Neurodynamics optimization-based grid-connected inverter continuous set model prediction control method

Abstract

The invention discloses a grid-connected inverter continuous set model prediction control method based on neurodynamics optimization. The method comprises the following steps: establishing a continuous time state space model of an LCL type grid-connected inverter in a static coordinate system; performing full-order discretization processing on the continuous time state space model through a zero-order retention method to obtain a discrete state space model; according to the discrete state space model, constructing a standard quadratic form cost function about an unknown control voltage increment sequence of the LCL type grid-connected inverter; solving the standard quadratic form cost function to obtain an unconstrained optimal control increment sequence; and according to the unconstrained optimal control increment sequence, constructing a nonlinear programming model of multiple heterogeneous constraints, and solving a nonlinear programming problem through a constructed neurodynamics optimizer so as to realize continuous set model prediction control of the LCL type grid-connected inverter. According to the method, the problem that the dynamic response speed and the control precision of the inverter are insufficient in the existing method is solved.

Description

Technical Field

[0001] This invention relates to the field of power electronics technology, and in particular to a continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization. Background Technology

[0002] Inverters, as the core devices for power conversion, play a crucial role in fields such as new energy grid connection, uninterruptible power supplies, and motor drives. To improve the dynamic response speed and control accuracy of inverters, Model Predictive Control (MPC) has become a research hotspot for next-generation high-performance control strategies due to its inherent fast dynamic response and multi-objective optimization capabilities.

[0003] However, applying MPC to practical inverter systems faces a core challenge: how to efficiently and accurately handle the system's physical constraints. The inverter's output voltage is limited by the DC bus voltage, and its boundary lies in the two-phase stationary coordinate system (…). In a coordinate system, it is represented as a hexagon; at the same time, in order to protect power devices from overcurrent damage under conditions such as sudden load increases and grid oscillations, the inverter's output current must also be strictly limited within a safe range. These two types of constraints together define the safe operating area of ​​the system.

[0004] Existing constraint handling methods have the following limitations: (1) Simple limiting method: This is the simplest method. When the voltage calculated by the controller exceeds the boundary, its amplitude or component is directly limited to the boundary value. Although this method is simple, it seriously destroys the optimal control direction calculated by the controller, resulting in a decrease in the dynamic performance of the system, distortion of the current waveform, and may cause system oscillation.

[0005] (2) Traditional Quadratic Programming (QP) Solvers: Standard MPC constraint problems can be modeled as QP problems and solved using traditional optimization algorithms such as the effective set method or interior point method. Although these algorithms can find exact solutions, they are complex to implement and have an excessive computational burden for high-speed inverter systems that need to complete calculations in each switching cycle. They can only run on specific high-performance controllers, which greatly restricts their industrial applications.

[0006] Therefore, there is an urgent need in this field for a continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization, which can maintain the fast dynamic response of MPC, accurately and efficiently handle the inherent multiple and heterogeneous (linear and nonlinear) constraints of the inverter, and ensure the computational feasibility of real-time implementation. Summary of the Invention

[0007] This invention provides a continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization to overcome the aforementioned technical problems.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows: A predictive control method for grid-connected inverters based on a continuous set model optimized by neurodynamics includes the following steps: S1: Establish a continuous-time state-space model of the LCL grid-connected inverter in the stationary coordinate system; S2: The continuous-time state-space model is discretized to obtain a discrete state-space model by using the zero-order preservation method. S3: Based on the discrete state-space model, construct a standard quadratic cost function for the unknown control voltage increment sequence of the LCL grid-connected inverter; S4: Solve the standard quadratic cost function to obtain the unconstrained optimal control increment sequence; S5: Based on the unconstrained optimal control increment sequence, a nonlinear programming model with multiple heterogeneous constraints is constructed, and the nonlinear programming model is solved by the constructed neurodynamics optimizer to achieve continuous set model predictive control of LCL grid-connected inverter.

[0009] Furthermore, the continuous-time state-space model of the LCL-type grid-connected inverter established in S1 in the stationary coordinate system is expressed as follows:

[0010]

[0011]

[0012] In the formula: Represents the state vector; express abbreviated form and ; Indicates the output vector; express The abbreviated form; These represent the inverter output side. Shaft current, output side Axis current, grid side Axis current, grid side shaft current, filter capacitor shaft voltage and filter capacitor Shaft voltage; These represent the system matrix, input matrix, disturbance matrix, and output matrix, respectively. express The first derivative; express The abbreviated form; Represents the input vector; Represents the perturbation vector; These represent the inverter-side inductance and parasitic resistance, respectively. These represent the grid-side inductance and parasitic resistance, respectively. Indicates the filter capacitor; Indicates the inverter output side shaft and Voltage on the shaft; Indicates the net side Axis voltage and grid side Shaft voltage.

[0013] Furthermore, the expression for the discrete state-space model described in S2 is:

[0014] , , ,

[0015] In the formula: Indicates the sampling time; j Indicates the prediction time domain and j = 1, 2, …, ; Indicates the step size in the prediction time domain; Represents the state vector of a discrete state-space model; F represents the input vector and disturbance vector of the discrete state-space model; , H represents the time-invariant discrete-time state matrix, input matrix, perturbation matrix, and output state matrix, respectively; I represents the identity matrix; and T represents the sampling time. This represents the output of the discrete state-space model.

[0016] Furthermore, step S3 specifically includes the following steps: S31: Based on the discrete state-space model, construct a cost function for the unknown control voltage increment sequence of the LCL-type grid-connected inverter. J ; And the cost function J The expression is

[0017]

[0018] In the formula: Represents the reference vector; This represents the difference in controller output between adjacent time points for an LCL-type grid-connected inverter; Indicate design parameters; This represents the weight matrix composed of the weight factors in the cost function; S32: Defining the Future The predicted output vector of the step The predicted output vector The corresponding function expression is based on the discrete state-space model and the current state vector. Future control increment sequence and future perturbation sequences Constructed; and , , ; Indicates the predicted output The elements, i.e., the output of the discrete state-space model. Transpose of; This represents the difference in controller output between adjacent times for an LCL-type grid-connected inverter. Transpose of; The perturbation vector representing the discrete state-space model Transpose of; The future Predicted output of step The function expression is

[0019] In the formula: Φ,Γ,Ξ represent parameter matrices; S33: According to the future Predicted output of step The function will be the cost function J Rewritten as an unknown control increment sequence The standard quadratic form function; And the expression for the standard quadratic form function is:

[0020]

[0021]

[0022] In the formula: Represents the Hessian matrix; Represents the augmented weight matrix; Represents the gradient vector; Represents the intermediate parameter vector and ; Represents the reference vector The transpose of .

[0023] Furthermore, the formula for obtaining the unconstrained optimal control increment sequence described in S4 is as follows:

[0024]

[0025] In the formula: Indicates the current sampling time k The elements corresponding to the optimal control increment sequence; Indicates the current sampling time k The corresponding control increment sequence; Indicates the current sampling time k The corresponding gradient vector.

[0026] Furthermore, S5 specifically includes the following steps: S51: Obtain the unconstrained control voltage of the LCL grid-connected inverter based on the unconstrained optimal control increment sequence; And the formula for obtaining the unconstrained control voltage is as follows:

[0027] In the formula: Indicates the current sampling time k The unconstrained control voltage corresponding to LCL-type grid-connected inverters; This indicates the actual control voltage applied in the previous control cycle; Elements representing the optimal control increment sequence; S52: Construct a nonlinear programming model with multiple heterogeneous constraints based on the unconstrained control voltage; Furthermore, the nonlinear programming model with multiple heterogeneous constraints includes optimization functions and mixed constraint conditions; The optimization function The expression is:

[0028] In the formula: This indicates the actual control voltage applied during the current control cycle; The expression for the hybrid constraint condition is:

[0029] In the formula: Indicates linear voltage constraint; Indicates nonlinear current constraint; Indicates the amount of intermediate parameters; This indicates extracting current from the state vector. The selection matrix and ; Indicates the current sampling time k The corresponding state vector; Describes a vector consisting of the maximum current values ​​and , These represent the inverter output side. The maximum value of the shaft current and the output side The maximum value of the shaft current; This indicates the DC bus voltage of the inverter; S53: Constructing a neurodynamics optimizer; And the expression for the neurodynamics optimizer is:

[0030]

[0031] In the formula: , Represents the time constant; Describes the gradient of the objective function and ; Represents the nonlinear current constraint function pair The gradient; Indicates the learning rate; Indicates time parameters; Indicates positive constant gain; Describes the projection operator projected onto a non-negative interval and ; Represents the Lagrange multipliers; S54: Solve the nonlinear programming problem using the neural dynamics optimizer to achieve continuous set model predictive control of the LCL grid-connected inverter.

[0032] Beneficial Effects: This invention provides a continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization. According to the discrete state-space model, a standard quadratic cost function is constructed for the unknown control voltage increment sequence of the LCL-type grid-connected inverter. This facilitates the automatic search for control sequences that can suppress capacitor voltage fluctuations during the optimization process, thereby effectively attenuating the resonance peak of the LCL filter and improving the system's stability and current quality under disturbances. Based on the solved and obtained unconstrained optimal control increment sequence, a nonlinear programming model with multiple heterogeneous constraints is constructed. The nonlinear programming problem is solved using the constructed neurodynamic optimizer to achieve continuous set model predictive control of the LCL-type grid-connected inverter. This invention maintains the fast dynamic response of continuous set model predictive control for LCL-type grid-connected inverters, accurately and efficiently handles the inherent multiple heterogeneous (linear and nonlinear) constraints of the inverter, and ensures the computational feasibility of real-time implementation. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 The flowchart shows the continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization according to the present invention. Figure 2 This is a schematic diagram of the three-level LCL grid-connected inverter circuit topology in this embodiment; Figure 3 This is a core block diagram of the method described in this embodiment; Figure 4 The above are simulation diagrams of grid-connected current and control output scatter plots for the steady-state neurodynamic optimization algorithm in this embodiment. Figure 5 The above are simulation diagrams of grid-connected current and control output scatter plots for the proposed neurodynamic optimization method under the simulated current surge condition in this embodiment. Figure 6 The above are simulation diagrams of grid-connected current and control output scatter plots under the simulated current surge condition using the traditional active set solver method in this embodiment. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] This embodiment provides a continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization, such as... Figure 1 As shown, the specific steps include: S1: Targeting Figure 2 The LCL grid-connected inverter shown is used to establish a continuous-time state-space model of the LCL grid-connected inverter in the stationary coordinate system. And the expression for the continuous-time state-space model is: (1) In the formula: Represents the state vector; express abbreviated form and ; Indicates the output vector; express The abbreviated form; These represent the inverter output side. Shaft current, output side Axis current, grid side Axis current, grid side shaft current, filter capacitor shaft voltage and filter capacitor Shaft voltage; These represent the system matrix, input matrix, disturbance matrix, and output matrix, respectively. express The first derivative; express The abbreviated form; Represents the input vector; Denotes the perturbation vector; where: (2) (3) (4) In the formula: These represent the inverter-side inductance and parasitic resistance, respectively. These represent the grid-side inductance and parasitic resistance, respectively. Indicates the filter capacitor; Indicates the inverter output side shaft and Voltage on the shaft; Indicates the net side Axis voltage and grid side Shaft voltage; S2: The continuous-time state-space model is discretized to obtain a discrete state-space model by using the zero-order preservation method. Specifically, in this embodiment, to deploy the model into the MPC algorithm, the continuous-time state-space model (1) needs to be discretized using the zero-order-preserving-full-order discretization method; that is, at the sampling time... k Department, targeting the future j Each prediction time domain ( j = 1, 2, …, The discrete state-space model of ) is (5) In the formula: , , , ; Indicates the sampling time;j Indicates the prediction time domain and j = 1, 2, …, ; Indicates the step size in the prediction time domain; Represents the state vector of a discrete state-space model; F represents the input vector and disturbance vector of the discrete state-space model; , H represents the time-invariant discrete-time state matrix, input matrix, perturbation matrix, and output state matrix, respectively; I represents the identity matrix; and T represents the sampling time. This represents the output of the discrete state-space model; S3: Based on the discrete state-space model, construct a standard quadratic cost function for the unknown control voltage increment sequence of the LCL grid-connected inverter; like Figure 3 As shown, the specific steps include: S31: In this embodiment, the goal of controller optimization is to find an optimal control increment sequence to minimize a comprehensive cost function, that is, to construct a cost function for the unknown control voltage increment sequence of the LCL grid-connected inverter based on the discrete state-space model. J ; And the cost function J The expression is (6) Where: cost function J It consists of two parts: the state tracking error term and the control smoothing term. Represents the reference vector; This indicates a reference value for the inverter-side current. The reference vector represents the capacitor voltage reference value; by incorporating the capacitor voltage into the cost function and setting its reference value to zero, it cleverly achieves a damping function. The optimization process automatically seeks a control sequence that can suppress capacitor voltage fluctuations, thereby effectively attenuating the resonance peak of the LCL filter and improving the system's stability and current quality under disturbances. This is more robust than the traditional method of controlling only the current. This represents the difference in controller output between adjacent times for an LCL-type grid-connected inverter. ; Used to punish the future The size of each control increment within a step is determined to ensure smooth control. Represents the norm; The weight matrix, composed of the weight factors in the cost function, represents: (7) In the formula: Indicate design parameters and , ; S32: Defining the Future The predicted output vector of the step The predicted output vector The corresponding function expression is based on the discrete state-space model and the current state vector. Future control increment sequence and future perturbation sequences Constructed; and , , ; Indicates the predicted output The elements, i.e., the output of the discrete state-space model. Transpose of; This represents the difference in controller output between adjacent times for an LCL-type grid-connected inverter. Transpose of; The perturbation vector representing the discrete state-space model Transpose of; The future Predicted output of step The function expression is (8) In the formula: Φ,Γ,Ξ represent the discretized model , and control increment The parameter matrix is ​​obtained by recursion based on expert experience; S33: According to the future Predicted output of step The function will be the cost function J Rewritten as an unknown control increment sequence The standard quadratic form function; And the expression for the standard quadratic form function is: (9) In the formula: Describe the Hessian matrix and ; Indicates by Extended from this, it is an augmented weight matrix that considers multi-step control of incremental smoothness based on expert experience, according to the prediction time domain. ,Will Repeated arrangement with The block diagonal matrix form of dimension matching, and then with Add them together to obtain the complete Hessian matrix. ; Describes the gradient vector and ; Represents the intermediate parameter vector and ; Represents the reference vector Transpose of; S4: Solve the standard quadratic cost function to obtain the unconstrained optimal control increment sequence; Specifically, by letting the standard quadratic form function right The partial derivatives are zero, allowing for the analytical calculation of the unconstrained optimal control increment sequence. Its expression is (10) (11) In the formula: Indicates the current sampling time k The elements corresponding to the optimal control increment sequence; Indicates the current sampling time k The corresponding control increment sequence; Indicates the current sampling time k The corresponding gradient vector. In this embodiment, based on the rolling optimization principle of model predictive control, although this embodiment calculates the future... The optimal control sequence for the step, but in the current sampling period k Only the first set of elements in the sequence is executed, i.e., the optimal control increment. ; S5: Based on the unconstrained optimal control increment sequence, a nonlinear programming model with multiple heterogeneous constraints is constructed, and the nonlinear programming model is solved by the constructed neurodynamics optimizer to achieve continuous set model predictive control of LCL grid-connected inverters. The specific steps include: S51: Based on the unconstrained optimal control increment sequence, obtain the theoretically optimal, but not yet physically constrained, unconstrained control voltage of the LCL grid-connected inverter; And the formula for obtaining the unconstrained control voltage is as follows: (12) In the formula: Indicates the current sampling time k The unconstrained control voltage corresponding to the LCL-type grid-connected inverter is described in this embodiment. It combines the foresight of multi-step prediction with the stability of active damping and is the theoretically optimal solution without considering any constraints. This indicates the actual control voltage applied in the previous control cycle; Elements representing the optimal control increment sequence; S52: In this embodiment, due to... To address the potential violation of inverter physical limits, a novel neurodynamics optimizer is employed to solve the following nonlinear programming problem with multiple heterogeneous constraints online and in real-time, thereby obtaining the optimal feasible voltage to be applied to the inverter. In each control cycle k In order to accurately describe the optimization problem solved in this embodiment, a nonlinear programming model with multiple heterogeneous constraints is constructed based on the unconstrained control voltage; and the nonlinear programming model with multiple heterogeneous constraints includes optimization functions and mixed constraint conditions. The expression for the optimization function is: (13) In the formula: This indicates the actual control voltage applied during the current control cycle; Indicates finding the variable Make the objective function The smallest optimization function; The expression for the hybrid constraint condition is: (14) In the formula: Indicates linear voltage constraint; Indicates nonlinear current constraint; Indicates the amount of intermediate parameters; This indicates extracting current from the state vector. The selection matrix and ; Indicates the current sampling time k The corresponding state vector; Describes a vector consisting of the maximum current values ​​and , These represent the inverter output side. The maximum value of the shaft current and the output side The maximum value of the shaft current is set automatically by factors such as power and components. The inverter DC bus voltage is represented; constraint (1) is a set of linear inequality constraints describing the hexagonal boundary of the SVPWM voltage; constraint (2) is a quadratic (nonlinear) inequality constraint describing the current amplitude limit on the inverter side. S53: Constructing a Neurodynamic Optimizer; To solve the above nonlinear programming model accurately and efficiently, this embodiment designs a hybrid constraint processing strategy based on the idea of ​​Lagrange programming neural networks, namely a neurodynamic optimizer. Its core lies in using different optimal processing methods for constraints of different properties: A. For nonlinear current constraints This embodiment introduces a Lagrange multiplier. To handle this constraint, according to the Karush-Kuhn-Tucker (KKT) conditions, at the optimal point, the Lagrange multiplier and the constraint function act together on the gradient of the system; this embodiment utilizes this principle to introduce a term related to the Lagrange multiplier during the dynamic evolution of the neurodynamic optimizer. and gradient The relevant dynamic correction term, which can be geometrically understood as a repulsive force, occurs when the solution... When approaching or attempting to cross the circular boundary of the current source, this repulsive force is activated, and its direction always points inward into the safe zone, thus guiding the solution. Move along a smooth circular boundary, rather than simply truncating the boundary.

[0037] B. Regarding linear voltage constraints For a set of simple linear inequality constraints, a feasible region of a convex polygon is defined. This embodiment employs a highly efficient projection operator. This operator can instantly and accurately map potentially out-of-bounds solutions back to the nearest point within the hexagonal feasible region after each iteration, thus rigidly ensuring the absolute satisfaction of voltage constraints.

[0038] The neurodynamic optimizer described in this embodiment is described by the following set of coupled differential equations, and is implemented in the digital controller in its discrete form: For the optimization equation of the projected part, i.e., the linear voltage constraint: (15) In the formula: Represents the time constant; Describes the gradient of the objective function and ; Represents the nonlinear current constraint function pair The gradient; Indicates the learning rate; This represents the time parameter; the equation describes the control voltage. The motion trend is influenced by the combined attraction from the objective function and the repulsion from the current constraint, and is superimposed with the projection operation of the voltage boundary. For dual variables The dynamic equation is the optimization equation with nonlinear current constraints: (16) In the formula: Represents the time constant; Indicates positive constant gain; Describes the projection operator projected onto a non-negative interval and This ensured ; This represents the Lagrange multipliers; the optimization equation describes the Lagrange multipliers. The evolutionary pattern enables it to dynamically sense violations of current constraints; S54: Solve the nonlinear programming problem using the neural dynamics optimizer to achieve continuous set model predictive control of the LCL grid-connected inverter.

[0039] Simulation verification: This embodiment uses Matlab / simulink for simulation, and the main parameters in the simulation are shown in Table 1. Table 1. Simulation Parameters

[0040] In the table: Indicates the sampling frequency; f Indicates the power grid frequency; Indicates the magnitude of the grid current; Vg Indicates the effective value of the grid voltage; like Figure 4 As shown, Figure 4 (a) shows the simulation results of the three-phase current of the power grid under steady-state operation using the method described in this embodiment. As can be seen from the figure, the total harmonic distortion (THD) of the power grid current is only 0.63%, which is extremely low. This result proves that the control framework of the present invention not only achieves excellent control performance but also does not exhibit overcurrent phenomena. Figure 4 (b) is a scatter plot of the control output, which clearly shows whether the voltage or current is in an effective state. The elliptical area above the scatter plot shows that some unconstrained solutions (marked with boxes) exist during the startup phase; these solutions exceed the hexagonal voltage constraint boundary. At this point, the neurodynamic optimizer is activated, projecting these unconstrained solutions into the hexagonal boundary to ensure that the control output does not exceed the boundary, verifying the effectiveness of the voltage constraint and achieving safe control. Figure 5 As shown, Figure 5 (a) shows the simulation results of the three-phase current of the power grid under a specific operating condition using the method proposed in this embodiment. The reference current is set to 16A and the constraint current to 15A, aiming to simulate an instantaneous overcurrent scenario and verify the current constraint handling capability of the method described in this embodiment. As can be seen from the figure, the current is stably controlled within the constraint range of 15A, proving that the method can effectively achieve current constraint control. Figure 5(b) is a scatter plot of the corresponding control output. It can be observed from the figure that, under current constraint, the control output exhibits jitter within the hexagonal region. This phenomenon is a result of the synergistic effect of the repulsive force under neurodynamic current constraint and LCL resonance suppression control. Figure 6 As shown, Figure 6 (a) shows the network-side current diagram of a traditional effective set QP solver under the same constraint conditions. This method requires linearization approximation of the current constraint circle, and the forced hard constraint during current constraint application exacerbates LCL resonance, leading to more harmonics and preventing precise current constraint control. The red dashed line in the diagram indicates a 15A current. Figure 5 (a) It is easy to see from the comparison that the traditional method is significantly inferior to the method of the present invention in terms of constraint handling capability in this scenario. Figure 6 (b) is a scatter plot of the control output of a traditional effective set QP solver. (Comparison) Figure 5 (b) It is evident that the scattered points are more chaotic, a problem also stemming from the resonance phenomenon caused by the approximately linearized hard constraints in traditional methods. Therefore, as can be seen from the above analysis, compared to the traditional MPC strategy, the method proposed in this embodiment can accurately handle current and voltage constraints, ensuring the stable operation of the LCL grid-connected inverter.

[0041] The beneficial effects of the method described in this embodiment are as follows: 1. High constraint accuracy and good waveform quality: The Lagrange multiplier method is used to directly handle the real quadratic circular current state constraint, avoiding the errors caused by the linearization approximation of traditional methods. When the constraint is activated, the controller smoothly scales the command instead of clipping, and can output a current waveform with limited amplitude but still maintaining high quality sinusoidal waveform, significantly reducing harmonic content.

[0042] 2. High safety: The method described in this embodiment can simultaneously and strictly guarantee both voltage and current constraints, effectively preventing power devices from being damaged due to overvoltage or overcurrent, and improving the reliability and robustness of the system.

[0043] 3. Fast dynamic response: It retains the advantage of fast response of MPC algorithm. At the same time, by optimally utilizing the complete feasible region defined by hexagonal voltage boundary and circular current boundary, it can provide greater control force in transient process compared with simple limiting method, thus achieving faster dynamic response.

[0044] 4. Computationally efficient and easy to implement: The neurodynamics optimizer is essentially a set of simple differential equations, whose discretization can be implemented in just a few lines of iterative code. Its simple structure makes it easy to implement on a digital signal processor. Its fast convergence characteristics ensure real-time computational feasibility at high switching frequencies.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization, characterized in that, Specifically, the following steps are included: S1: Establish a continuous-time state-space model of the LCL grid-connected inverter in the stationary coordinate system; S2: The continuous-time state-space model is discretized to obtain a discrete state-space model by using the zero-order preservation method. S3: Based on the discrete state-space model, construct a standard quadratic cost function for the unknown control voltage increment sequence of the LCL grid-connected inverter; S4: Solve the standard quadratic cost function to obtain the unconstrained optimal control increment sequence; S5: Based on the unconstrained optimal control increment sequence, a nonlinear programming model with multiple heterogeneous constraints is constructed, and the nonlinear programming model is solved by the constructed neurodynamics optimizer to achieve continuous set model predictive control of LCL grid-connected inverter.

2. The method for predictive control of a grid-connected inverter based on a continuous set model optimized by neurodynamics, as described in claim 1, is characterized in that... The continuous-time state-space model of the LCL-type grid-connected inverter established in S1 in the stationary coordinate system is expressed as follows: In the formula: Represents the state vector; express abbreviated form and ; Indicates the output vector; express The abbreviated form; These represent the inverter output side. Shaft current, output side Axis current, grid side Axis current, grid side shaft current, filter capacitor shaft voltage and filter capacitor Shaft voltage; These represent the system matrix, input matrix, disturbance matrix, and output matrix, respectively. express The first derivative; express The abbreviated form; Represents the input vector; Represents the perturbation vector; These represent the inverter-side inductance and parasitic resistance, respectively. These represent the grid-side inductance and parasitic resistance, respectively. Indicates the filter capacitor; Indicates the inverter output side shaft and Voltage on the shaft; Indicates the net side Axis voltage and grid side Shaft voltage.

3. The method for continuous set model predictive control of grid-connected inverters based on neurodynamic optimization according to claim 2, characterized in that, The expression for the discrete state-space model described in S2 is: , , , In the formula: Indicates the sampling time; j Indicates the prediction time domain and j = 1, 2, …, ; Indicates the step size in the prediction time domain; Represents the state vector of a discrete state-space model; F represents the input vector and disturbance vector of the discrete state-space model; , H represents the time-invariant discrete-time state matrix, input matrix, perturbation matrix, and output state matrix, respectively; I represents the identity matrix; and T represents the sampling time. This represents the output of the discrete state-space model.

4. The continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization according to claim 2, characterized in that, S3 specifically includes the following steps: S31: Based on the discrete state-space model, construct a cost function for the unknown control voltage increment sequence of the LCL-type grid-connected inverter. J ; And the cost function J The expression is In the formula: Represents the reference vector; This represents the difference in controller output between adjacent time points for an LCL-type grid-connected inverter; Indicate design parameters; This represents the weight matrix composed of the weight factors in the cost function; S32: Defining the Future The predicted output vector of the step The predicted output vector The corresponding function expression is based on the discrete state-space model and the current state vector. Future control increment sequence and future perturbation sequences Constructed; and , , ; Indicates the predicted output The elements, i.e., the output of the discrete state-space model. Transpose of; This represents the difference in controller output between adjacent times for an LCL-type grid-connected inverter. Transpose of; The perturbation vector representing the discrete state-space model Transpose of; The future Predicted output of step The function expression is In the formula: Φ,Γ,Ξ represent parameter matrices; S33: According to the future Predicted output of step The function will be the cost function J Rewritten as an unknown control increment sequence The standard quadratic form function; And the expression for the standard quadratic form function is: In the formula: Represents the Hessian matrix; Represents the augmented weight matrix; Represents the gradient vector; Represents the intermediate parameter vector and ; Represents the reference vector The transpose of .

5. The method for continuous set model predictive control of grid-connected inverters based on neurodynamic optimization according to claim 4, characterized in that, The formula for obtaining the unconstrained optimal control increment sequence described in S4 is as follows: In the formula: Indicates the current sampling time k The elements corresponding to the optimal control increment sequence; Indicates the current sampling time k The corresponding control increment sequence; Indicates the current sampling time k The corresponding gradient vector.

6. The continuous set model predictive control method for grid-connected inverters based on neurodynamic optimization according to claim 5, characterized in that, S5 specifically includes the following steps: S51: Obtain the unconstrained control voltage of the LCL grid-connected inverter based on the unconstrained optimal control increment sequence; And the formula for obtaining the unconstrained control voltage is as follows: In the formula: Indicates the current sampling time k The unconstrained control voltage corresponding to LCL-type grid-connected inverters; This indicates the actual control voltage applied in the previous control cycle; Elements representing the optimal control increment sequence; S52: Construct a nonlinear programming model with multiple heterogeneous constraints based on the unconstrained control voltage; Furthermore, the nonlinear programming model with multiple heterogeneous constraints includes optimization functions and mixed constraint conditions; The expression for the optimization function is: In the formula: This indicates the actual control voltage applied during the current control cycle; The expression for the hybrid constraint condition is: In the formula: Indicates linear voltage constraint; Indicates nonlinear current constraint; Indicates the amount of intermediate parameters; This indicates extracting current from the state vector. The selection matrix and ; Indicates the current sampling time k The corresponding state vector; Describes a vector consisting of the maximum current values ​​and , These represent the inverter output side. The maximum value of the shaft current and the output side The maximum value of the shaft current; This indicates the DC bus voltage of the inverter; S53: Constructing a neurodynamics optimizer; And the expression for the neurodynamics optimizer is: In the formula: , Represents the time constant; Describes the gradient of the objective function and ; Represents the nonlinear current constraint function pair The gradient; Indicates the learning rate; Indicates time parameters; Indicates positive constant gain; Describes the projection operator projected onto a non-negative interval and ; Represents the Lagrange multipliers; S54: Solve the nonlinear programming problem using the neural dynamics optimizer to achieve continuous set model predictive control of the LCL grid-connected inverter.

Images