Microgrid group topology switching stability analysis method and system

Abstract

The invention provides a microgrid group topology switching stability analysis method and system. The method comprises the steps that modeling is conducted on a microgrid system, and a Gaboff equation is introduced; solving a giboff equation in the micro-grid system by using a neural network to obtain a Lyapunov function; analyzing stability through a Lyapunov function, and calculating an attraction domain of the system; and further analyzing the stability of the micro-grid group system under different topology configurations so as to ensure that the system topology switching is stable. The technical problems that the dependence degree on explicit mathematical derivation or an approximate model is high, error accumulation stability evaluation is not accurate, and stability evaluation is not comprehensive are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of microgrid cluster topology switching stability analysis, and particularly relates to a microgrid cluster topology switching stability analysis method. BACKGROUND

[0002] Microgrids play an important role in modern power systems, especially in ensuring reliable power distribution in isolated or small areas. Microgrid topology switching is a key factor affecting system stability. Accurate prediction and analysis of the impact of microgrid topology changes on system stability can effectively prevent power system failures and improve overall system performance.

[0003] Traditional microgrid cluster topology switching stability analysis methods mainly rely on analytical methods and numerical simulation. Analytical methods use mathematical models to analyze system stability using classical methods such as Lyapunov stability theory or Zubov equation. These methods are usually suitable for simpler or lower-dimensional systems and can provide clear stability judgments. However, as the size and complexity of the system increases, traditional methods face significant limitations, especially when dealing with high-dimensional and nonlinear systems, analytical methods often cannot find effective analysis functions or equations. In addition, numerical simulation, although it can handle more complex systems, is also limited by computing resources and time, with high computational complexity and precision problems when dealing with large-scale data.

[0004] To solve the above problems, existing technologies use data-driven methods, especially in the absence of a complete analytical model, to use historical data for system modeling and analysis. Data-driven methods can more efficiently handle high-dimensional complex systems by extracting features and patterns from the system's historical behavior, especially in microgrid topology switching stability analysis. Data-driven methods can automatically extract potential patterns of system behavior from large-scale operational data, and then perform stability evaluation. This makes it possible to effectively handle high-dimensional and nonlinear systems, reducing the dependence on complex analytical solutions, improving the efficiency and adaptability of system analysis, and reducing the dependence on high-dimensional complex calculations. However, these methods still face some challenges, such as: high cost of data acquisition and processing, data quality and completeness have a greater impact on the results, and in the absence of a clear physical model, data-driven methods may be subject to overfitting and inaccurate model risks.

[0005] The prior art patent application document with the publication number CN119209702A discloses a method for analyzing the large signal stability of a complex micro-grid, which includes: establishing a nonlinear state space model of each component of the micro-grid, using the model of each component to constitute a system-level nonlinear model of the alternating current micro-grid, using the Takagi-Sugeno (TS) multiple modeling method to segment linearize the original nonlinear model, and constructing a linear matrix inequality set. Then, the linear matrix inequality set is solved, an energy function of the micro-grid is constructed, and the energy function is mapped to a two-dimensional plane to obtain an analytical expression of an attraction domain. Finally, the influence mechanism of the system parameters on the large signal stability is obtained by adjusting the system parameters and solving the attraction domain under each working condition. However, the foregoing prior art scheme has the following technical defects: 1. Segmentation linearization error: the Takagi-Sugeno (TS) multiple modeling method is used to segment linearize the nonlinear model, which may simplify the dynamic characteristics of a high-dimensional system and introduce model errors; 2. High computational complexity: the energy function and the attraction domain are solved by constructing a linear matrix inequality (LMI), which has a large computational burden and low efficiency in a high-dimensional system; 3. Fuzzy boundary of the attraction domain: after the energy function based on the LMI is mapped to a two-dimensional plane, the boundary of the attraction domain is an approximate analytical expression, and the accuracy is limited; The prior art patent application document with the publication number CN111210145A discloses a method for analyzing the transient stability of a power system based on coupling evaluation indexes, which includes: establishing a spatial model for the state of the power system, decoupling the spatial model by using the coupling evaluation indexes and a nonlinear decoupling method, and approximately converting the original high-order nonlinear system into a series of decoupled first-order quadratic and second-order quadratic systems; then, the attraction domain of each state variable is determined by using a low-order quadratic system analysis tool, and the transient stability of the power system is reflected by the relationship between the numerical value of the state variable and the obtained attraction domain. However, the foregoing prior art scheme has the following technical defects: 1. Coupling effect is missed in the decoupling process: the high-order system is decoupled into low-order quadratic systems based on the coupling evaluation indexes, which may ignore part of the dynamic coupling, resulting in incomplete stability analysis; 2. Limitations of the low-order model: the decoupled first-order / second-order system is difficult to completely reflect the complex interaction of the original high-dimensional system, especially in the strong nonlinear scenario; 3. Insufficient conservatism: the characteristic values of the linear part are used to screen the coupling pairs, which may result in high conservatism of the results and cannot accurately guide the parameter design.

[0006] The prior art patent application document with the publication number CN118863262A, "DC micro-grid large signal stability analysis method based on DNN Lyapunov function", the prior method includes: establishing an equivalent reduced circuit mathematical model, and building a simulation model in MATLAB / Simulink; constructing the main structure of the Lyapunov function through DNN; constructing a loss function that guides the DNN to optimize in the direction of the Lyapunov stability condition; determining the structure parameters of the DNN; performing iterative training, stopping iterative training when the training process is less than the loss function set value epsilon, giving the Lyapunov function analytical expression according to the training parameters, outputting the Lyapunov three-dimensional graph and performing strict verification, and then combining the LaSalle invariant set principle to estimate the attraction domain, and evaluating the DNN through the loss function convergence curve. However, the foregoing prior art has the following technical defects: 1. High data dependency: based on DNN to construct Lyapunov function needs to rely on a large amount of training data, and the loss function convergence is difficult to guarantee strictly; 2. Poor adaptability to topology switching: the topology switching scenario of the micro-grid is not mentioned, and the dynamic adjustment capability is lacking; 3. Attraction domain boundary is fuzzy: the traditional Lyapunov method estimates the attraction domain by relying on the LaSalle invariant set principle, and the boundary is an approximate result, which is insufficient in accuracy.

[0007] In summary, the prior art has the technical problems of high dependence on explicit mathematical derivation or approximate model, inaccurate error accumulation stability evaluation, and incomplete stability evaluation. SUMMARY

[0008] The technical problem to be solved by the present application is how to solve the technical problems of high dependence on explicit mathematical derivation or approximate model, inaccurate error accumulation stability evaluation, and incomplete stability evaluation in the prior art.

[0009] The present application solves the above technical problems by adopting the following technical solutions: a micro-grid group topology switching stability analysis method includes: S1, modeling the micro-grid group system, and introducing the Zubov equation; S2, solving the Zubov equation of the micro-grid group system by using a multi-layer feedforward neural network to obtain a Lyapunov function, and designing and using a loss function to determine the boundary condition loss , Zubov equation residual , to minimize the residual between the network output and the Zubov equation, and to train and optimize the multi-layer feedforward neural network; S3, calculating the Lyapunov function of the micro-grid group system under different topology switching the value of the system, whether the micro-grid group system is stable under the whether the micro-grid group system is stable under the S4, the stability of the micro-grid group system under different topological configurations is analyzed to control the topological switching operation of the micro-grid group system.

[0010] The loss function capable of meeting the Zubov condition and the boundary condition is designed.

[0011] Compared with the prior art which can only provide an approximate boundary of the attraction domain, the exact boundary of the system attraction domain can be accurately calculated, and the accuracy and stability of the complex system can be evaluated.

[0012] The Zubov equation can handle high-dimensional micro-grid group systems involving a large number of subsystems by combining the powerful representation ability of the neural network.

[0013] When the system faces topology switching, the solution of the Zubov equation is further verified to ensure that the system can still maintain a stable operating state under the new topological configuration after switching.

[0014] In a more specific technical solution, in S1, the dynamic equation of the first sub-micro-grid in the micro-grid group is defined by using the following logic:

[0015]

[0016] In the formula, is the angle deviation, is the voltage amplitude, is the derivative of the angle deviation, is the derivative of the voltage amplitude, and is the tracking time constant, and are the steady-state injections of active power and reactive power, and represent the active power and reactive power injections, and are the drop gains; The Zubov equation is expressed by using the following logic:

[0017] In the formula, is the state vector of the system, is the Lyapunov function, is a given positive definite function related to the system state, in the Zubov equation, the approaching information of is used to determine the stable state of the system.

[0018] The application can directly deal with nonlinearity: by combining the Zubov equation with the neural network, avoiding piecewise linearization, and preserving the complete nonlinear characteristics of the original system.

[0019] The application ensures accuracy through theoretical application: by combining the mathematical framework of the Zubov equation, the correctness of the solution is ensured, the conservativeness is reduced, and reliable parameter design basis is provided.

[0020] In a more specific technical solution, in S2, a multi-layer feedforward neural network is constructed; wherein the input of the multi-layer feedforward neural network is the state vector of the microgrid group system :

[0021] wherein the state of the microgrid group system includes: angle deviation and voltage amplitude; The intermediate layer of the multi-layer feedforward neural network includes: no less than 2 hidden layers, and each hidden layer includes no less than 2 neurons; wherein the output of each layer of hidden layer is expressed by using the following logic:

[0022] wherein, is the output of the layer, , is a weight matrix, is a bias vector, is an activation function , denotes the total number of layers in the neural network.

[0023] In a more specific technical solution, the Lyapunov function output by the multi-layer feedforward neural network is :

[0024] wherein, is the parameter of the network, including weight, bias.

[0025] In a more specific technical solution, a loss function is designed and used to minimize the residual error between the network output and the Zubov equation; The derivative of the multi-layer feedforward neural network with respect to time is calculated:

[0026] wherein, is the derivative of the network output with respect to time Next The state of each sampling point It is a neural network Output under input; The state of the sampled microgrid group system is used to calculate the physical information loss function; among other things, the residuals of the Zuboff equation are calculated. For each sampling point in the microgrid cluster system Calculate the physical information loss function:

[0027] In the formula, This refers to the number of sampling points selected. It is a stability function learned through a neural network; Determine boundary condition loss For boundary points, it is assumed that... The value is 1 at the boundary of the attraction domain. Using the following logic, the boundary condition loss function of the microgrid system is defined:

[0028] In the formula, This refers to the number of boundary points selected. According to the residuals of the Zuboff equation Boundary condition loss Find the total loss function :

[0029] In the formula, These are weighting coefficients; Minimize the total loss function The parameters of a multilayer feedforward neural network are updated using the gradient descent optimization algorithm. ; Among them, for the total loss function By taking the derivative, we can obtain the parameter gradient. :

[0030] Update parameters using gradient descent optimization algorithm :

[0031] In the formula, It is the learning rate, which determines the step size for each parameter update; Through iterative operations, the parameters are optimized. Convergence, making It satisfies the Zuboff equation and boundary conditions.

[0032] The stability analysis of the system can be ensured by adopting the construction of Lyapunov function, so as to provide a theoretical basis for the dynamic behavior of the system. The traditional stability analysis method often depends on explicit mathematical derivation or approximate model, while the neural network can automatically extract rules from data through learning, and provides a more flexible and efficient solving method.

[0033] The application has adaptive learning ability: by using the flexibility of neural network, the complex mapping relationship is automatically extracted from data, and the strong nonlinear scene is adapted.

[0034] By combining the Lyapunov equation and the neural network, the application can accurately calculate the exact boundary of the system attractor, and avoid the error accumulation problem in the traditional method. The accurate boundary of the attractor not only helps to more comprehensively evaluate the stability of the system, but also can provide more reliable prediction for the actual operation of the microgrid system.

[0035] The boundary calculation of the application is more accurate: the loss function meeting the Lyapunov condition is designed to ensure the strict mathematical derivation of the attractor boundary and avoid the approximation error.

[0036] The application can strictly calculate the boundary: the attractor boundary is accurately solved by combining the Lyapunov equation, and the estimation error of the traditional Lyapunov method is avoided.

[0037] The application can perform theory-driven training: the loss function design is constrained by the Lyapunov equation to reduce the dependence on data and ensure that the network convergence direction meets the stability condition.

[0038] In a more specific technical solution, in S3, the derivative of Lyapunov function with respect to time is calculated :

[0039] In the formula, is the Lyapunov function with respect to time , and is the change rate of the system state .

[0040] The application can dynamically verify the topology switching: when the system topology changes, the stability under the new configuration is quickly verified based on the Lyapunov equation.

[0041] The sign information of the derivative is judged to analyze the stability of the microgrid system; The value of Lyapunov function is analyzed to determine the attractor , wherein the region with the value of Lyapunov function less than 1 is determined as the attractor : .

[0043] In a more specific technical solution, in S4, the topology switching of the microgrid group will change the connection mode between subsystems, according to the topology change information, adjusting the dynamic equation of the microgrid group system, using the following logic, expressing the dynamics model of the microgrid group system:

[0044] In the formula, is the state vector of the system, indicating the voltage and angle of all subsystems of the system, is the nonlinear dynamics function of the system, which depends on the state and topology configuration of the system , is the topology configuration of the system at time , indicating the connection mode of the microgrid system subsystem.

[0045] In a more specific technical solution, in S4, the derivative of Lyapunov function is calculated by using the neural network Zubov equation, and the stability of the system under different topology configurations is analyzed; wherein, at time , the topology configuration is selected, according to the Zubov equation, the Lyapunov function satisfies the following relationship:

[0046] In the formula, is the Lyapunov function of the system under the topology configuration , is the derivative of time ; so that the change rate of the Lyapunov function satisfies the following condition, to stabilize the microgrid group system:

[0047] Among them, is the attractor domain of the system under the topology configuration , indicating the area where the system can be stable. And all states in the attractor domain satisfy , indicating that the state will tend to the equilibrium point.

[0048] The attractor domain under different topology configurations is calculated, so that when the topology switching of the microgrid group system occurs, the running state after the switching will fall into the attractor domain, realizing the topology safe reconstruction.

[0049] The application can retain complete dynamic characteristics: the high-dimensional nonlinearity of the original system is directly modeled by the neural network, without decoupling, and the key coupling effects are avoided.

[0050] Before and after the topology switching, the first condition, the second condition and the stability condition are ensured to be met.

[0051] The neural network used in the application can learn deep features in the data and perform efficient calculation in large-scale systems.

[0052] The application can efficiently solve high-dimensional systems: the function approximation capability of the neural network is used to directly learn high-dimensional dynamic relationships from data, reducing the computational complexity.

[0053] In a more specific technical solution, in the first condition, the state before switching falls within the attractor domain of the topology configuration before switching :

[0054] In the second condition, the state after switching falls within the attractor domain of the topology configuration after switching :

[0055] In the second condition and the stability condition, the state after switching is ensured to meet the stability condition: .

[0057] Since topology switching often introduces uncertainty in the system, the application can ensure stable transition of the system state through accurate attractor domain analysis, avoiding unstable dynamic behavior, thereby providing reliable stability guarantee for the practical application of microgrid clusters.

[0058] In a more specific technical solution, the microgrid cluster topology switching stability analysis system comprises: A microgrid cluster modeling module is used to model the microgrid cluster system and introduce the Zubov equation; A neural network training optimization module is used to solve the Zubov equation of the microgrid cluster system by using a multi-layer feedforward neural network, obtain a Lyapunov function, and design and use a loss function to determine the boundary condition loss , the Zubov equation residual ​​The multi-layer feedforward neural network is trained and optimized by seeking and minimizing the residual error of the network output and the Zubov equation. The micro-grid cluster attraction domain determination module is used to calculate the value of the Lyapunov function of the micro-grid cluster system under different topology switching, judge whether the micro-grid cluster system is stable under different topology switching, calculate the boundary of the attraction domain, and determine the attraction domain of the micro-grid cluster system. The topology switching module is used to analyze the stability of the micro-grid cluster system under different topology configurations, so as to control the topology switching operation of the micro-grid cluster system. The topology switching module is connected with the micro-grid cluster attraction domain determination module. The topology switching module is used to analyze the stability of the micro-grid cluster system under different topology configurations, so as to control the topology switching operation of the micro-grid cluster system.

[0059] Compared with the prior art, the present application has the following advantages: The loss function designed in the present application can satisfy the Zubov condition and the boundary condition. Through the training of the neural network, the Zubov equation of the system can be effectively solved, and the Lyapunov function can be derived.

[0060] Compared with the prior art which can only provide an approximate boundary of the attraction domain, the present application can accurately calculate the exact boundary of the system attraction domain, and can accurately and stably evaluate the complex system.

[0061] The present application can handle high-dimensional micro-grid cluster systems involving a large number of subsystems by combining the powerful representation ability of the neural network.

[0062] When the system faces topology switching, the present application further verifies the solution of the Zubov equation to ensure that the system can still maintain a stable operating state under the new topology configuration after switching.

[0063] The construction of the Lyapunov function can ensure the stability analysis of the system, thereby providing a theoretical basis for the dynamic behavior of the system. Traditional stability analysis methods often rely on explicit mathematical derivation or approximate models, while neural networks can automatically extract rules from data through learning, providing a more flexible and efficient solution.

[0064] By combining the Zubov equation and the neural network, the present application can accurately calculate the exact boundary of the system attraction domain, avoiding the error accumulation problem in traditional methods. The accurate boundary of the attraction domain not only helps to more comprehensively evaluate the stability of the system, but also provides more reliable prediction for the actual operation of the micro-grid system.

[0065] Compared with the traditional method, the neural network can learn deep features in data and perform efficient calculation in a large system.

[0066] Since the topology switching often introduces uncertainty of the system, the attractor domain analysis can ensure stable transition of the system state and avoid unstable dynamic behavior, thereby providing reliable stability guarantee for practical application of the microgrid group.

[0067] The present application provides a new method for large signal stability analysis of the microgrid, which is high-precision, low-conservatism and strong adaptability, especially suitable for high-dimensional nonlinear systems and dynamic topology scenarios, and provides a reliable tool for stability evaluation and parameter design in practical engineering.

[0068] The present application solves the technical problems of high dependence on explicit mathematical derivation or approximate model, inaccurate error accumulation stability evaluation and incomplete stability evaluation in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0069] Figure 1 The figure is a basic step diagram of the microgrid group topology switching stability analysis method of the present application. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below, obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0071] Embodiment 1 As shown in the figure, the present application provides a microgrid group topology switching stability analysis method. The method comprises the following basic steps: Figure 1 S1, modeling the microgrid system and introducing the Zubov equation; In this embodiment, the dynamic equation of the first sub-microgrid in the microgrid group can be written as:

[0072] wherein, is the angle deviation, is the voltage amplitude, is the derivative of the angle deviation, is the derivative of the voltage amplitude, and is the tracking time constant,​​ P and Q are the steady-state injections of active and reactive power, respectively, and P and Q represent the active and reactive power injections, respectively, and is the decrease gain.

[0073] In this embodiment, the general form of the Zubov equation is:

[0074] where, x is the state vector of the system, which can contain the state variables of multiple microgrid subsystems, V(x) is the Lyapunov function, which is used to describe the stability of the system. V(x) is a given positive definite function related to the state of the system, which is used to represent stability. In this equation, when approaches 1, the system is close to an unstable state; when approaches 0, the system is in a stable state.

[0075] S2, solving the Zubov equation in the microgrid system using a neural network to obtain the Lyapunov function; In this embodiment, a multi-layer feedforward neural network is constructed. The specific structure of the neural network is as follows: the input of the neural network is the state vector of the system, and the states of the microgrid group system include the angle deviation and voltage amplitude:

[0076] The intermediate layer of the neural network includes multiple hidden layers, and each hidden layer contains multiple neurons. The output of each layer is given by the following formula:

[0077] where, is the output of the th layer, is the weight matrix, is the bias vector, is the activation function , and usually represents the total number of layers in the neural network.

[0078] The Lyapunov function output by the neural network is represented as follows:

[0079] where, is the parameter of the network, including all weights and biases.

[0080] In this embodiment, the loss function is designed as The training objective of the neural network is to minimize the residual error between the network output and the Zubov equation through the loss function.

[0081] First, the derivative of the neural network with respect to time is calculated:

[0082] where is the state of the th sampling point at time , and is the output of the neural network under the input.

[0083] Then the physical information loss function is calculated by sampling the state of the system. Specifically, in the loss function design operation, the residual error of the Zubov equation For each sampling point , the physical information loss is calculated as:

[0084] where is the number of selected sampling points, is the stability function learned by the neural network.

[0085] The boundary condition loss is determined: for the boundary points, assume that the value at the attractor domain boundary is 1, and the boundary condition loss function of the system is:

[0086] where is the number of selected boundary points.

[0087] The total loss function is the weighted sum of the above loss:

[0088] where is the weight coefficient, used to adjust the relative importance of each loss term in the total loss function.

[0089] In this embodiment, the training and optimization of the neural network. The training process of the neural network involves minimizing the total loss function First, the parameters of the neural network are updated by the gradient descent optimization algorithm . The gradient of the parameter is obtained by taking the derivative of the loss function :

[0090] Then the parameters are updated by gradient descent method :

[0091] where, is the learning rate, which determines the step size of each parameter update.

[0092] Finally, multiple iterations are adopted, and the network parameters are gradually converged, so that satisfy the Zubov equation and the boundary conditions.

[0093] S3, analyze the stability by Lyapunov function and calculate the attractor domain of the system; In this embodiment, after the neural network training is completed, the Lyapunov function output by the network is used to verify the stability.

[0094] First, the output of the neural network is obtained .

[0095] Then calculate , that is the derivative of time:

[0096] where, is the gradient of Lyapunov function with respect to time , and is the rate of change of the system state .

[0097] Then analyze the stability by judging the sign of : if , the system is stable; if , the system is unstable.

[0098] Finally, the value of is analyzed to determine the attractor domain , where the attractor domain is the area with a value less than 1:

[0099] By calculating the value of the micro-grid group under different topological switching, we can judge whether the system can maintain stability under and calculate the boundary of the attractor domain.

[0100] S4, further analyze the stability of the micro-grid group system under different topological configurations to ensure the stability of the system topological switching.​

[0101] In this embodiment, the topology switching of the microgrid cluster will change the connection mode between subsystems, so the dynamic equation of the system will change with the change of the topology. The dynamic model of the system can be expressed as:

[0102] wherein, is the state vector of the system, representing the voltage and angle of all subsystems of the system, is the nonlinear dynamics function of the system, which depends on the state and topology configuration of the system , is the topology configuration of the system at time , representing the connection mode of the subsystems of the microgrid system.

[0103] In this embodiment, the derivative of the Lyapunov function is calculated by the neural network Zhabotinsky equation to analyze the stability of the system under different topology configurations. At time , the topology configuration is selected, then according to the Zhabotinsky equation, the Lyapunov function satisfies the following relationship:

[0104] wherein, is the Lyapunov function of the system under the topology configuration , is the derivative thereof with respect to time .

[0105] Then, to ensure the stability of the system after topology switching under all states, the change rate of the Lyapunov function satisfies the following condition:

[0106] wherein is the attractor domain of the system under the topology configuration , representing the area in which the system can be stable. And all states in the attractor domain satisfy , indicating that the state will tend to the equilibrium point.

[0107] In this embodiment, through the above analysis, the attractor domain under different topology configurations is calculated. When the topology of the system is switched, it is necessary to ensure that the running state after the switching falls within the stable domain, so as to ensure the topology switching stability of the microgrid cluster system and realize the topology safe reconstruction.

[0108] Specifically, before and after the topology switching, the following conditions are ensured: The first condition, the state before switching Attractor region of the pre-switch topology configuration , i.e.:

[0109] Second condition, post-switch state Attractor region of the post-switch topology configuration , i.e.:

[0110] Second condition, stability condition: ensure that the post-switch state satisfies the stability condition, i.e.:

[0111] This ensures the stability of the system after topology switching, ensuring smooth transition of the system state and stable operation under the new topology configuration.

[0112] In summary, the present application designs a loss function that can satisfy the Zubov condition and the boundary condition. Through the training of the neural network, the Zubov equation of the system can be effectively solved, and the Lyapunov function can be derived therefrom.

[0113] Compared with the prior art which can only provide an approximate boundary of the attractor region, the present application can accurately calculate the exact boundary of the attractor region of the system, and can accurately and stably evaluate the complex system.

[0114] The present application combines the powerful representation ability of the neural network, so that the Zubov equation can handle high-dimensional microgrid system involving a large number of subsystems.

[0115] When the system faces topology switching, the present application further verifies the solution of the Zubov equation to ensure that the system can still maintain a stable operating state under the new topology configuration after switching.

[0116] The present application uses the construction of the Lyapunov function to ensure the stability analysis of the system, thereby providing a theoretical basis for the dynamic behavior of the system. Traditional stability analysis methods often rely on explicit mathematical derivation or approximate models, while neural networks can automatically extract rules from data through learning, providing a more flexible and efficient solution.

[0117] The present application can accurately calculate the exact boundary of the attractor region of the system by combining the Zubov equation with the neural network, thereby avoiding the error accumulation problem in traditional methods. The accurate boundary of the attractor region not only helps to more comprehensively evaluate the stability of the system, but also provides more reliable prediction for the actual operation of the microgrid system.

[0118] ​​Compared with the traditional method, the neural network can learn deep features in data and perform efficient calculation in a large-scale system.

[0119] Since the topology switching often introduces uncertainty of the system, the accurate attraction domain analysis can ensure stable transition of the system state and avoid unstable dynamic behavior, thereby providing reliable stability guarantee for practical application of the micro-grid group.

[0120] The present application solves the technical problems of the prior art, such as high dependence on explicit mathematical derivation or approximate model, inaccurate error accumulation stability evaluation, and incomplete stability evaluation.

[0121] The above examples are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for analyzing the stability of microgrid topology switching, characterized in that, The method includes: S1. Model the microgrid group system and introduce the Zuboff equation; S2. Solve the Zubov equation of the microgrid group system using a multilayer feedforward neural network to obtain the Lyapunov function, and design and utilize the loss function. Determine the boundary condition loss Zuboff equation residuals The residual between the network output and the Zubov equation is calculated and minimized, and the multilayer feedforward neural network is trained and optimized. S3. Calculate the Lyapunov function of the microgrid group system under different topology switching conditions. The value is used to determine the microgrid group system in [the context of the system]. To determine whether the microgrid cluster system remains stable, calculate the boundary of the attraction domain and determine the attraction domain of the microgrid cluster system. S4. Perform stability analysis on the microgrid system under different topology configurations to control the microgrid system to perform topology switching operations.

2. The microgrid group topology switching stability analysis method according to claim 1, characterized in that, In S1, the following logic is used to define the first [unit / group] in the microgrid group. Dynamic equations of individual microgrids: In the formula, Due to angular deviation, Voltage amplitude, The derivative of the angular deviation, The derivative of the voltage amplitude, and It is the tracking time constant. These are the steady-state injections of active power and reactive power, respectively. and These represent active power injection and reactive power injection, respectively. and It is a decrease in gain; The Zuboff equation can be expressed using the following logic: In the formula, It is the system's state vector. It is a Lyapunov function. It is a given positive definite function related to the system state, in the Zuboff equation, according to The approach information determines the stable state of the system.

3. The microgrid group topology switching stability analysis method according to claim 1, characterized in that, In step S2, the multi-layer feedforward neural network is constructed; The input to the multilayer feedforward neural network is the state vector of the microgrid swarm system. : The states of the microgrid cluster system include: angle deviation and voltage amplitude; The intermediate layers of the multilayer feedforward neural network include at least two hidden layers, each of which includes at least two neurons; wherein the output of each hidden layer is expressed using the following logic: In the formula, It is the first The output of the layer, , It is a weight matrix. It is a bias vector. It is an activation function , This represents the total number of layers in a neural network.

4. The microgrid group topology switching stability analysis method according to claim 3, characterized in that, The Lyapunov function is output through the multilayer feedforward neural network. : In the formula, These are the network parameters, including weights and biases.

5. The microgrid group topology switching stability analysis method according to claim 3, characterized in that, Design and utilize loss functions Minimize the residual between the network output and the Zubov equation; Calculate the time derivative of the multilayer feedforward neural network: In the formula, It is in time Next The state of each sampling point It is a neural network Output under input; The state of the microgrid cluster system is sampled, and the physical information loss function is calculated; wherein, the Zuboff equation residual is calculated. For each sampling point in the microgrid cluster system Calculate the physical information loss function: In the formula, This refers to the number of sampling points selected. It is a stability function learned through a neural network; Determine the boundary condition loss For boundary points, it is assumed that... The value is set to 1 at the boundary of the attraction domain. The boundary condition loss function of the microgrid group system is defined using the following logic: In the formula, This refers to the number of boundary points selected. According to the residual of the Zubov equation The boundary condition loss Find the total loss function : In the formula These are weighting coefficients; Minimize the total loss function The parameters of a multilayer feedforward neural network are updated using the gradient descent optimization algorithm. ; Among them, for the total loss function By taking the derivative, we can obtain the parameter gradient. : Update parameters using gradient descent optimization algorithm : In the formula, It is the learning rate, which determines the step size for each parameter update; Through iterative operations, the parameters are optimized. Convergence, making The Zuboff equation and boundary conditions are satisfied.

6. The microgrid group topology switching stability analysis method according to claim 1, characterized in that, In step S3, the derivative of the Lyapunov function with respect to time is calculated. : In the formula, It is a Lyapunov function Regarding time gradient, System status The rate of change. Determine the derivative The symbol information is used to analyze the stability of the microgrid system; Analysis of the Lyapunov function The value determines the attraction domain. Among them, the Lyapunov function The region with a value less than 1 is defined as the attraction domain. : 。 7. The microgrid group topology switching stability analysis method according to claim 1, characterized in that, In step S4, the topology switching of the microgrid group will change the connection mode between subsystems. Based on the topology change information, the dynamic equations of the microgrid group system are adjusted, and the dynamic model of the microgrid group system is expressed using the following logic: In the formula, It is the system's state vector, representing the voltage and angle of all subsystems of the system. It is a nonlinear dynamic function of the system, which depends on the system's state and topology. , Is the system in time The topology configuration at any given time indicates the connection method of the subsystems in the microgrid system.

8. The microgrid group topology switching stability analysis method according to claim 1, characterized in that, In step S4, the Lyapunov function is calculated using the Zubov equation of the neural network. The derivative is used to analyze the stability of the system under different topological configurations; where, at time... Selected topology configuration According to the Zubov equation, the Lyapunov function The following relationship must be satisfied: In the formula, It is the system in topology configuration Lyapunov function under the following conditions It is about time The derivative; The rate of change of the Lyapunov function must satisfy the following condition to stabilize the microgrid cluster system: in It is the system in topology configuration The region of attraction is defined as the region where the system can be stable. Furthermore, all states within the region of attraction satisfy... This indicates that the state will tend towards an equilibrium point. Find different topology configurations The attraction domain mentioned below ensures that when the topology switch occurs in the microgrid group system, the operating state after the switch will fall into the attraction domain, thereby achieving topology safe reconstruction. Before and after the topology switch, ensure that the first condition, the second condition, and the stability condition are met.

9. The microgrid group topology switching stability analysis method according to claim 8, characterized in that, In the first condition, the state before the switch. Falling on topology configuration before handover attraction domain Inside: In the second condition, the state after the switch Falling on the topology configuration after the switch attraction domain Inside: In the second condition and the stability condition, ensure that the state after the switch satisfies the stability condition: 。 10. A microgrid group topology switching stability analysis system, characterized in that, The system includes: The microgrid group modeling module is used to model microgrid group systems, introducing the Zuboff equation; The neural network training and optimization module is used to solve the Zubov equation of the micro-network swarm system using a multilayer feedforward neural network, obtain the Lyapunov function, and design and utilize the loss function. Determine the boundary condition loss Zuboff equation residuals The residual between the network output and the Zubov equation is calculated and minimized, and the multilayer feedforward neural network is trained and optimized. The neural network training and optimization module is connected to the micro-network group modeling module. The microgrid group attraction domain determination module is used to calculate the Lyapunov function of the microgrid group system under different topology switching conditions. The value is used to determine the microgrid group system in [the context of the system]. To determine whether the system remains stable, the boundary of the attraction domain is calculated, and the attraction domain of the micro-network group system is determined. The micro-network group attraction domain determination module is connected to the neural network training and optimization module. The topology switching module is used to perform stability analysis on the microgrid system under different topology configurations in order to control the microgrid system to perform topology switching operations. The topology switching module is connected to the microgrid attraction domain determination module.

Images