u Cq , define the error e u , e u =u Cd -u Cq , select the first power evaluation function ITSE for e u Calculation is performed to obtain the fitness value in this scenario, wherein the first power evaluation function ITSE is the integral of any time t during the operation of the grid-connected inverter multiplied by the square of the error;

[0024] Use the particle swarm optimization algorithm to make the fitness value approach 0 or reach the preset number of iterations to obtain the global optimal fitness value g best , and the global optimal fitness value g best The corresponding estimated values ​​of the two control parameters are the optimal first injection damping gain r1' and the optimal second injection damping gain r2'.

[0025] Preferably, the implementation process of step 3 is as follows:

[0026] Step 3.1, define the expected equilibrium point x of the Hamiltonian model ＊ , whose expression is:

[0027]

[0028] in, represents the expected value of the dq-axis component of the inductor current on the bridge arm side, Represents the expected value of the dq-axis component of the filter capacitor voltage;

[0029] Step 3.2, define the initial Hamiltonian energy storage function H(x) as the sum of the energy stored in the inductor and filter capacitor on the bridge arm side, that is:

[0030]

[0031] Among them, x T is the transpose of the state variable x, Q is the diagonal parameter matrix, Q = diag{L1, L1, C, C};

[0032] Step 3.3, define the closed-loop Hamiltonian energy storage function H d (xx * ), whose expression is:

[0033]

[0034] Step 3.4, introduce the total structure matrix J d And the total damping matrix R d , whose expressions are:

[0035]

[0036] Step 3.5: According to step 3.4, the closed-loop system expression of the LCL grid-connected inverter based on the Hamiltonian equation is obtained:

[0037]

[0038] in, is the derivative of the expected value of the state variable of the Hamiltonian model;

[0039] Step 3.6, the expression of the passive feedback control law of the passive feedback controller u of the LCL type grid-connected inverter is:

[0040]

[0041] Where t is any moment in the operation of the grid-connected inverter.

[0042] Preferably, the global optimal fitness value g in step 4 is best The solution process is as follows:

[0043] Establish the evaluation function f o (e u ), whose expression is:

[0044]

[0045] Use the particle swarm optimization algorithm to find the estimated values ​​of the two control parameters that meet the requirements, and make the particle swarm optimization evaluation function f o (e u ) is the smallest. The specific steps are as follows:

[0046] Step 4.1, set the particle swarm size to n, particle dimension d to 2, particle swarm optimization algorithm learning factor 1 to c1, particle swarm optimization algorithm learning factor 2 to c2, particle swarm optimization algorithm initial inertia factor w1, particle swarm optimization algorithm end inertia factor w2, particle swarm optimization iteration number K, particle velocity minimum value v min , the maximum particle velocity v max ;

[0047] Step 4.2, set the speed range, position range and minimum fitness accuracy fit, and initialize the speed and position of each particle in the particle swarm optimization algorithm;

[0048] Step 4.3, calculate the fitness value of each particle of the particle swarm optimization algorithm, and record the smallest fitness value as the individual optimal fitness value p best , at this time p best =g best ;

[0049] Step 4.4, update the position and velocity of each particle in the particle swarm optimization algorithm, and perform K updates, and record any update in the K updates as the kth update, k = 1, 2, ..., K, k is a positive integer;

[0050] The velocity of the particle swarm optimization algorithm particle at the kth update is v k , position x k , the update formula is as follows:

[0051]

[0052] x k =x k-1 +v k

[0053] Wherein, rand is a random number between 0 and 1, k = 1, 2, ..., K;

[0054] Step 4.5, recalculate the fitness value of each particle in the particle swarm optimization algorithm. Denote the minimum fitness value at the k-th update as the minimum fitness value at the (k - 1)-th update as Compare with . If , then update the value of to p best . Otherwise, the value of p best remains unchanged. Then compare the value of p best with the value of g best . If p best < g best , then update the value of p best to g best . Otherwise, the value of g best remains unchanged;

[0055] Step 4.6, repeat the iterative process. When g best < fit or when the K-th update is reached, end the iteration and output the global best fitness value g best , and obtain the estimated value of the first injection damping gain r1 and the estimated value of the second injection damping gain r2 corresponding to this global best fitness value g best . Denote these two estimated values as the optimal first injection damping gain and the optimal second injection damping gain respectively.

[0056] Preferably, the expressions of the anti-symmetric interconnection matrix J, the positive semi-definite dissipation matrix R, the input matrix G, and the disturbance term matrix ξ in Step 1 are respectively:

[0057]

[0058] ξ = [-u Cd / L1 -u Cq / L1 - ωCu Cq ωCu Cd .

[0059] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0060] 1. The present invention is not only simple to implement, but also by using the artificial intelligence algorithm PSO algorithm for parameter tuning, it greatly reduces the difficulty of implementation, and at the same time can ensure good stability performance of the grid-connected inverter and robustness to parameter changes;

[0061] 2. By means of online parameter tuning, the present invention can obtain the real-time influence effect of parameter changes on system stability, and more conveniently analyze the system stability;

[0062] 3. The present invention can improve the effectiveness of parameters by improving intelligent algorithms and selecting other algorithms to ensure a more stable system;

[0063] 4. The existing scheme uses modeling analysis and other means to set parameters, which has problems such as repeated iterations and huge amount of calculation. The present invention adopts the data-driven technology of particle swarm algorithm to ensure the optimal design of parameters without the cost of manual calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a control diagram of the optimization method of the present invention.

[0065] Figure 2 It is a fitness evolution curve diagram of the optimization method of the present invention under the first power evaluation function ITSE.

[0066] Figure 3 for Figure 2 The damping gain obtained in the scenario is substituted into the waveform of the d-axis component of the inverter output voltage obtained by simulation.

[0067] Figure 4 for Figure 2 The damping gain obtained in the scenario is substituted into the waveform of the q-axis component of the inverter output voltage obtained by simulation.

[0068] Figure 5 It is a flow chart of the implementation steps of the present invention. DETAILED DESCRIPTION

[0069] The example of the present invention provides a mathematical model of an LCL type grid-connected inverter in a grid-type operation mode, and adopts a PS0 algorithm to adjust the parameters of a passive controller online. The present invention is not only simple to implement, but also can observe the influence of different damping coefficients on system stability in real time, which is convenient for subsequent stability analysis.

[0070] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0071] Figure 1 is a control diagram of the optimization method of the present invention, Figure 1 It can be seen that the grid-connected inverter is an LCL-type grid-connected inverter, including a DC source, an inverter, an LCL-type filter, a line equivalent inductance and a three-phase power grid connected in series in sequence, and the LCL-type filter includes a bridge arm side inductance, a bridge arm side inductance equivalent resistance, a grid side inductance and a filter capacitor;

[0072] In this embodiment, L1 = 2mH, r f =0.1Ω, C=30μF, L2=lmH.

[0073] Figure 4 Flow chart of the implementation steps of the present invention. Figure 4It can be seen that the present invention provides a data-driven grid-connected inverter grid-connected passive control parameter optimization method, the optimization method designs a passive feedback controller by establishing a mathematical model of an LCL-type grid-connected inverter in a grid-connected operation mode and a Hamiltonian model with the LCL-type grid-connected inverter as the control object, and adopts a data-driven method to optimize two control parameters in the passive control law, the two control parameters being a first injection damping gain r1 and a second injection damping gain r2; the specific steps are as follows:

[0074] Step 1: Establish a mathematical model of the LCL grid-connected inverter in the grid-connected operation mode, and its expression is:

[0075]

[0076] In the formula, L1 is the inductance of the bridge arm side, C is the capacitance of the filter capacitor, r f is the resistance value of the inductor equivalent resistance on the bridge arm side, i 1d ,i 1q is the dq axis component of the inductor current on the bridge arm side, ω is the rated angular frequency of the power grid, u d ,u q is the dq axis component of the inverter output voltage, u Cd ,u Cq is the dq axis component of the filter capacitor voltage, i gd ,i gq is the dq-axis component of the grid-side current.

[0077] Step 2: Based on the dissipative passivity theory, a Hamiltonian model with the LCL grid-connected inverter as the control object is established, and its expression is:

[0078]

[0079] Where u is the passive feedback controller, u=[u d u q ] T ; x is the state variable of the Hamiltonian model, denoted as state variable x, x = [i 1d i 1q u Cd u Cq ] T , is the derivative of the state variable x; J is the antisymmetric interconnection matrix, R is the semi-positive definite dissipation matrix, G is the input matrix, ξ is the interference matrix; H(x) is the initial Hamiltonian energy storage function;

[0080] Substitute the mathematical model of the LCL grid-connected inverter in the grid-connected operation mode of step 1 into the Hamiltonian model to obtain the antisymmetric interconnection matrix J, semi-positive definite dissipation matrix R, input matrix G, and interference matrix ξ.

[0081] In this embodiment, the expressions of the antisymmetric interconnection matrix J, the semi-positive definite dissipation matrix R, the input matrix G, and the interference matrix ξ are respectively:

[0082]

[0083]

[0084] ξ=[-u Cd / L1 -u Cq / / L1 -ωCu Cq ωCu Cd ].

[0085] Step 3, according to the Hamiltonian model established in step 2, based on the interconnection and damping distribution passive control method, the passive control law of the passive feedback controller is obtained, and the expression of the passive control law includes the first injection damping gain r1 and the second injection damping gain r2.

[0086] In this embodiment, the implementation process of step 3 is as follows:

[0087] Step 3.1, define the expected equilibrium point x of the Hamiltonian model ＊ , whose expression is:

[0088]

[0089] in, represents the expected value of the dq-axis component of the inductor current on the bridge arm side, Represents the expected value of the dq-axis component of the filter capacitor voltage;

[0090] Step 3.2, define the initial Hamiltonian energy storage function H(x) as the sum of the energy stored in the inductor and filter capacitor on the bridge arm side, that is:

[0091]

[0092] Among them, x T is the transpose of the state variable x, Q is the diagonal parameter matrix, Q = diag{L1, L1, C, C};

[0093] Step 3.3, define the closed-loop Hamiltonian energy storage function H d (xx * ), whose expression is:

[0094]

[0095] Step 3.4, introduce the total structure matrix J d And the total damping matrix R d , whose expressions are:

[0096]

[0097] Step 3.5, according to step 3.4, the closed-loop system expression of the LCL grid-connected inverter based on the Hamiltonian equation is obtained:

[0098]

[0099] in, is the derivative of the expected value of the state variable of the Hamiltonian model;

[0100] Step 3.6, the expression of the passive feedback control law of the passive feedback controller u of the LCL type grid-connected inverter is:

[0101]

[0102] Where t is any moment in the operation of the grid-connected inverter.

[0103] Step 4: Sample the dq-axis component u of the filter capacitor voltage Cd ,u Cq , define the error e u , e u =u Cd -u Cq , select the first power evaluation function ITSE for e u Calculation is performed to obtain the fitness value in this scenario, wherein the first power evaluation function ITSE is the integral of any time t during the operation of the grid-connected inverter multiplied by the square of the error;

[0104] Use the particle swarm optimization algorithm to make the fitness value approach 0 or reach the preset number of iterations to obtain the global optimal fitness value g best , and the global optimal fitness value g best The corresponding estimated values ​​of the two control parameters are the optimal first injection damping gain r1' and the optimal second injection damping gain r2'.

[0105] In this embodiment, the global optimal fitness value g best The solution process is as follows:

[0106] Establish the evaluation function f o (e u ), whose expression is:

[0107]

[0108] Use the particle swarm optimization algorithm to find the estimated values ​​of the two control parameters that meet the requirements, and make the particle swarm optimization evaluation function f o (e u ) is the smallest. The specific steps are as follows:

[0109] Step 4.1, set the particle swarm size to n, particle dimension d to 2, particle swarm optimization algorithm learning factor 1 to c1, particle swarm optimization algorithm learning factor 2 to c2, particle swarm optimization algorithm initial inertia factor w1, particle swarm optimization algorithm end inertia factor w2, particle swarm optimization iteration number K, particle velocity minimum value v min , the maximum particle velocity v max ;

[0110] Step 4.2, set the speed range, position range and minimum fitness accuracy fit, and initialize the speed and position of each particle in the particle swarm optimization algorithm;

[0111] Step 4.3, calculate the fitness value of each particle of the particle swarm optimization algorithm, and record the smallest fitness value as the individual optimal fitness value p best , at this time p best =g best ;

[0112] Step 4.4, update the position and velocity of each particle in the particle swarm optimization algorithm, and perform K updates, and record any update in the K updates as the kth update, k = 1, 2, ..., K, k is a positive integer;

[0113] The velocity of the particle swarm optimization algorithm particle at the kth update is v k , position x k , the update formula is as follows:

[0114]

[0115] x k =x k-1 +v k

[0116] Wherein, rand is a random number between 0 and 1, k = 1, 2, ..., K;

[0117] Step 4.5, recalculate the fitness value of each particle of the particle swarm optimization algorithm, and record the minimum fitness value at the kth update as The minimum fitness value at the k-1th update is Will and For comparison, if Then Update the value of p best In contrast, p best The value of p remains unchanged; best The value of g best The value of p is compared. best <g best , then pbest The value is updated to g best In, conversely, g best The value remains unchanged;

[0118] Step 4.6, repeat the iterative process. When g best <fit or reaches the Kth update, end the iteration, output the global best fitness value g best And obtain the estimated value of the first injection damping gain r1 and the estimated value of the second injection damping gain r2 corresponding to the global best fitness value g best Record these two estimated values as the optimal first injection damping gain and the optimal second injection damping gain respectively.

[0119] To prove the beneficial effects of the present invention, an online simulation of the present invention was carried out.

[0120] Figure 2 is the fitness evolution curve diagram of the optimization method of the present invention under the first-power evaluation function ITSE. After 30 iterations, the fitness value is reduced by 49.7% from 0.0600 to 0.0302, that is, the global best fitness value g best is 0.0302; Figure 3 and Figure 4 are respectively Figure 2 The waveforms of the d-axis and q-axis of the inverter output voltage obtained by substituting the damping gains obtained in the scenario into the simulation. It can be seen from the figure that the response time for the dq-axis components of the inverter output voltage to converge and reach stability is short, and the stability effect is good.

Claims

1. A data-driven grid-connected inverter grid-type passive control parameter optimization method, wherein the grid-connected inverter is an LCL-type grid-connected inverter, comprising a DC source, an inverter, an LCL-type filter, a line equivalent inductance and a three-phase power grid connected in series in sequence, wherein the LCL-type filter comprises a bridge arm side inductance, a bridge arm side inductance equivalent resistance, a grid side inductance and a filter capacitor; It is characterized in that The optimization method designs a passive feedback controller by establishing a mathematical model of an LCL-type grid-connected inverter in a grid-connected operation mode and a Hamiltonian model with the LCL-type grid-connected inverter as a control object, and adopts a data-driven method to optimize two control parameters in the passive control law, wherein the two control parameters are a first injection damping gain r1 and a second injection damping gain r2; the specific steps are as follows: Step 1: Establish a mathematical model of the LCL grid-connected inverter in the grid-connected operation mode, and its expression is: In the formula, L1 is the inductance of the bridge arm side, C is the capacitance of the filter capacitor, r f is the resistance value of the inductor equivalent resistance on the bridge arm side, i 1d ,i 1q is the dq axis component of the inductor current on the bridge arm side, ω is the rated angular frequency of the power grid, u d ,u q is the dq axis component of the inverter output voltage, u Cd ,u Cq is the dq axis component of the filter capacitor voltage, i gd ,i gq is the dq axis component of the grid-side current; Step 2: Based on the dissipative passivity theory, a Hamiltonian model with the LCL grid-connected inverter as the control object is established, and its expression is: Where u is the passive feedback controller, u=[u d u q ] T ; x is the state variable of the Hamiltonian model, denoted as state variable x, is the derivative of the state variable x; J is the antisymmetric interconnection matrix, R is the semi-positive definite dissipation matrix, G is the input matrix, ξ is the interference matrix; H(x) is the initial Hamiltonian energy storage function; Substitute the mathematical model of the LCL grid-connected inverter in the grid-connected operation mode in step 1 into the Hamiltonian model to obtain the antisymmetric interconnection matrix J, the semi-positive definite dissipation matrix R, the input matrix G, and the interference matrix ξ; Step 3, according to the Hamiltonian model established in step 2, a passive control law of a passive feedback controller is obtained based on an interconnection and damping distribution passive control method, wherein the expression of the passive control law includes a first injection damping gain r1 and a second injection damping gain r2; Step 4: Sample the dq-axis component u of the filter capacitor voltage Cd ,u Cq , define the error e u , e u =u Cd -u cq , select the first power evaluation function ITSE for e u Calculation is performed to obtain the fitness value in this scenario, wherein the first power evaluation function ITSE is the integral of any time t during the operation of the grid-connected inverter multiplied by the square of the error; Use the particle swarm optimization algorithm to make the fitness value approach 0 or reach the preset number of iterations to obtain the global optimal fitness value g best , and the global optimal fitness value g best The corresponding estimated values ​​of the two control parameters are the optimal first injection damping gain r1' and the optimal second injection damping gain r2'.

2. The data-driven grid-connected inverter grid-connected passive control parameter optimization method according to claim 1, characterized in that: The implementation process of step 3 is as follows: Step 3.1, define the expected equilibrium point x of the Hamiltonian model ＊ , whose expression is: in, represents the expected value of the dq-axis component of the inductor current on the bridge arm side, Represents the expected value of the dq-axis component of the filter capacitor voltage; Step 3.2, define the initial Hamiltonian energy storage function H(x) as the sum of the energy stored in the inductor and filter capacitor on the bridge arm side, that is: Among them, x T is the transpose of the state variable x, Q is the diagonal parameter matrix, Q = diag{L1, L1, C, C}; Step 3.3, define the closed-loop Hamiltonian energy storage function H d (xx * ), whose expression is: Step 3.4, introduce the total structure matrix J d And the total damping matrix R d , whose expressions are: Step 3.5: According to step 3.4, the closed-loop system expression of the LCL grid-connected inverter based on the Hamiltonian equation is obtained: in, is the derivative of the expected value of the state variable of the Hamiltonian model; Step 3.6, the expression of the passive feedback control law of the passive feedback controller u of the LCL type grid-connected inverter is: Where t is any moment in the operation of the grid-connected inverter.

3. The data-driven grid-connected inverter grid-connected passive control parameter optimization method according to claim 2, characterized in that: The global optimal fitness value g in step 4 best The solution process is as follows: Establish the evaluation function f o (e u ), whose expression is: Use the particle swarm optimization algorithm to find the estimated values ​​of the two control parameters that meet the requirements, and make the particle swarm optimization evaluation function f o (e u ) is the smallest. The specific steps are as follows: Step 4.1, set the particle swarm size to n, particle dimension d to 2, particle swarm optimization algorithm learning factor 1 to c1, particle swarm optimization algorithm learning factor 2 to c2, particle swarm optimization algorithm initial inertia factor w1, particle swarm optimization algorithm end inertia factor w2, particle swarm optimization iteration number K, particle velocity minimum value v min , the maximum particle velocity v max ; Step 4.2, set the speed range, position range and minimum fitness accuracy fit, and initialize the speed and position of each particle in the particle swarm optimization algorithm; Step 4.3, calculate the fitness value of each particle of the particle swarm optimization algorithm, and record the smallest fitness value as the individual optimal fitness value p best , at this time p best =g best ; Step 4.4, update the position and velocity of each particle in the particle swarm optimization algorithm, and perform K updates, and record any update in the K updates as the kth update, k = 1, 2, ..., K, k is a positive integer; The velocity of the particle swarm optimization algorithm particle at the kth update is v k , position x k , the update formula is as follows: x k =x k-1 +v k Wherein, rand is a random number between 0 and 1, k = 1, 2, ..., K; Step 4.5, recalculate the fitness value of each particle of the particle swarm optimization algorithm, and record the minimum fitness value at the kth update as The minimum fitness value at the k-1th update is Will and For comparison, if Then Update the value of p best In contrast, p best The value of p remains unchanged; best The value of g best The value of p is compared. best <g best , then p best Update the value of g best In contrast, g best The value of remains unchanged; Step 4.6, repeat the iterative process, when g best When <fit or the Kth update is reached, the iteration ends and the global optimal fitness value g is output best , and obtain the global optimal fitness value g best The corresponding estimated value of the first injection damping gain r1 and the estimated value of the second injection damping gain r2 are respectively recorded as the optimal first injection damping gain and the optimal second injection damping gain.

4. The data-driven grid-connected inverter grid-connected passive control parameter optimization method according to claim 1, characterized in that: The expressions of the antisymmetric interconnection matrix J, the semi-positive definite dissipation matrix R, the input matrix G, and the interference matrix ξ in step 2 are respectively: ξ=[-u Cd / L1 -u Cq / L1 -ωCu Cq ωCu Cd ]。

Images