Power grid voltage fluctuation suppression method and system based on dynamic voltage regulation

Abstract

The invention discloses a power grid voltage fluctuation suppression method and system based on dynamic voltage regulation, and the method comprises the steps: synchronously collecting a three-phase voltage signal through a sensor array, generating a multi-dimensional feature vector through Park transformation, and extracting a time-frequency joint feature through a mixed multi-scale decomposition algorithm, and inputting a model fusing fuzzy reasoning and a space-time diagram neural network, generating a cooperative adjustment strategy in combination with power grid topology and equipment constraints, and optimizing model parameters through a closed-loop correction mechanism driven by a dynamic forgetting factor. The system comprises a signal acquisition module, a characteristic decomposition module, an intelligent decision module and a closed loop correction module. According to the method, high-precision suppression of power grid voltage fluctuation is realized, the multi-node cooperative adjustment capability is improved, the method can be adaptive to power grid operation state changes, technical feasibility and economic rationality are considered, and an effective scheme is provided for modern power grid voltage stable control.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid voltage regulation, and in particular to a method for suppressing grid voltage fluctuations based on dynamic voltage regulation. Background Art

[0002] With the increasing penetration of renewable energy and power electronics into the power grid, voltage fluctuations are becoming increasingly problematic, leading to degraded power quality and increased equipment operational risks. Existing methods for mitigating grid voltage fluctuations primarily rely on static VAR compensation devices or fixed-rule transformer regulation. These methods suffer from issues such as delayed response, insufficient harmonic suppression, and voltage coupling caused by local regulation.

[0003] In recent years, voltage fluctuation control methods based on intelligent algorithms have gradually been applied, such as predictive control based on neural networks or fuzzy logic control. However, these methods lack multi-scale feature analysis capabilities and do not consider grid topology constraints, resulting in poor global coordination of regulation strategies, ignoring the topological correlation of the grid and the dynamic response characteristics of equipment. Control deviations are prone to accumulation during multi-node coordinated regulation, resulting in unsatisfactory suppression effects. Summary of the Invention

[0004] The main purpose of this application is to provide a method and system for suppressing grid voltage fluctuations based on dynamic voltage regulation, aiming to solve the problems of slow response speed, poor coordination and weak adaptability of grid voltage fluctuation suppression in the existing technology.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for suppressing grid voltage fluctuations based on dynamic voltage regulation, comprising the following steps: S1. Synchronously collect the three-phase voltage signals of the grid nodes through the sensor array deployed at the grid nodes, and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transform; S2. Using the multidimensional feature vector as input, a hybrid multi-scale decomposition algorithm is used to separate the fundamental wave and harmonics, and to extract the time-frequency joint features of the voltage fluctuation amplitude, frequency deviation, and phase mutation rate; S3. Input the extracted time-frequency joint features into an intelligent decision-making model that integrates fuzzy reasoning and spatiotemporal graph neural network. Combined with the grid topology constraints and the dynamic response characteristics of the equipment, it generates target voltage optimization, adaptive reactive power compensation, and transformer ratio coordinated adjustment strategies. S4. Design a closed-loop correction mechanism driven by a dynamic forgetting factor, iteratively optimize the parameters of the intelligent decision-making model through real-time error feedback, and achieve adaptive suppression of voltage fluctuations.

[0006] Optionally, step S1 specifically includes: Three synchronous voltage sensors are configured at each grid node, corresponding to phases A, B, and C of the three-phase AC power; Collecting three-phase voltage signals by the synchronous voltage sensor; The collected three-phase voltage signal is input into the Parker transformation module, and the real-time phase angle and frequency of the power grid are estimated in real time through the phase-locked loop; The Parker transformation module converts the three-phase voltage signal into a voltage component in a two-phase rotating coordinate system, and generates a multidimensional feature vector in combination with the real-time frequency and phase angle. The multidimensional feature vector includes the two-phase voltage components, the real-time frequency and the real-time phase angle.

[0007] Optionally, the Parker transformation module converts the three-phase voltage signal into a voltage component in a two-phase rotating coordinate system, and generates a multi-dimensional feature vector in combination with the real-time frequency and phase angle, specifically including: The three-phase voltage signal 、 、 Input to the Parker transform module, the transformation matrix of the Parker transform module is defined as:

[0008] in, , , is the initial phase, The real-time frequency of the power grid is estimated in real time through a phase-locked loop; After conversion, the voltage component in the dq coordinate system is obtained ; Combined with the real-time frequency of the power grid and phase angle , generate multidimensional feature vectors .

[0009] Optionally, the hybrid multi-scale decomposition algorithm in step S2 includes: For the voltage components in the multidimensional feature vector Perform 3-layer wavelet packet decomposition to obtain 8 frequency band components , where the fourth frequency band component corresponds to the fundamental component; Perform fast Fourier transform on the fundamental component to calculate the fundamental amplitude 、 , fundamental frequency and fundamental phase ; The harmonic component is defined as the difference between the original signal and the fundamental component: ,

[0010] The voltage fluctuation amplitude is calculated as the norm of the harmonic components:

[0011] The frequency deviation is calculated as:

[0012] in, is the rated frequency of the grid; The phase mutation rate is calculated by numerical differentiation:

[0013] in, is the sampling time interval.

[0014] Optionally, step S3 specifically includes: The voltage fluctuation amplitude in the time-frequency joint feature , frequency deviation Phase mutation rate Map them to the corresponding fuzzy levels respectively and calculate the membership of each input variable to different fuzzy levels; Based on the preset fuzzy rule base, the membership of the input variables is logically combined, and the accurate reactive compensation coefficient is output after defuzzification by the centroid method. , voltage optimization coefficient and transformer ratio adjustment coefficient ; Build a power grid topology map , where the vertex set Represents a grid node, edge set Represents the admittance relationship between nodes. Node characteristics include voltage fluctuation amplitude, frequency deviation, and phase mutation rate. Power grid topology map through spatiotemporal graph neural network Perform multi-layer graph convolution operations and update the dynamic characteristics of each node through adjacent node information transmission to extract the global spatiotemporal correlation characteristics of the power grid; By integrating fuzzy inference output and node features after graph convolution, a collaborative regulation strategy is generated with the goal of minimizing voltage fluctuation, frequency deviation and equipment regulation cost.

[0015] Optionally, the node state update equation of the spatiotemporal graph neural network is:

[0016] in, is the adjacency matrix with self-loops added, is the adjacency matrix of the admittance connections of the grid nodes, is the identity matrix; for The degree matrix of ; For the Layer node feature matrix, is the number of nodes, is the feature dimension, and the initial features include 、 and ; is the learnable weight matrix; is the ReLU activation function.

[0017] Optionally, the step of generating the collaborative adjustment strategy includes: Construct the optimization objective function:

[0018] in, 、 、 The preset weight coefficients control the optimization priorities of voltage fluctuation, frequency deviation and equipment regulation cost respectively; To regulate the number of devices; Equipment adjustment cost Defined as:

[0019] in, 、 、 is the adjustment coefficient of the device at the previous moment; Node features based on graph convolution and fuzzy inference output ,The optimal adjustment coefficient that satisfies the constraints is solved by the gradient descent method, and the coordinated adjustment instructions of each node are generated.

[0020] Optionally, the coordinated regulation strategy processes the grid topology constraints by: Inter-node voltage coupling constraint: For adjacent nodes and , its voltage regulation satisfies:

[0021] in, is the inter-node admittance, is the admittance phase angle; Equipment dynamic response characteristic constraints: Reactive power compensation equipment response time constraints: ; Transformer ratio adjustment minimum interval constraint: ; Based on the power grid topology The adjacency matrix , the adjustment influence weights of adjacent nodes are calculated through the graph attention mechanism:

[0022] in, is the attention weight vector, For nodes The set of adjacent nodes of For the Layer Node The eigenvector of .

[0023] Optionally, step S4 specifically includes: Defining the dynamic forgetting factor , and its update formula is:

[0024] in, is the initial forgetting factor, is the forgetting rate parameter; for The voltage prediction error at time t, is the sliding window length; Iterative optimization of model parameters based on forgetting factor:

[0025] in, is the parameter set of the intelligent decision-making model; is the learning rate; is the gradient of the objective function with respect to the parameters; Error feedback process of closed-loop correction mechanism: Real-time acquisition of the actual voltage signal after adjustment and calculation of the prediction error ; like If it is greater than the preset threshold, the forgetting factor update and parameter iteration are triggered, otherwise the current parameters are maintained.

[0026] In a second aspect, the present application provides a grid voltage fluctuation suppression system based on dynamic voltage regulation, comprising: The signal acquisition module is used to synchronously collect three-phase voltage signals through a sensor array deployed at the grid nodes and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transform; A feature decomposition module is used to separate the fundamental wave and harmonics of the multidimensional feature vector using a hybrid multi-scale decomposition algorithm, and to extract the time-frequency joint features of the voltage fluctuation amplitude, frequency deviation and phase mutation rate; An intelligent decision-making module is used to input the time-frequency joint features into a model that integrates fuzzy reasoning and spatiotemporal graph neural networks, and generate target voltage optimization, adaptive reactive power compensation, and transformer ratio coordinated adjustment strategies based on grid topology constraints and equipment dynamic response characteristics; The closed-loop correction module is used to iteratively optimize the parameters of the intelligent decision-making model based on real-time error feedback through a mechanism driven by a dynamic forgetting factor, thereby achieving adaptive suppression of voltage fluctuations.

[0027] Through the above technical scheme, the beneficial effects of the present invention are as follows: the present application generates a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through a sensor array and Park transformation, comprehensively capturing the dynamic characteristics of voltage fluctuations, and effectively solving the problem of incomplete characterization of voltage fluctuations by traditional single feature analysis; using a hybrid multi-scale decomposition algorithm to achieve precise separation of fundamental and harmonics, quickly capture transient voltage fluctuation characteristics, and improve the analysis ability of time-frequency characteristics of voltage fluctuations; integrating fuzzy reasoning and spatiotemporal graph neural network intelligent decision-making model, combining grid topology constraints and equipment dynamic response characteristics to generate collaborative adjustment strategies, improve multi-node collaborative adjustment capabilities, reduce equipment adjustment costs, and ensure that the strategy is both technically feasible and economically reasonable; a closed-loop correction mechanism driven by a dynamic forgetting factor, iteratively optimizes model parameters based on real-time error feedback, so that the system can adapt to changes in grid operating status, continuously maintain voltage suppression accuracy, reduce equipment action frequency, and extend equipment service life, providing a reliable solution for voltage stability control of modern power grids with a high proportion of renewable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention, and the embodiments in the drawings do not constitute any limitation to the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 A flowchart of a method for suppressing grid voltage fluctuations based on dynamic voltage regulation provided in one embodiment of the present application; Figure 2 A schematic diagram of the process of step S1 provided in one embodiment of the present application; Figure 3 A schematic diagram of the process of step S3 provided in one embodiment of the present application; Figure 4 A schematic diagram of the process of step S4 provided in one embodiment of the present application; Figure 5A schematic diagram of the structure of a grid voltage fluctuation suppression system based on dynamic voltage regulation provided in one embodiment of the present application; Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application; The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0030] It should be understood that the specific embodiments described herein are merely for explaining the present application and are not intended to limit the present application. Rather, these embodiments are provided to make the present disclosure more thorough and complete and to fully convey the scope of the present disclosure to those skilled in the art.

[0031] The above and other technical contents, features and effects of the present invention are described below with reference to the attached Figure 1-6 The detailed description of the embodiments will clearly show that the structural contents mentioned in the following embodiments are all based on the accompanying drawings.

[0032] Various exemplary embodiments of the present invention will be described below with reference to the accompanying drawings.

[0033] In an exemplary embodiment, Figure 1 As shown, a method for suppressing grid voltage fluctuations based on dynamic voltage regulation is provided. The method is executed by a control terminal with computing capabilities. The control terminal can be specifically executed by a power system dedicated server, an edge computing unit and / or an embedded processing device. In an embodiment of the present invention, the method includes the following steps S1 to S4. Among them: Step S1: synchronously collect the three-phase voltage signals of the grid nodes through the sensor array deployed at the grid nodes, and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transformation.

[0034] In this step, an array of three synchronized voltage sensors is deployed at each node in the power grid. Each sensor corresponds to phase A, phase B, and phase C of the three-phase AC power. A unified clock synchronization mechanism is used to ensure consistency in sampling time. The sampling frequency of the sensor is not less than twice the highest harmonic frequency of the power grid (typical value ≥10kHz) to meet the requirements of the Nyquist sampling theorem and avoid signal aliasing. The collected three-phase voltage signal is represented as a time series 、 、 ,in is the sampling time.

[0035] The Park Transformation is a mathematical transformation method that converts an AC signal in a three-phase stationary coordinate system (ABC coordinate system) into a two-phase rotating coordinate system (dq coordinate system). Its core purpose is to convert time-varying AC signals into DC quantities through coordinate transformation, thereby simplifying dynamic analysis and control design. This embodiment decouples complex three-phase AC signals into easily processable DC quantities through coordinate transformation. Combined with the real-time frequency and phase angle estimated by the phase-locked loop, the dq-axis voltage components are fused with frequency and phase information to generate a multidimensional vector containing dynamic amplitude, phase, and frequency characteristics. This vector serves as the input for subsequent hybrid multi-scale decomposition, achieving separation of fundamental and harmonic characteristics and extraction of voltage fluctuation characteristics.

[0036] Step S2: using the multidimensional feature vector as input, a hybrid multi-scale decomposition algorithm is used to separate the fundamental wave and harmonics, and the time-frequency joint features of the voltage fluctuation amplitude, frequency deviation and phase mutation rate are extracted.

[0037] This step uses a hybrid multiscale decomposition algorithm to perform a hierarchical frequency-domain decomposition of the voltage components in the multidimensional feature vector generated in step S1. This algorithm combines wavelet packet decomposition with the fast Fourier transform (FFT) to perform time-frequency analysis of the multidimensional feature vector. This process achieves precise separation of the fundamental and harmonics, extracting key characteristics reflecting voltage quality from both the time and frequency dimensions, and providing a quantitative basis for subsequent intelligent decision-making.

[0038] In step S3, the extracted time-frequency joint features are input into an intelligent decision-making model that integrates fuzzy reasoning and spatiotemporal graph neural network, and combined with the grid topology constraints and equipment dynamic response characteristics to generate target voltage optimization, adaptive reactive power compensation and transformer ratio coordinated adjustment strategies.

[0039] This step inputs the combined time-frequency features into a model that integrates fuzzy reasoning and a spatiotemporal graph neural network. Fuzzy logic is used to map continuous features to discrete regulation strategies. Specifically, parameters such as voltage fluctuation amplitude and frequency deviation are first fuzzily graded. Regulation coefficients for reactive power compensation, voltage optimization, and transformer ratio are then inferred based on a pre-set rule base. Simultaneously, a grid topology is constructed, and multi-layer graph convolution operations within the spatiotemporal graph neural network are used to capture the spatiotemporal correlations between nodes. The voltage characteristics and topological constraints of adjacent nodes are then incorporated into the decision-making process. Ultimately, with the goal of minimizing voltage fluctuation, frequency deviation, and equipment regulation costs, a regulation strategy is generated that balances local optimization with global coordination, enabling coordinated control of grid equipment.

[0040] Step S4: designing a closed-loop correction mechanism driven by a dynamic forgetting factor, iteratively optimizing the parameters of the intelligent decision-making model through real-time error feedback, and realizing adaptive suppression of voltage fluctuations.

[0041] This step establishes a closed-loop correction system using a dynamic forgetting factor mechanism. The forgetting factor is dynamically adjusted based on the historical voltage forecast error. Specifically, the larger the error, the faster the forgetting factor decays, allowing the model to focus more on recent data. Based on the forgetting factor, the intelligent decision-making model parameters are iteratively optimized using gradients. When the error between the actual voltage collected in real time and the predicted voltage exceeds a threshold, a parameter update is triggered. The learning rate controls the iteration step size, ensuring that the model parameters adapt to dynamic grid changes. This mechanism empowers the system with self-learning capabilities, enabling real-time correction of decision-making deviations and continuously improving the robustness and accuracy of voltage fluctuation suppression.

[0042] In order to unify the amplitude, frequency and phase information of the grid voltage in one feature space, such as Figure 2 As shown, the above step S1 further includes the following steps S101 to S104 instead, specifically: Step S101: configure three synchronous voltage sensors at each grid node, corresponding to phase A, phase B, and phase C of the three-phase alternating current respectively; Step S102: The three-phase voltage signal is collected by the synchronous voltage sensor; the sensor realizes microsecond-level time synchronization of the entire network based on the IEEE1588 precision clock synchronization protocol, and collects the three-phase voltage signal at a sampling frequency of 12.8kHz. 、 and ; Step S103: input the collected three-phase voltage signal into the Parker transformation module, and estimate the real-time phase angle of the power grid in real time through the phase-locked loop (PLL) algorithm. and frequency ; Step S104: the Parker transformation module converts the three-phase voltage signal into a voltage component in a two-phase rotating coordinate system, and generates a multi-dimensional feature vector by combining the real-time frequency and phase angle. The multi-dimensional feature vector includes the two-phase voltage components , real-time frequency and real-time phase angle .

[0043] In one embodiment, the Parker transformation module converts the three-phase voltage signal into voltage components in a two-phase rotating coordinate system and generates a multi-dimensional feature vector based on the real-time frequency and phase angle, specifically including: The three-phase voltage signal 、 、 Input to the Parker transform module, the transformation matrix of the Parker transform module is defined as:

[0044] in, , , is the initial phase, The real-time frequency of the power grid is estimated in real time through the phase-locked loop algorithm; Specifically, the three-phase voltage signal With the transformation matrix Multiplying them, we get the voltage components in the two-phase rotating coordinate system (dq coordinate system):

[0045] in, is the direct-axis voltage component that rotates synchronously with the grid, is the quadrature-axis voltage component, and the two correspond to the voltage amplitude and phase information respectively.

[0046] Combined with the voltage component of the dq coordinate system and the real-time frequency of the power grid and phase angle , construct a multidimensional vector containing dynamic amplitude, phase and frequency characteristics: .

[0047] In order to jointly extract the voltage fluctuation amplitude, frequency deviation, and phase mutation rate from the time domain and frequency domain to comprehensively characterize the grid voltage quality problem, the hybrid multi-scale decomposition algorithm in step S2 above specifically includes: For the voltage components in the multidimensional feature vector Perform 3-layer wavelet packet decomposition to evenly divide the signal band into 8 frequency band components , where the fourth frequency band component It corresponds to the fundamental component of the power grid, and the other frequency bands correspond to the harmonic components.

[0048] Perform fast Fourier transform on the fundamental component to calculate the fundamental amplitude 、 , fundamental frequency and fundamental phase ,This process accurately obtains the fundamental wave characteristics through frequency domain analysis,,providing a reference benchmark for harmonic calculation.

[0049] The harmonic component is defined as the difference between the original signal and the fundamental component: ,

[0050] The voltage fluctuation amplitude is calculated as the norm of the harmonic components:

[0051] The frequency deviation is calculated as:

[0052] in, is the rated frequency of the grid; The phase mutation rate is calculated by numerical differentiation:

[0053] in, is the sampling time interval.

[0054] Through the above-mentioned hybrid multi-scale decomposition algorithm, it uses the multi-scale characteristics of wavelet packet decomposition to adapt to the non-stationary characteristics of the grid voltage, accurately separates the fundamental wave and harmonics, and quantifies the voltage fluctuation characteristics from the time and frequency dimensions, providing comprehensive and accurate feature input for subsequent intelligent decision-making, effectively solving the problem that traditional frequency domain analysis cannot take into account the local characteristics of time and frequency.

[0055] In order to ensure the effectiveness of the intelligent decision-making model and the accuracy of the collaborative adjustment strategy, Figure 3 As shown, the above step S3 further includes the following steps S301 to S305 instead, specifically: S301, the voltage fluctuation amplitude in the time-frequency joint feature , frequency deviation Phase mutation rate They are mapped to the corresponding fuzzy levels respectively, and the membership of each input variable to different fuzzy levels is calculated.

[0056] In this step, define the voltage fluctuation amplitude , frequency deviation Phase mutation rate The blur levels are as follows: The fuzzy set is {extremely small, small, medium, large, extremely large}, and the domain is , and its Gaussian membership function parameter is (center value) and (width); The fuzzy set is {negative large, negative small, zero, positive small, positive large}, and the domain is , the parameters are and ; The fuzzy set is {low, medium, high}, and the domain is , the parameters are and .

[0057] Calculate the membership of each input variable to each fuzzy level: , , .

[0058] S302, based on the preset fuzzy rule base, the membership of the input variables is logically combined, and the accurate reactive compensation coefficient is output after defuzzification using the centroid method. , voltage optimization coefficient and transformer ratio adjustment coefficient .

[0059] For example, 27 fuzzy rules are preset, and the rule format is as follows: "like for and for and for ,but for 、 for 、 for ” in, 、 、 is the input fuzzy set, 、 、 is the output fuzzy set; Using the Mamdani reasoning method, the “AND” operation is performed on each rule to obtain the rule activation strength ; For output variables 、 、 The fuzzy set is weighted averaged and the exact value is obtained by defuzzification using the centroid method: , 、 Calculate similarly; in, For the The corresponding rules Outputs the center value.

[0060] S303, constructing a power grid topology map , where the vertex set Represents a grid node, edge set Represents the admittance relationship between nodes. Node characteristics include 、 、 and the connection weights of adjacent nodes.

[0061] S304, the power grid topology map is analyzed through the spatiotemporal graph neural network (ST-GNN) Multi-layer graph convolution operations are performed, and the dynamic characteristics of each node are updated through adjacent node information transmission to extract the global spatiotemporal correlation characteristics of the power grid.

[0062] In this step, the node state update equation of the spatiotemporal graph neural network is:

[0063] in, is the adjacency matrix with self-loops added, is the adjacency matrix of the admittance connections of the grid nodes, is the identity matrix; for The degree matrix of ; For the Layer node feature matrix, is the number of nodes, is the feature dimension, and the initial features include 、 and ; is the learnable weight matrix; is the ReLU activation function.

[0064] This embodiment aggregates multi-hop neighborhood information through three layers of graph convolution and outputs a global spatiotemporal feature matrix .

[0065] S305 , by fusing the fuzzy inference output and the node features after graph convolution, a collaborative regulation strategy is generated with the goal of minimizing voltage fluctuation, frequency deviation, and equipment regulation cost.

[0066] In this step, the steps for generating the collaborative adjustment strategy include: Construct the optimization objective function:

[0067] in, 、 、 The preset weight coefficients control the optimization priorities of voltage fluctuation, frequency deviation and equipment regulation cost respectively; To regulate the number of devices; Equipment adjustment cost Defined as:

[0068] in, 、 、 is the adjustment coefficient of the device at the previous moment; 、 、 is the current adjustment coefficient; the device action intensity is quantified by calculating the absolute value of the difference to avoid frequent adjustments.

[0069] Constraints include: ; And it is an integer multiple of the tap adjustment step length; , is the allowable voltage fluctuation ratio.

[0070] Nodes in a power grid are coupled to each other through line admittance. Voltage regulation at any node affects the voltage state of adjacent nodes through the topological structure. Ignoring topological constraints can lead to global voltage imbalances while optimizing local optimization. Therefore, topological constraints must be used to ensure that regulation strategies adhere to the electrical laws of the physical connections of the power grid. Coordinated regulation strategies address topological constraints in the following ways: (1) Voltage coupling constraint between nodes: For adjacent nodes and , its voltage regulation satisfies:

[0071] in, is the inter-node admittance, is the admittance phase angle.

[0072] This constraint is implemented by the topological graph adjacency matrix Implementation: If the node and If connected, ,otherwise , ensuring that the voltage regulation amount complies with electrical laws in the topology network.

[0073] (2) Equipment dynamic response characteristic constraints: Reactive power compensation equipment response time constraint: The response time must meet , to avoid adjustment instructions exceeding the physical limits of the equipment; Minimum interval constraint for transformer ratio adjustment: The ratio adjustment amount at adjacent moments must satisfy , to prevent equipment damage caused by excessive adjustment.

[0074] Based on the power grid topology The adjacency matrix , the adjustment influence weights of adjacent nodes are calculated through the graph attention mechanism (GAT):

[0075] in, is the attention weight vector, For nodes The set of adjacent nodes of For the Layer Node The eigenvector of .

[0076] Node features based on graph convolution and fuzzy inference output , the optimal adjustment coefficient that satisfies the constraints is solved by the gradient descent method, and the optimized adjustment coefficient is converted into specific device control instructions, as follows: For reactive power compensation equipment, according to Calculate the target reactive power compensation amount: , is the rated capacity of the equipment; For transformer ratio adjustment equipment, according to Generate tap adjustment gear: , is the initial position of the transformer, Adjust the step size for transformer tap changer; For voltage optimization equipment, according to Generate target voltage reference value: , is the rated voltage of the grid.

[0077] By implementing the above steps S301 to S305, on the one hand, fuzzy reasoning can transform continuous voltage fluctuation characteristics into interpretable regulation rules, adapt to the uncertainty and nonlinear characteristics of power grid operation, and quickly generate preliminary regulation strategies; on the other hand, the spatiotemporal graph neural network models the power grid topology and uses graph convolution operations to capture the voltage coupling relationship and dynamic propagation law between nodes, thereby making up for the lack of global coordination in fuzzy reasoning; after the fusion of the two, the collaborative regulation strategy generated with multi-objective optimization as the goal has both the flexibility of local regulation and the consideration of the global topological constraints of the power grid and the dynamic response characteristics of the equipment, thereby achieving precise suppression of voltage fluctuations and coordinated control of multiple devices, effectively improving the voltage stability and regulation efficiency of the power grid.

[0078] Since the grid operation environment has strong time-varying and nonlinear characteristics, if the parameters of the intelligent decision-making model are fixed, the prediction error will accumulate due to the migration of the grid state. Therefore, in step S4, a closed-loop correction mechanism driven by a dynamic forgetting factor is used to iteratively optimize the model parameters through real-time error feedback, so that the system has the ability of "self-learning". Figure 4 As shown, step S4 specifically includes: Step S401, defining a dynamic forgetting factor , and its update formula is:

[0079] in, is the initial forgetting factor; is the forgetting rate parameter, which is used to control the decay rate of the error to the forgetting factor; for The voltage prediction error at time t, is the sliding window length; When the error When it increases, Rapid decay makes the model pay more attention to the current data; when the error decreases, Tend to be stable and retain historical effective experience.

[0080] Step S402, iterative optimization of model parameters based on forgetting factor:

[0081] in, It is a parameter set of the intelligent decision-making model, including fuzzy inference rule weights and graph neural network weights ; is the learning rate, which controls the iteration step size; is the gradient of the objective function with respect to the parameters, the forgetting factor The gradient direction is weighted so that the update amplitude increases when the error is large and the update tends to be gentle when the error is small.

[0082] Step S403, error feedback process of the closed-loop correction mechanism: Real-time acquisition of the actual voltage signal after adjustment and calculation of the prediction error ; like If it is greater than the preset threshold, the forgetting factor update and parameter iteration are triggered, otherwise the current parameters are maintained.

[0083] By implementing the above steps S401 to S403 and dynamically adjusting the weights according to historical errors using the forgetting factor, it is possible to quickly adapt to scenarios such as load mutations and topology changes, and simultaneously balance computing efficiency and accuracy through a sliding window mechanism. For model mismatches caused by factors such as transient disturbances or equipment aging, it is possible to iteratively optimize parameters through error feedback to avoid interference from historical invalid data, achieve "self-learning" correction, and ensure that the voltage suppression effect maintains high accuracy and robustness during full-operation conditions of the power grid.

[0084] Based on the same inventive concept, embodiments of the present application also provide a system for suppressing grid voltage fluctuations based on dynamic voltage regulation, for implementing the aforementioned method for suppressing grid voltage fluctuations based on dynamic voltage regulation. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the system for suppressing grid voltage fluctuations based on dynamic voltage regulation provided below can be found in the aforementioned limitations of the method for suppressing grid voltage fluctuations based on dynamic voltage regulation, and will not be further elaborated here.

[0085] In an exemplary embodiment, Figure 5As shown, a grid voltage fluctuation suppression system based on dynamic voltage regulation is provided, which includes a signal acquisition module, a feature decomposition module, an intelligent decision module and a closed-loop correction module. The signal acquisition module is used to synchronously collect three-phase voltage signals through a sensor array deployed at the grid nodes, and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transform.

[0086] Among them, the sensor array achieves microsecond-level time synchronization across the entire network through the IEEE1588 precision clock synchronization protocol, collects three-phase voltage signals at a sampling frequency of 12.8kHz, and generates a multidimensional feature vector containing the voltage components in the dq coordinate system, real-time frequency, and phase angle through Park transformation.

[0087] The feature decomposition module is used to separate the fundamental wave and harmonics of the multidimensional feature vector using a hybrid multi-scale decomposition algorithm, and extract the time-frequency joint features of voltage fluctuation amplitude, frequency deviation and phase mutation rate.

[0088] Specifically, the feature decomposition module performs three-layer wavelet packet decomposition on the multidimensional feature vector to separate the fundamental and harmonic components, calculates the fundamental parameters through FFT, and then extracts the time-frequency joint features of voltage fluctuation amplitude, frequency deviation and phase mutation rate.

[0089] The intelligent decision-making module is used to input the time-frequency joint features into a model that integrates fuzzy reasoning and spatiotemporal graph neural network, and combines the grid topology constraints and equipment dynamic response characteristics to generate target voltage optimization, adaptive reactive power compensation and transformer ratio coordinated adjustment strategies.

[0090] During the operation of the intelligent decision-making module, the joint time-frequency features are first input into a model that integrates fuzzy reasoning and a spatiotemporal graph neural network. The fuzzy reasoning unit maps the voltage fluctuation amplitude, frequency deviation, and phase mutation rate into fuzzy levels. Preliminary values ​​for the reactive power compensation coefficient, voltage optimization coefficient, and transformer ratio adjustment coefficient are output based on preset fuzzy rules. This process effectively handles the uncertainty and nonlinear relationships in voltage fluctuations. The spatiotemporal graph neural network performs multi-layer graph convolution operations based on the power grid topology. Node features include voltage fluctuation parameters, and edge features represent the admittance relationships between nodes. Graph convolution captures the global spatiotemporal correlation characteristics of the power grid through graph convolution, extracting the propagation pattern and mutual influence of voltage fluctuations at each node. Then, the local regulation coefficient output by fuzzy reasoning and the global topological features extracted by graph neural network are integrated to construct an optimization objective function that includes voltage fluctuation suppression effect and equipment regulation cost. At the same time, grid topology constraints and equipment dynamic response characteristics constraints such as inter-node voltage coupling constraints, reactive compensation equipment response time constraints, and transformer ratio adjustment minimum interval constraints are considered. The projected gradient method is used to solve the optimal regulation coefficient that meets the constraints. Finally, the target voltage optimization, adaptive reactive compensation and transformer ratio collaborative regulation strategies are generated to achieve coordinated optimization control of multiple devices, reducing equipment regulation costs while ensuring voltage quality.

[0091] The closed-loop correction module is used to iteratively optimize the parameters of the intelligent decision-making model based on real-time error feedback through a mechanism driven by a dynamic forgetting factor, thereby achieving adaptive suppression of voltage fluctuations.

[0092] Among them, this implementation method realizes the synchronous collection of multi-dimensional voltage characteristics through sensor array and Park transform, extracts time-frequency joint features with the help of hybrid multi-scale decomposition, integrates fuzzy reasoning and spatiotemporal graph neural network to generate a collaborative adjustment strategy that takes into account the topological constraints of the power grid and the response characteristics of the equipment, and uses the closed-loop correction driven by the dynamic forgetting factor to realize the adaptive optimization of model parameters. It has the ability of high-precision suppression of voltage fluctuations, multi-node collaborative adjustment, topology constraint processing and adaptive environmental changes, which can effectively improve the stability of the power grid and reduce the cost of equipment adjustment.

[0093] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6As shown. The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, memory and input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store grid voltage fluctuation suppression data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a grid voltage fluctuation suppression method based on dynamic voltage regulation is implemented.

[0094] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0095] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0096] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0097] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0098] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0099] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0100] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0101] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for suppressing grid voltage fluctuations based on dynamic voltage regulation, characterized in that: The steps include: S1. Synchronously collect the three-phase voltage signals of the grid nodes through the sensor array deployed at the grid nodes, and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transform; S2. Using the multidimensional feature vector as input, a hybrid multi-scale decomposition algorithm is used to separate the fundamental wave and harmonics, and to extract the time-frequency joint features of the voltage fluctuation amplitude, frequency deviation, and phase mutation rate; S3. Input the extracted time-frequency joint features into an intelligent decision-making model that integrates fuzzy reasoning and spatiotemporal graph neural network. Combined with the grid topology constraints and the dynamic response characteristics of the equipment, it generates target voltage optimization, adaptive reactive power compensation, and transformer ratio coordinated adjustment strategies. S4. Design a closed-loop correction mechanism driven by a dynamic forgetting factor, iteratively optimize the parameters of the intelligent decision-making model through real-time error feedback, and achieve adaptive suppression of voltage fluctuations.

2. The method according to claim 1, characterized in that The step S1 specifically includes: Three synchronous voltage sensors are configured at each grid node, corresponding to phases A, B, and C of the three-phase AC power; Collecting three-phase voltage signals by the synchronous voltage sensor; The collected three-phase voltage signal is input into the Parker transformation module, and the real-time phase angle and frequency of the power grid are estimated in real time through the phase-locked loop; The Parker transformation module converts the three-phase voltage signal into a voltage component in a two-phase rotating coordinate system, and generates a multidimensional feature vector in combination with the real-time frequency and phase angle. The multidimensional feature vector includes the two-phase voltage components, the real-time frequency and the real-time phase angle.

3. The method according to claim 2, characterized in that The Parker transformation module converts the three-phase voltage signal into a voltage component in a two-phase rotating coordinate system and generates a multi-dimensional feature vector by combining the real-time frequency and phase angle. Specifically, it includes: The three-phase voltage signal 、 、 Input to the Parker transform module, the transformation matrix of the Parker transform module is defined as: in, , , is the initial phase, The real-time frequency of the power grid is estimated in real time through a phase-locked loop; After conversion, the voltage component in the dq coordinate system is obtained ; Combined with the real-time frequency of the power grid and phase angle , generate multidimensional feature vectors .

4. The method according to claim 3, characterized in that The hybrid multi-scale decomposition algorithm in step S2 includes: For the voltage components in the multidimensional feature vector Perform 3-layer wavelet packet decomposition to obtain 8 frequency band components , where the fourth frequency band component corresponds to the fundamental component; Perform fast Fourier transform on the fundamental component to calculate the fundamental amplitude 、 , fundamental frequency and fundamental phase ; The harmonic component is defined as the difference between the original signal and the fundamental component: ， The voltage fluctuation amplitude is calculated as the norm of the harmonic components: The frequency deviation is calculated as: in, is the rated frequency of the grid; The phase mutation rate is calculated by numerical differentiation: in, is the sampling time interval.

5. The method according to claim 4, characterized in that The step S3 specifically includes: The voltage fluctuation amplitude in the time-frequency joint feature , frequency deviation Phase mutation rate Map them to the corresponding fuzzy levels respectively and calculate the membership of each input variable to different fuzzy levels; Based on the preset fuzzy rule base, the membership of the input variables is logically combined, and the accurate reactive compensation coefficient is output after defuzzification by the centroid method. , voltage optimization coefficient and transformer ratio adjustment coefficient ; Build a power grid topology map , where the vertex set Represents a grid node, edge set Represents the admittance relationship between nodes. Node characteristics include voltage fluctuation amplitude, frequency deviation, and phase mutation rate. Power grid topology map through spatiotemporal graph neural network Perform multi-layer graph convolution operations and update the dynamic characteristics of each node through adjacent node information transmission to extract the global spatiotemporal correlation characteristics of the power grid; By integrating fuzzy inference output and node features after graph convolution, a collaborative regulation strategy is generated with the goal of minimizing voltage fluctuation, frequency deviation and equipment regulation cost.

6. The method according to claim 5, characterized in that The node state update equation of the spatiotemporal graph neural network is: in, is the adjacency matrix with self-loops added, is the adjacency matrix of the admittance connections of the grid nodes, is the identity matrix; for The degree matrix of ; For the Layer node feature matrix, is the number of nodes, is the feature dimension, and the initial features include 、 and ; is the learnable weight matrix; is the ReLU activation function.

7. The method according to claim 6, characterized in that The steps of generating the collaborative regulation strategy include: Construct the optimization objective function: in, 、 、 The preset weight coefficients control the optimization priorities of voltage fluctuation, frequency deviation and equipment regulation cost respectively; To regulate the number of devices; Equipment adjustment cost Defined as: in, 、 、 is the adjustment coefficient of the device at the previous moment; Node features based on graph convolution and fuzzy inference output ,The optimal adjustment coefficient that satisfies the constraints is solved by the gradient descent method, and the coordinated adjustment instructions of each node are generated.

8. The method according to claim 7, characterized in that The coordinated regulation strategy handles grid topology constraints by: Inter-node voltage coupling constraint: For adjacent nodes and , its voltage regulation satisfies: in, is the inter-node admittance, is the admittance phase angle; Equipment dynamic response characteristic constraints: Reactive power compensation equipment response time constraints: ; Transformer ratio adjustment minimum interval constraint: ; Based on the power grid topology The adjacency matrix , the adjustment influence weights of adjacent nodes are calculated through the graph attention mechanism: in, is the attention weight vector, For nodes The set of adjacent nodes of For the Layer Node The eigenvector of .

9. The method according to claim 8, characterized in that The step S4 specifically includes: Defining the dynamic forgetting factor , and its update formula is: in, is the initial forgetting factor, is the forgetting rate parameter; for The voltage prediction error at time t, is the sliding window length; Iterative optimization of model parameters based on forgetting factor: in, is the parameter set of the intelligent decision-making model; is the learning rate; is the gradient of the objective function with respect to the parameters; Error feedback process of closed-loop correction mechanism: Real-time acquisition of the actual voltage signal after adjustment and calculation of the prediction error ; like If it is greater than the preset threshold, the forgetting factor update and parameter iteration are triggered, otherwise the current parameters are maintained.

10. A grid voltage fluctuation suppression system based on dynamic voltage regulation, characterized in that: include: The signal acquisition module is used to synchronously collect three-phase voltage signals through a sensor array deployed at the grid nodes and generate a multi-dimensional feature vector containing dynamic amplitude, phase angle and frequency components through Park transform; A feature decomposition module is used to separate the fundamental wave and harmonics of the multidimensional feature vector using a hybrid multi-scale decomposition algorithm, and to extract the time-frequency joint features of the voltage fluctuation amplitude, frequency deviation and phase mutation rate; An intelligent decision-making module is used to input the time-frequency joint features into a model that integrates fuzzy reasoning and spatiotemporal graph neural networks, and generate target voltage optimization, adaptive reactive power compensation, and transformer ratio coordinated adjustment strategies based on grid topology constraints and equipment dynamic response characteristics; The closed-loop correction module is used to iteratively optimize the parameters of the intelligent decision-making model based on real-time error feedback through a mechanism driven by a dynamic forgetting factor, thereby achieving adaptive suppression of voltage fluctuations.

Images