Distributed energy storage grid-connected voltage stability prediction method, medium and system

Abstract

The invention provides a distributed energy storage grid-connected voltage stability prediction method, medium and system, and belongs to the technical field of distributed energy storage, and the method comprises the steps: obtaining the parameters of each node according to a distributed energy storage urban power grid topological graph; an energy storage grid-connected voltage influence equation set is established according to the parameters, and the stability range of each node parameter is obtained through calculation. Based on the topological graph and the node parameter stability range, the voltage stability margin of each node is calculated, and a power grid voltage stability margin topological graph is formed. And inputting the voltage stability margin matrix at a plurality of continuous moments into a pre-trained prediction model to predict the voltage stability margin of each node in a period of time in the future. The system comprehensively considers the influence of distributed energy storage on the voltage stability of the power grid, and solves the technical problem that the existing distributed energy storage grid-connected voltage stability prediction method is difficult to solve the problem that the voltage stability analysis generalization ability is poor due to the randomness of load and power generation and the uncertainty of power transmission line parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed energy storage, and in particular, relates to a distributed energy storage grid-connected voltage stability prediction method, medium and system. Background Art

[0002] With the continuous advancement of renewable energy generation technology and the rapid development of power electronics technology, distributed generation and energy storage technology are increasingly being used in power systems. Unlike the traditional centralized power generation model, distributed power sources can be connected to the distribution network nearby, providing users with a more flexible and reliable power supply. The introduction of energy storage systems further enhances the peak load regulation and support capabilities of distributed power sources and improves the operational stability of the power grid.

[0003] At present, the deployment of distributed power sources and energy storage systems in urban power grids has become a general trend. However, the access of a large number of decentralized distributed power sources and energy storage devices to the power grid has also brought new challenges to the operation and dispatch of the power grid. Among them, the problem of grid voltage stability is particularly prominent. The grid connection of distributed power sources and energy storage systems will change the power flow distribution of the power grid, thereby affecting the voltage amplitude and phase angle of each node. If this change exceeds the tolerance of the power grid, it may lead to local or overall voltage instability, and even the risk of voltage collapse. Therefore, how to accurately predict and effectively control the impact of distributed grid connection on grid voltage stability has become one of the key issues in optimizing the operation of the power system.

[0004] Traditional voltage stability analysis methods mainly include load characteristic analysis, energy function method and modal analysis method. These methods usually require the establishment of a complete power grid topology model and power flow equation, and use numerical calculation to solve them. However, in actual power grids, voltage stability analysis becomes very complicated due to factors such as the randomness of load and power generation, the uncertainty of transmission line parameters, and the dynamic characteristics of distributed power sources and energy storage systems. In particular, when a large number of distributed energy storage is connected to the power grid, its charging and discharging behavior will further aggravate the uncertainty of the power grid and bring greater challenges to voltage stability analysis. In other words, the existing distributed energy storage grid-connected voltage stability prediction method is difficult to solve the technical problem of poor generalization ability of voltage stability analysis caused by factors such as the randomness of load and power generation and the uncertainty of transmission line parameters. Summary of the invention

[0005] In view of this, the present invention provides a distributed energy storage grid-connected voltage stability prediction method, medium and system, which can solve the technical problem that the existing distributed energy storage grid-connected voltage stability prediction method is difficult to solve due to the randomness of load and power generation, the uncertainty of transmission line parameters and other factors. The generalization ability of voltage stability analysis is poor.

[0006] The present invention is achieved in that:

[0007] A first aspect of the present invention provides a method for predicting voltage stability of a distributed energy storage grid, comprising the following steps:

[0008] S10, establishing a topological diagram of an urban power grid including distributed energy storage, wherein the topological diagram includes nodes and edges, wherein the nodes include power-input nodes, power generation nodes, load nodes, and energy storage nodes, and the edges represent power transmission paths between nodes, and obtaining the value of a node parameter of each node;

[0009] S20, establishing a group of energy storage grid-connected voltage influence equations according to the node parameters, including a power balance equation, a node voltage equation, an energy storage state equation, a power flow equation, and a voltage stability index equation;

[0010] S30, calculating the node parameter stability range of each node according to the energy storage grid-connected voltage influence equation group and the initial values ​​and constraints of each node parameter, wherein the node parameter stability range refers to the range of node parameter values ​​that ensures the node voltage stability;

[0011] S40, based on the urban power grid topology, the stability range of each node parameter and the current voltage state of the node acquired in real time, calculating the voltage stability margin of each node, wherein the voltage stability margin represents the distance between the current voltage state of the node and the critical point of voltage instability;

[0012] S50, setting the value of each node of the urban power grid topology map as the voltage stability margin of the node to form a power grid voltage stability margin topology map, and converting the power grid voltage stability margin topology map into a power grid voltage stability margin matrix using an adjacency matrix method;

[0013] S60, repeatedly executing S40-S50 to obtain a grid voltage stability margin matrix at a plurality of consecutive acquisition moments;

[0014] S70, input the obtained grid voltage stability margin matrix of multiple consecutive acquisition moments into a pre-trained grid voltage stability margin prediction model, and output the voltage stability margin prediction value of each node in the future period as the distributed energy storage grid-connected voltage stability prediction result.

[0015] Among them, the node parameters include power-in node parameters, power generation node parameters, load node parameters and energy storage node parameters. The power-in node parameters include voltage amplitude, phase angle and short-circuit capacity. The power generation node parameters include active power, reactive power and voltage amplitude. The load node parameters include active load, reactive load and voltage amplitude. The energy storage node parameters include charging and discharging power, energy storage capacity and voltage amplitude.

[0016] Among them, the initial values ​​and constraints of the parameters of each node are specifically as follows: the voltage amplitude of the power-in node is ±5% of the rated value, the phase angle is 0°, and the short-circuit capacity is determined according to the actual system parameters; the active power of the power generation node is between 0 and the rated power, the reactive power is between -30% and +30% of the rated power, and the voltage amplitude is ±5% of the rated value; the active load and reactive load of the load node are determined according to historical data and load forecasts, and the voltage amplitude is allowed to be between 90% and 110% of the rated value; the charging and discharging power of the energy storage node does not exceed the rated power, the energy storage capacity is between 20% and 80%, and the voltage amplitude is ±5% of the rated value.

[0017] The voltage state refers to a complex number representation consisting of the amplitude and phase angle of the node voltage.

[0018] The method for obtaining the voltage instability critical point is: through continuous power flow calculation and PV curve analysis, gradually increase the system load until the power flow calculation does not converge or the voltage collapse point appears, and the non-convergence or voltage collapse point is the voltage instability critical point.

[0019] The following is the formula expression of each equation in the equation group of the influence of energy storage grid voltage:

[0020] 1. Power balance equation:

[0021]

[0022] Where P i ,Q i is the injected active and reactive power of node i; V i ,V k is the complex voltage at nodes i and k; Y ik is the element in row i and column k in the node admittance matrix; P L,i ,Q L,i is the active and reactive power of the load at node i; P S,i ,Q S,i is the active and reactive power injected into the energy storage system of node i; n is the total number of nodes in the system; j is the imaginary unit; is the complex conjugate.

[0023] 2. Node voltage equation:

[0024]

[0025] In the formula, |V i | is the voltage amplitude of node i; θ i is the voltage phase angle of node i; are the real and imaginary parts of the voltage at node i; t is time;

[0026] Among them, the rate of change of voltage amplitude and phase angle can be calculated by the following equation:

[0027]

[0028] 3. Energy storage state equation:

[0029]

[0030] In the formula, E S,i is the energy state of the energy storage system at node i; is the charging and discharging power of the energy storage system; η c,i ,η d,i is the charging and discharging efficiency; P loss,i is the self-discharge loss of the energy storage system; is the maximum capacity of the energy storage system; is the maximum charging and discharging power of the energy storage system.

[0031] 4. Flow equation:

[0032]

[0033] In the formula, G ik ,B ik is the real and imaginary part of the element in row i and column k in the node admittance matrix; θ ik =θ i -θ k is the phase angle difference between nodes i and k.

[0034] 5. Voltage stability index equation:

[0035]

[0036] Where, L i is the L index of node i; VSI i is the voltage stability index of node i; Y ii are the diagonal elements of the node admittance matrix;

[0037] Among them, the voltage stability margin can be calculated by the following equation:

[0038] VSM i =VSI i -VSI i,crit ;

[0039] Among them, VSI i,crit It is the voltage stability index value at the critical stability point, which can be determined by continuously increasing the load and performing power flow calculations until the system collapses.

[0040] Among them, the grid voltage stability margin prediction model adopts a multi-task learning structure, including a time series feature extraction subnetwork, a spatial feature extraction subnetwork, a change factor processing subnetwork and a fusion subnetwork.

[0041] Furthermore, the input of the time series feature extraction subnetwork is the grid voltage stability margin matrix at multiple consecutive acquisition moments, and the output is a time series feature vector, which is used to capture the time variation law of the voltage stability margin, and its structure is a recurrent neural network based on a long short-term memory network;

[0042] The input of the spatial feature extraction subnetwork is the grid voltage stability margin matrix at each acquisition moment, and the output is a spatial feature vector, which is used to extract the spatial correlation in the grid topology structure, and its structure is based on a graph convolutional neural network;

[0043] The input of the change factor processing subnetwork is load forecast and energy storage processing forecast data, and the output is a load-energy storage forecast feature vector, which is used to consider the impact of load changes on voltage stability. Its structure is a multi-layer feedforward neural network;

[0044] The input of the fusion subnetwork is the concatenation of the time series feature vector, the spatial feature vector and the load-energy storage prediction feature vector, and the output is the predicted value of the voltage stability margin of each node in the future period, which is used to integrate various features for final prediction. Its structure is a multi-layer feedforward neural network.

[0045] The training steps of the grid voltage stability margin prediction model are briefly described as follows:

[0046] Step 1: Initialize model parameters;

[0047] Step 2: Use the training set data for forward propagation;

[0048] Step 3: Calculate the loss between the predicted value and the actual value;

[0049] Step 4: Back propagation updates model parameters;

[0050] Step 5: Evaluate model performance on the validation set.

[0051] Step 6: Repeat steps 2-5 until the preset number of training rounds or performance index is reached;

[0052] Step 7: Use the test set for final evaluation.

[0053] The specific steps of acquiring the training data set of the power grid voltage stability margin prediction model include:

[0054] Step 1: Using steps S10-S60, obtain the grid voltage stability margin matrix of multiple consecutive historical acquisition moments, and then perform the following steps:

[0055] Step 2: Collect forecast data related to power grid operation, including load forecast data and energy storage output forecast data. The load forecast data is the predicted active load and reactive load data of the load node, and the energy storage output forecast data is the predicted charging and discharging power of the energy storage node. The forecasting method is to form a continuous data curve every 15 minutes based on historical records, and then use a time series forecasting model (such as ARIMA, Prophet or LSTM network) to predict the data for the next 24 hours. These forecasts take into account seasonal changes, cyclical patterns and long-term trends. For load forecasting, the influence of external variables such as weather factors and holiday effects must also be considered. For energy storage output forecasting, it is necessary to combine factors such as the operation strategy, charging and discharging efficiency and remaining capacity of the energy storage system.

[0056] Step 3: Arrange the grid voltage stability margin matrices of multiple consecutive historical acquisition moments in chronological order to form a time series data set;

[0057] Step 4: Use the sliding window method to process the time series data set, set the input window length and prediction window length, and generate input-output sample pairs;

[0058] Step 5: Align the predicted data with the grid voltage stability margin matrix samples at the corresponding time to obtain a training data set, where the training input is:

[0059] The grid voltage stability margin matrix sequence for the past n time steps, where n is the set input window length.

[0060] The load forecast data for the corresponding time period includes the active load and reactive load forecast values ​​for each load node.

[0061] The energy storage output forecast data for the corresponding time period includes the predicted charging and discharging power of each energy storage node.

[0062] The grid topology information can be represented by an adjacency matrix.

[0063] The output of the training is:

[0064] The predicted voltage stability margin of each node in the next ω time steps, where ω is the set prediction window length. These predicted values ​​form a three-dimensional tensor with dimensions (ω, number of nodes, 1), where 1 represents a voltage stability margin value for each node.

[0065] Furthermore, the frequency of repeatedly executing S40 to S50 is once every 15 minutes.

[0066] Specifically, step S10 includes: first, obtaining the basic topological information of the urban power grid of the research object, including the type of each node, the connection relationship between nodes, and the relevant parameters of each node. Node types include power-input nodes, power generation nodes, load nodes, and energy storage nodes, and the connection relationship between nodes can be represented by edges. Each node has its own node parameters, including voltage amplitude, phase angle, active power, reactive power, load power, energy storage capacity, and charging and discharging power. These parameters can be obtained through power grid modeling software or on-site measurements. On this basis, a power grid topology diagram containing distributed energy storage is constructed, and the connection relationship between nodes is described using methods such as adjacency matrix or adjacency table, and various types of nodes are identified, and finally the parameter value of each node is obtained.

[0067] Wherein, the step S20 specifically includes: according to the grid topology information and node parameters obtained in step S10, a mathematical model describing the influence of distributed energy storage on grid voltage stability is established, including power balance equation, node voltage equation, energy storage state equation, power flow equation and voltage stability index equation. The power balance equation describes the balance relationship of active and reactive power of nodes, including transmission line loss, load demand and energy storage system injection power; the node voltage equation describes the variation law of node voltage amplitude and phase angle over time; the energy storage state equation describes the charging and discharging process and energy state of the energy storage system; the power flow equation is used to calculate the active and reactive power of each node; the voltage stability index equation defines the voltage stability index and gives the formula for calculating the voltage stability margin. By establishing such a mathematical model, the foundation is laid for the subsequent voltage stability analysis and prediction.

[0068] Wherein, the step S30 specifically includes: first, according to the actual operation of the power grid, the initial value range of each node parameter is given, for example, the voltage amplitude of the power-in node is ±5% of the rated value, the active power of the power generation node is between 0 and the rated power, and the voltage amplitude of the load node is allowed to be between 90% and 110% of the rated value. Then these initial value ranges are substituted into the mathematical model established in step S20, and each equation is solved by numerical calculation to obtain the node parameter value range that meets the voltage stability constraint condition. For a certain node parameter, if its value exceeds the initial range, it is necessary to appropriately adjust the value range of the parameter, and repeat the calculation until the parameter value range that meets the voltage stability requirements is found. In this way, the stable value range of each node parameter is obtained, providing the necessary constraint conditions for the subsequent voltage stability prediction.

[0069] Wherein, the step S40 specifically includes: first, according to the power grid topology diagram of step S10 and the node parameter stability range obtained in step S30, the mathematical equation in step S20 is solved by numerical calculation method to obtain the current voltage amplitude and phase angle of each node. Then, these voltage state parameters are substituted into the voltage stability index equation to calculate the voltage stability index value of each node. Then, by continuously increasing the load until the critical point where the power flow calculation does not converge or the voltage collapses, the voltage stability index value at the critical point is obtained. Finally, according to the difference between the current voltage stability index value of the node and the critical point index value, the voltage stability margin of each node is calculated. These voltage stability margin values ​​can intuitively reflect the current voltage stability state of the power grid.

[0070] Wherein, the step S50 specifically includes: marking each node in the power grid topology diagram of step S10 with the voltage stability margin value calculated in step S40 to form a power grid voltage stability margin topology diagram. Then, the adjacency matrix method is used to convert this topology diagram into a power grid voltage stability margin matrix, and the matrix elements represent the voltage stability margin values ​​between the nodes. This matrix form not only includes the voltage stability margin of each node itself, but also reflects the mutual influence relationship between the voltage stability of the nodes, providing a data form that is easy to process and analyze for the subsequent voltage stability prediction.

[0071] Among them, the steps S60 and S70 specifically include: firstly, setting a suitable sampling time interval, repeating steps S40-S50, and obtaining a matrix sequence of voltage stability margin of the power grid at multiple consecutive sampling moments. These data not only include the time-varying characteristics of the voltage stability margin of each node itself, but also reflect the dynamic characteristics of the mutual influence of voltage stability between nodes. Then, using the pre-trained power grid voltage stability margin prediction model, the voltage stability margin matrix sequence obtained in step S60 and the related load and energy storage prediction data are used as input to output the voltage stability margin prediction value of each node in the future. The prediction model adopts a multi-task learning structure, including modules such as a time series feature extraction subnetwork, a spatial feature extraction subnetwork, a change factor processing subnetwork, and a fusion subnetwork, which can comprehensively consider the influence of factors such as power grid topology, load changes, and energy storage systems on voltage stability.

[0072] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned distributed energy storage grid-connected voltage stability prediction method.

[0073] A third aspect of the present invention provides a distributed energy storage grid-connected voltage stability prediction system, which includes the above-mentioned computer-readable storage medium.

[0074] Compared with the prior art, the beneficial effects of the distributed energy storage grid-connected voltage stability prediction method, medium and system provided by the present invention are: the present invention fully considers the influence of factors such as grid topology, load changes and energy storage system operation on voltage stability, and can more accurately predict the voltage stability state of the future grid.

[0075] Specifically, the present invention first establishes a city power grid topology model including distributed energy storage and obtains parameter information of each node. Then, based on the power grid model, a set of mathematical equations describing the impact of distributed energy storage on voltage stability are derived, including power balance equations, node voltage equations, energy storage state equations, power flow equations, and voltage stability index equations. Using these equations, the voltage stability margin of each node in the power grid is calculated, reflecting the distance between the current voltage state of the node and the critical point of instability.

[0076] Next, the present invention converts the voltage stability margin information of the grid nodes into a matrix form and obtains the voltage stability margin matrix sequence at multiple consecutive moments. Finally, a grid voltage stability margin prediction model based on multi-task learning is designed, which can comprehensively consider factors such as time series characteristics, spatial correlation, and load and energy storage changes, and output the voltage stability margin prediction value of each node in the future period of time.

[0077] Compared with the traditional voltage stability analysis method, the advantages of the present invention are mainly reflected in the following aspects:

[0078] 1) The dynamic characteristics of the distributed energy storage system are fully considered, and its impact on the grid voltage stability is included in the analysis scope, which is more in line with the complex situation of actual grid operation.

[0079] 2) The voltage stability margin information is processed in matrix representation, which not only reflects the stability of the node itself, but also describes the mutual influence between nodes, providing richer data input for subsequent predictive analysis.

[0080] 3) The prediction model based on machine learning can automatically learn the complex laws of grid voltage stability, and has stronger generalization ability and prediction accuracy than traditional numerical calculation methods.

[0081] Therefore, the distributed energy storage urban power grid voltage stability prediction method proposed in the present invention can effectively make up for the shortcomings of the existing technology and solve the technical problem that the existing distributed energy storage grid-connected voltage stability prediction method is difficult to solve the poor generalization ability of voltage stability analysis caused by factors such as the randomness of load and power generation and the uncertainty of transmission line parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0084] like Figure 1 FIG. 1 is a flow chart of a method for predicting voltage stability of a distributed energy storage grid-connected power system provided by the present invention. The method comprises the following steps:

[0085] S10, establishing a topological diagram of an urban power grid including distributed energy storage, wherein the topological diagram includes nodes and edges, wherein the nodes include power-input nodes, power generation nodes, load nodes, and energy storage nodes, and the edges represent power transmission paths between nodes, and obtaining the value of a node parameter of each node;

[0086] S20, establishing a group of energy storage grid-connected voltage influence equations according to the node parameters, including a power balance equation, a node voltage equation, an energy storage state equation, a power flow equation, and a voltage stability index equation;

[0087] S30, calculating the node parameter stability range of each node according to the energy storage grid-connected voltage influence equation group and the initial values ​​and constraints of each node parameter, where the node parameter stability range refers to the range of node parameter values ​​that ensure node voltage stability;

[0088] S40, based on the urban power grid topology, the stability range of each node parameter and the current voltage state of the node acquired in real time, calculating the voltage stability margin of each node, where the voltage stability margin represents the distance between the current voltage state of the node and the critical point of voltage instability;

[0089] S50, setting the value of each node of the urban power grid topology map as the voltage stability margin of the node to form a power grid voltage stability margin topology map, and converting the power grid voltage stability margin topology map into a power grid voltage stability margin matrix using an adjacency matrix method;

[0090] S60, repeatedly executing S40-S50 to obtain a grid voltage stability margin matrix at a plurality of consecutive acquisition moments;

[0091] S70, input the obtained grid voltage stability margin matrix of multiple consecutive acquisition moments into a pre-trained grid voltage stability margin prediction model, and output the voltage stability margin prediction value of each node in the future period as the distributed energy storage grid-connected voltage stability prediction result.

[0092] The specific implementation methods of the above steps are described in detail below:

[0093] Step S10: Establishing a city power grid topology map including distributed energy storage

[0094] The specific implementation method of this step is as follows:

[0095] First, it is necessary to obtain the basic topological information of the urban power grid of the research object. This includes the types of nodes in the power grid, the connection relationship between nodes, and the relevant parameters of each node. Specifically, the node types include power-input nodes, power generation nodes, load nodes, and energy storage nodes. The connection relationship between nodes can be represented by edges, which represent the transmission path between nodes. Each node has its own node parameters, including voltage amplitude, phase angle, active power, reactive power, load power, energy storage capacity, and charging and discharging power. These parameters can be obtained through power grid modeling software or field measurements.

[0096] After obtaining the basic topology information of the power grid, the next step is to construct a power grid topology diagram that includes distributed energy storage. The specific approach is:

[0097] 1) Based on the acquired node and edge information, a topological structure diagram of the power grid is established. This can be done by using the representation method of nodes and edges in graph theory, such as using an adjacency matrix or adjacency table to describe the connection relationship between nodes.

[0098] 2) Based on the topological structure, identify various nodes in the power grid, including power-input nodes, power generation nodes, load nodes, and energy storage nodes. These nodes are represented by different symbols or colors in the topological diagram.

[0099] 3) For each node, obtain its corresponding node parameter value. These parameter values ​​can be obtained through power flow calculation results of power grid simulation software, field measurement data or relevant literature.

[0100] Through the above steps, a topological diagram of the urban power grid including distributed energy storage is established, and the parameter information of each node is obtained. This topological diagram lays the foundation for subsequent analysis and calculation. The main purpose of this step is to build a mathematical model that can reflect the actual situation of the power grid and provide necessary input data for subsequent voltage stability analysis.

[0101] Step S20: Establishing the energy storage grid voltage influence equation group according to the node parameters

[0102] Based on the grid topology information and node parameters obtained in step S10, the purpose of this step is to establish a mathematical model that describes the impact of distributed energy storage on grid voltage stability. This model includes the following equations:

[0103] 1) Power balance equation:

[0104]

[0105] This equation describes the active power P of node i. i and reactive power Q i The balance between transmission line losses, load demand, and energy storage system injection power. i and V k is the complex voltage at nodes i and k, Y ik is the corresponding element in the node admittance matrix, P L,i and Q L,i is the active and reactive power of the load at node i, P S,i and Q S,i Active and reactive power injected into the energy storage system of node i.

[0106] 2) Node voltage equation:

[0107]

[0108] This set of equations describes the variation of the amplitude and phase angle of the node voltage over time, and provides the formula for calculating the voltage amplitude and phase angle change rate.

[0109] 3) Energy storage state equation:

[0110]

[0111] These equations describe the energy state changes of the energy storage system, including the charging, discharging and self-discharging processes, and define the capacity and power boundary conditions of the energy storage system.

[0112] 4) Flow equation:

[0113]

[0114] This set of equations is used to calculate the active power P of each node i and reactive power Q i , where G ik and B ik is the real and imaginary part of the node admittance matrix, θ ik is the phase angle difference between nodes i and k.

[0115] 5) Voltage stability index equation:

[0116]

[0117] VSM i =VSI i -VSI i,crit

[0118] This set of equations defines the voltage stability index L i and VSI i, and the voltage stability margin VSM is calculated i The formula is: ii is the diagonal element of the node admittance matrix, VSI i,crit It is the voltage stability index value of the critical stability point.

[0119] Through the above equations, a mathematical model describing the impact of distributed energy storage on grid voltage stability is established. This model can be used to analyze the voltage stability margin of each node in the grid and provide a basis for subsequent voltage stability prediction.

[0120] Step S30: Calculate the stability range of node parameters according to the energy storage grid voltage influence equation group

[0121] Based on the energy storage grid-connected voltage influence equations established in step S20, as well as the initial values ​​and constraints of each node parameter, the purpose of this step is to calculate the node parameter value range that ensures node voltage stability. The specific steps are as follows:

[0122] 1) Initialize the value range of each node parameter:

[0123] -The voltage amplitude of the input node is ±5% of the rated value, the phase angle is 0°, and the short-circuit capacity is determined according to the actual system parameters;

[0124] - The active power of the power generation node is between 0 and the rated power, the reactive power is between -30% and +30% of the rated power, and the voltage amplitude is ±5% of the rated value;

[0125] -The active load and reactive load of the load node are determined based on historical data and load forecasts, and the voltage amplitude is allowed to be between 90% and 110% of the rated value;

[0126] -The charging and discharging power of the energy storage node does not exceed the rated power, the energy storage capacity is between 20% and 80%, and the voltage amplitude is ±5% of the rated value.

[0127] 2) Substitute the initial value ranges of the above node parameters into the energy storage grid-connected voltage influence equation group established in step S20, and use numerical calculation methods to solve these equations to obtain the value ranges of each node parameter that meets the voltage stability constraint conditions.

[0128] 3) For a node parameter, if it exceeds the initial value range during the solution process, it means that the value range of the parameter is not enough to ensure voltage stability. At this time, it is necessary to appropriately adjust the value range of the parameter and recalculate until the parameter value range that meets the voltage stability requirements is found.

[0129] Through the above steps, the stable value range of each node parameter is obtained, which can ensure that the power grid maintains voltage stability when distributed energy storage is connected to the grid. These results provide necessary constraints for subsequent voltage stability prediction.

[0130] Step S40: Calculate the voltage stability margin based on the grid topology and node parameter stability range

[0131] In step S30, the stable value range of the parameters of each node of the power grid has been obtained. The purpose of this step is to further calculate the voltage stability margin of each node. The voltage stability margin is an important indicator to characterize the voltage stability of the node. It reflects the distance between the current voltage state of the node and the critical point of voltage instability. The specific implementation steps are as follows:

[0132] 1) Based on the urban power grid topology established in step S10 and the stability range of each node parameter obtained in step S30, the energy storage grid-connected voltage influence equation group in step S20 is solved by numerical calculation method to obtain the current voltage amplitude and phase angle of each node.

[0133] 2) Substitute the current voltage state (amplitude and phase angle) of each node into the voltage stability index equation in step S20 to calculate the voltage stability index VSI of each node. i .

[0134] 3) By continuously increasing the system load, the voltage stability index VSI is gradually solved using the calculation method in step 2 i , until a critical point occurs where the power flow calculation does not converge or the voltage collapses. The voltage stability index value corresponding to this critical point is recorded as VSI i,crit , which is the critical point of voltage instability.

[0135] 4) Based on the results obtained in steps 2 and 3, calculate the voltage stability margin VSM of each node i =VSI i -VSI i,crit The larger the voltage stability margin, the more stable the node voltage is, and vice versa, the closer the voltage is to the critical point of instability.

[0136] Through the above steps, the voltage stability margin value of each node in the power grid is obtained. These results can intuitively reflect the current voltage stability state of the power grid and provide necessary input data for subsequent voltage stability prediction.

[0137] Step S50: Convert the grid voltage stability margin information into a matrix representation

[0138] Based on the voltage stability margin of each node of the power grid calculated in step S40, the purpose of this step is to convert this information into a matrix form to provide input data for subsequent voltage stability prediction. The specific implementation steps are as follows:

[0139] 1) Each node in the urban power grid topology map established in step S10 is marked with the voltage stability margin value calculated by the node, thus forming a power grid voltage stability margin topology map.

[0140] 2) Adopting the adjacency matrix method, the above-mentioned grid voltage stability margin topology graph is converted into a grid voltage stability margin matrix VSM. Specifically, the number of rows and columns of the matrix is ​​equal to the total number of nodes n in the grid, and the matrix element VSM ij represents the voltage stability margin value between node i and node j. If there is no direct connection between node i and node j, then VSM ij = 0. Diagonal elements VSM ii It represents the voltage stability margin of node i itself.

[0141] Through the above steps, the voltage stability margin information of each node in the power grid is converted into a matrix representation. Such a matrix not only contains the voltage stability margin of each node itself, but also reflects the mutual influence relationship between the voltage stability of the nodes. This provides a data form that is easy to process and analyze for subsequent voltage stability prediction.

[0142] Step S60: Repeat S40-S50 to obtain the grid voltage stability margin matrix at multiple consecutive acquisition moments.

[0143] In steps S40-S50, the voltage stability margin matrix of the power grid at a certain moment has been calculated and converted. In order to analyze the change law of power grid voltage stability over time, the purpose of this step is to repeatedly execute steps S40-S50 to obtain a sequence of power grid voltage stability margin matrices at multiple consecutive acquisition moments.

[0144] The specific steps are as follows:

[0145] 1) Set an appropriate sampling interval, such as sampling every 15 minutes.

[0146] 2) For each sampling moment, steps S40-S50 are repeated to calculate and transform the grid voltage stability margin matrix at that moment.

[0147] 3) The grid voltage stability margin matrices obtained by continuous sampling at multiple moments are arranged in chronological order to form a three-dimensional tensor VSM(t), in which the first dimension corresponds to the time series, the second dimension corresponds to the grid nodes, and the third dimension corresponds to the voltage stability margin value of each node.

[0148] Through the above steps, the change process of the voltage stability margin of the power grid at multiple consecutive moments is obtained, which provides the required time series data for the subsequent voltage stability prediction. These data not only include the time-varying characteristics of the voltage stability margin of each node itself, but also reflect the dynamic characteristics of the mutual influence of voltage stability between nodes.

[0149] Step S70: Predict using the pre-trained grid voltage stability margin prediction model

[0150] Based on the grid voltage stability margin matrix sequence of multiple consecutive acquisition moments obtained in step S60, combined with the prediction data of load and energy storage system, the purpose of this step is to use the pre-trained grid voltage stability margin prediction model to output the voltage stability margin prediction value of each node in the future. The specific implementation steps are as follows:

[0151] 1) Structure of the grid voltage stability margin prediction model:

[0152] The prediction model adopts a multi-task learning structure, including the following four sub-networks:

[0153] -Time series feature extraction sub-network: A recurrent neural network based on the long short-term memory network (LSTM) is used to capture the changing pattern of the voltage stability margin time series.

[0154] -Spatial feature extraction sub-network: Based on graph convolutional neural network (GCN), it is used to extract spatial correlation in power grid topology.

[0155] -Change factor processing subnetwork: a multi-layer feedforward neural network used to consider the impact of load and energy storage system changes on voltage stability.

[0156] - Fusion sub-network: A multi-layer feed-forward neural network used to integrate timing characteristics, spatial characteristics, and load-storage prediction characteristics to make the final voltage stability margin prediction.

[0157] 2) The training of the prediction model includes: initializing model parameters; using the training set data for forward propagation; calculating the loss between the predicted value and the actual value; updating the model parameters by backpropagation; evaluating the model performance on the validation set; repeating the above steps until the preset number of training rounds or performance indicators are reached; and finally, using the test set for final evaluation.

[0158] 3) Acquisition of training data set:

[0159] Using the method of steps S10-S60, a voltage stability margin matrix of the power grid at a plurality of consecutive historical acquisition moments is obtained;

[0160] Collect relevant load forecast data and energy storage output forecast data, and use time series forecasting models for forecasting;

[0161] The historical grid voltage stability margin matrix sequence and the load-energy storage forecast data at the corresponding time are combined into training samples.

[0162] 4) Prediction process:

[0163] Inputting the grid voltage stability margin matrix sequence of a plurality of consecutive acquisition moments obtained in step S60 into the pre-trained prediction model;

[0164] Other data input into the model include: load forecast data for the corresponding time period, energy storage output forecast data, and grid topology information;

[0165] The model outputs the predicted value of the voltage stability margin of each node in multiple future time steps, forming a three-dimensional tensor.

[0166] Through the above steps, the voltage stability margin prediction model of the pre-trained power grid is used to output the voltage stability margin prediction value of each node in the future. These prediction results can provide a reference for power grid operation and dispatch, and help to timely discover and prevent potential risks of voltage instability.

[0167] The core idea of ​​the whole method is: first establish a grid topology model including distributed energy storage, then derive a set of mathematical equations describing voltage stability based on the model; then calculate the voltage stability margin of each node and convert it into a matrix form; finally, use a machine learning model to predict the voltage stability margin. This method fully considers the impact of factors such as grid topology, load changes and energy storage systems on voltage stability, and can more accurately predict the voltage stability state of the future grid.

[0168] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned distributed energy storage grid-connected voltage stability prediction method.

[0169] A third aspect of the present invention provides a distributed energy storage grid-connected voltage stability prediction system, which includes the above-mentioned computer-readable storage medium.

[0170] Specifically, the principle of the present invention is:

[0171] 1. Establish a power grid topology model including distributed energy storage and obtain parameter information of each node.

[0172] First, it is necessary to collect the basic topological information of the urban power grid of the research object, including various nodes (power-input nodes, power generation nodes, load nodes and energy storage nodes) and their connection relationships, as well as the voltage amplitude, phase angle, active power, reactive power, load power and energy storage capacity of each node. With this information, a mathematical model that can reflect the actual operation of the power grid is established.

[0173] 2. Derive a set of mathematical equations describing voltage stability based on the power grid model.

[0174] Based on the grid topology information and node parameters obtained in step 1, a set of mathematical equations including power balance equation, node voltage equation, energy storage state equation, power flow equation and voltage stability index equation is established. These equations can describe the impact of the distributed energy storage system on the voltage amplitude, phase angle and stability of each node in the grid. By solving these equations, the voltage stability margin of each node can be calculated, that is, the distance between the current voltage state of the node and the critical point of voltage instability.

[0175] 3. Use machine learning models to predict voltage stability margin.

[0176] In step 2, the voltage stability margin information of each node in the power grid at multiple consecutive moments is obtained and converted into a matrix form. This matrix representation not only includes the voltage stability of the node itself, but also reflects the mutual influence between nodes. A prediction model based on multi-task learning is designed, which can comprehensively consider factors such as time series characteristics, spatial correlation, and load and energy storage changes, and output the voltage stability margin prediction value of each node in the future period.

[0177] There are several reasons for adopting this approach:

[0178] First, the dynamic characteristics of distributed energy storage systems will have an important impact on the voltage stability of the power grid, so they must be included in the analysis. Only by establishing a power grid topology model that includes distributed energy storage can the actual operation of the power grid be described more accurately.

[0179] Secondly, the matrix representation can better reflect the interaction between power grid nodes. Compared with considering the stability of each node separately, this method can more comprehensively analyze the overall stability state of the power grid.

[0180] Finally, the prediction model based on machine learning has a strong adaptive learning ability and can effectively capture the complex spatiotemporal characteristics of grid voltage stability. Compared with traditional numerical calculation methods, this model has better generalization and prediction accuracy when processing large-scale, high-dimensional grid data.

[0181] In summary, the distributed energy storage urban power grid voltage stability prediction method proposed in the present invention fully considers the various complex factors in the actual operation of the power grid, and adopts a combination of mathematical modeling and machine learning to achieve effective prediction and early warning of the power grid voltage stability state, providing important support for the safe and stable operation of the power grid.

[0182] In order to better understand and implement the present invention, a specific embodiment 1 of the present invention is provided below. The specific steps of embodiment 1 are described in detail as follows:

[0183] Step S10: Establishing a city power grid topology map including distributed energy storage

[0184] The specific implementation method of this step is as follows:

[0185] First, we need to obtain the basic topological information of the urban power grid of the research object. This includes the types of nodes in the power grid, the connection relationship between nodes, and the relevant parameters of each node. Specifically, the node type includes the power-input node Power generation node Load Node and energy storage nodes The connection relationship between nodes can be expressed by edges. It is represented by the transmission path between nodes. Each node i has its own node parameters, including voltage amplitude |V i |、Phase angle θ i , Active power P i , reactive power Q i , Load active power P L,i , load reactive power Q L,i , Energy storage capacity And the charge and discharge power P S,i Etc. These parameters can be obtained through power grid modeling software or on-site measurements.

[0186] After obtaining the basic topology information of the power grid, the next step is to construct a power grid topology diagram that includes distributed energy storage. The specific approach is:

[0187] 1) Based on the acquired node and edge information, establish a schematic diagram of the topological structure of the power grid. This can be done by using the representation method of nodes and edges in graph theory, such as using an adjacency matrix A or an adjacency table to describe the connection relationship between nodes. The element A of the adjacency matrix A is ij Defined as:

[0188]

[0189] 2) Based on the topological structure, identify various nodes in the power grid, including the power-input nodes Power generation node Load Node and energy storage nodes These nodes are represented by different symbols or colors in the topology diagram.

[0190] 3) For each node i, obtain its corresponding node parameter value These parameter values ​​can be obtained through power flow calculation results of power grid simulation software, field measurement data or relevant literature.

[0191] Through the above steps, a topological diagram of the urban power grid including distributed energy storage is established, and the parameter information of each node is obtained. This topological diagram lays the foundation for subsequent analysis and calculation. The main purpose of this step is to build a mathematical model that can reflect the actual situation of the power grid and provide necessary input data for subsequent voltage stability analysis.

[0192] Step S20: Establishing the energy storage grid voltage influence equation group according to the node parameters

[0193] Based on the grid topology information and node parameters obtained in step S10, the purpose of this step is to establish a mathematical model that describes the impact of distributed energy storage on grid voltage stability. This model includes the following equations:

[0194] 1) Power balance equation:

[0195]

[0196] This equation describes the active power P of node i. i and reactive power Q i The balance between transmission line losses, load demand, and energy storage system injection power. i and V k is the complex voltage at nodes i and k, Y ik is the corresponding element in the node admittance matrix Y, P L,i and Q L,i is the active and reactive power of the load at node i, P S,i and Q S,i Active and reactive power injected into the energy storage system of node i.

[0197] 2) Node voltage equation:

[0198]

[0199] This set of equations describes the magnitude of the node voltage |V i | and phase angle θ i The changing rules over time are discussed, and the formulas for calculating the voltage amplitude and phase angle changing rate are given.

[0200] 3) Energy storage state equation:

[0201]

[0202] These equations describe the energy state changes of the energy storage system, including charging Discharge and self-discharge P loss,i process and limits the capacity of the energy storage system and power Boundary conditions. Among them, η c,i and η d,i are the charge and discharge efficiencies, respectively.

[0203] 4) Flow equation:

[0204]

[0205] This set of equations is used to calculate the active power P of each node i and reactive power Q i , where G ik and B ik is the real and imaginary part of the node admittance matrix Y, θ ik =θ i -θ k is the phase angle difference between nodes i and k.

[0206] 5) Voltage stability index equation:

[0207]

[0208]

[0209] VSM i =VSI i -VSI i,crit

[0210] This set of equations defines the voltage stability index L i and VSI i , and the voltage stability margin VSM is calculated i The formula is: ii is the diagonal element of the node admittance matrix Y, VSI i,crit It is the voltage stability index value of the critical stability point.

[0211] Through the above equations, a mathematical model describing the impact of distributed energy storage on grid voltage stability is established. This model can be used to analyze the voltage stability margin of each node in the grid and provide a basis for subsequent voltage stability prediction.

[0212] Step S30: Calculate the stability range of node parameters according to the energy storage grid voltage influence equation group

[0213] Based on the energy storage grid-connected voltage influence equations established in step S20, as well as the initial values ​​and constraints of each node parameter, the purpose of this step is to calculate the node parameter value range that ensures node voltage stability. The specific steps are as follows:

[0214] 1) Initialize the value range of each node parameter:

[0215] Input node voltage amplitude |V i | ±5% of the rated value, phase angle θ i is 0°, and the short-circuit capacity is determined according to the actual system parameters;

[0216] Active power of power generation node P i Between 0 and rated power, reactive power Q i Between -30% and +30% of rated power, the voltage amplitude is |V i | ±5% of the rated value;

[0217] Load node active load P L,i and reactive load Q L,i Determined based on historical data and load forecast, the voltage amplitude |V i | Allowed between 90% and 110% of the rated value;

[0218] Energy storage node charging and discharging power P S,i Not exceeding rated power Energy storage capacity Between 20% and 80%, the voltage amplitude |V i |±5% of the rated value.

[0219] 2) Substitute the initial value ranges of the above node parameters into the energy storage grid-connected voltage influence equation group established in step S20, and use numerical calculation methods to solve these equations to obtain the value ranges of each node parameter that meets the voltage stability constraint conditions.

[0220] 3) For a node parameter, if it exceeds the initial value range during the solution process, it means that the value range of the parameter is not enough to ensure voltage stability. At this time, it is necessary to appropriately adjust the value range of the parameter and recalculate until the parameter value range that meets the voltage stability requirements is found.

[0221] Through the above steps, the stable value range of each node parameter is obtained, which can ensure that the power grid maintains voltage stability when distributed energy storage is connected to the grid. These results provide necessary constraints for subsequent voltage stability prediction.

[0222] Step S40: Calculate the voltage stability margin based on the grid topology and node parameter stability range

[0223] In step S30, the stable value range of the parameters of each node of the power grid has been obtained. The purpose of this step is to further calculate the voltage stability margin of each node. The voltage stability margin is an important indicator to characterize the voltage stability of the node. It reflects the distance between the current voltage state of the node and the critical point of voltage instability. The specific implementation steps are as follows:

[0224] 1) According to the urban power grid topology established in step S10 and the parameter stability range of each node obtained in step S30, the energy storage grid voltage influence equation group in step S20 is solved by numerical calculation method to obtain the current voltage amplitude of each node |V i | and phase angle θ i .

[0225] 2) The current voltage state of each node (|V i |,θ i ) is substituted into the voltage stability index equation in step S20 to calculate the voltage stability index VSI of each node i .

[0226] 3) By continuously increasing the system load, the voltage stability index VSI is gradually solved using the calculation method in step 2 i , until a critical point occurs where the power flow calculation does not converge or the voltage collapses. The voltage stability index value corresponding to this critical point is recorded as VSI i,crit , which is the critical point of voltage instability.

[0227] 4) Based on the results obtained in steps 2 and 3, calculate the voltage stability margin VSM of each node i =VSI i -VSI i,crit The larger the voltage stability margin, the more stable the node voltage is, and vice versa, the closer the voltage is to the critical point of instability.

[0228] Through the above steps, the voltage stability margin value of each node in the power grid is obtained. These results can intuitively reflect the current voltage stability state of the power grid and provide necessary input data for subsequent voltage stability prediction.

[0229] Step S50: Convert the grid voltage stability margin information into a matrix representation

[0230] Based on the voltage stability margin of each node of the power grid calculated in step S40, the purpose of this step is to convert this information into a matrix form to provide input data for subsequent voltage stability prediction. The specific implementation steps are as follows:

[0231] 1) For each node i in the urban power grid topology diagram established in step S10, the voltage stability margin value VSM calculated by the node i is used. iThis forms a grid voltage stability margin topology diagram.

[0232] 2) Adopting the adjacency matrix method, the above-mentioned grid voltage stability margin topology graph is converted into a grid voltage stability margin matrix VSM. Specifically, the number of rows and columns of the matrix is ​​equal to the total number of nodes n in the grid, and the matrix element VSM ij represents the voltage stability margin value between node i and node j. If there is no direct connection between node i and node j, then VSM ij = 0. Diagonal elements VSM ii It represents the voltage stability margin of node i itself. The grid voltage stability margin matrix VSM can be expressed as:

[0233]

[0234] Through the above steps, the voltage stability margin information of each node in the power grid is converted into a matrix representation. Such a matrix not only contains the voltage stability margin of each node itself, but also reflects the mutual influence relationship between the voltage stability of the nodes. This provides a data form that is easy to process and analyze for subsequent voltage stability prediction.

[0235] Step S60: Repeat S40-S50 to obtain the grid voltage stability margin matrix at multiple consecutive acquisition moments.

[0236] In steps S40-S50, the voltage stability margin matrix of the power grid at a certain moment has been calculated and converted. In order to analyze the change law of power grid voltage stability over time, the purpose of this step is to repeatedly execute steps S40-S50 to obtain a sequence of power grid voltage stability margin matrices at multiple consecutive acquisition moments.

[0237] The specific steps are as follows:

[0238] 1) Set a suitable sampling time interval Δt, for example, sampling every 15 minutes.

[0239] 2) For each sampling time t, repeat steps S40-S50 to calculate and transform the grid voltage stability margin matrix VSM(t) at that time.

[0240] 3) Arrange the grid voltage stability margin matrices {VSM(t), VSM(t+Δt), …, VSM(t+mΔt)} obtained by continuous sampling at multiple moments in time in chronological order to form a three-dimensional tensor VSM(t), where the first dimension corresponds to the time series, the second dimension corresponds to the grid nodes, and the third dimension corresponds to the voltage stability margin value of each node. The three-dimensional tensor VSM(t) can be expressed as:

[0241]

[0242] Through the above steps, the change process of the voltage stability margin of the power grid at multiple consecutive moments is obtained, which provides the required time series data for the subsequent voltage stability prediction. These data not only include the time-varying characteristics of the voltage stability margin of each node itself, but also reflect the dynamic characteristics of the mutual influence of voltage stability between nodes.

[0243] Step S70: Predict using the pre-trained grid voltage stability margin prediction model

[0244] Based on the grid voltage stability margin matrix sequence of multiple consecutive acquisition moments obtained in step S60, combined with the prediction data of load and energy storage system, the purpose of this step is to use the pre-trained grid voltage stability margin prediction model to output the voltage stability margin prediction value of each node in the future. The specific implementation steps are as follows:

[0245] 1) Structure of the grid voltage stability margin prediction model:

[0246] The prediction model adopts a multi-task learning structure, including the following four sub-networks:

[0247] Time series feature extraction sub-network: A recurrent neural network based on the long short-term memory network (LSTM) is used to capture the changing law of the voltage stability margin time series. The input of the LSTM network is the grid voltage stability margin matrix sequence VSM(t) at multiple consecutive acquisition moments, and the output is the time series feature vector h t .

[0248] Spatial feature extraction subnetwork: Based on graph convolutional neural network (GCN), it is used to extract spatial correlation in the power grid topology. The input of the GCN network is the power grid voltage stability margin matrix VSM(t) at each acquisition time, and the output is the spatial feature vector h s .

[0249] Change factor processing subnetwork: a multi-layer feedforward neural network, used to consider the impact of load and energy storage system changes on voltage stability. The input of this subnetwork is the load forecast {P L,i ,Q L,i} and energy storage processing prediction Data, output is the characteristic vector of the change factor h c .

[0250] Fusion sub-network: multi-layer feedforward neural network, used to integrate temporal features h t , spatial features h s and the changing factor characteristics h c , and the final voltage stability margin prediction is performed. The output of this subnetwork is the voltage stability margin prediction value of each node in the next m time steps.

[0251] 2) Training of prediction model:

[0252] Initialize model parameters

[0253] Forward propagation using training set data

[0254] Calculate predicted values The loss function between the actual value VSM

[0255] Update the model parameters through the back-propagation algorithm to minimize the loss function

[0256] Evaluate model performance on the validation set, such as prediction accuracy, etc.

[0257] Repeat the above steps until the preset number of training rounds or performance indicators are reached

[0258] Final evaluation using the test set

[0259] 3) Acquisition of training data set:

[0260] Using the method of steps S10-S60, a grid voltage stability margin matrix sequence {VSM(t), VSM(t+Δt), ..., VSM(t+mΔt)} of multiple consecutive historical acquisition moments is obtained.

[0261] Collect relevant load forecast data L,i ,Q L,i} and energy storage output forecast data Use time series forecasting models (such as ARIMA, Prophet, or LSTM networks) for forecasting

[0262] The historical grid voltage stability margin matrix sequence and the corresponding time load-storage forecast data are used to form training samples

[0263] 4) Prediction process:

[0264] The grid voltage stability margin matrix sequence VSM(t) obtained in step S60 at multiple consecutive acquisition moments is input into the pre-trained prediction model

[0265] Other data input into the model include: load forecast data for the corresponding time period {P L,i ,Q L,i}、Energy storage output forecast data And the grid topology information A

[0266] The model outputs the predicted value of voltage stability margin for each node in the next m time steps. This forms a three-dimensional tensor.

[0267] Through the above steps, the voltage stability margin prediction model of the pre-trained power grid is used to output the voltage stability margin prediction value of each node in the future. These prediction results can provide a reference for power grid operation and dispatch, and help to timely discover and prevent potential risks of voltage instability.

[0268] The core idea of ​​this embodiment 1 is: first establish a grid topology model including distributed energy storage, then derive a set of mathematical equations describing voltage stability based on the model; then calculate the voltage stability margin of each node and convert it into a matrix form; finally, use a machine learning model to predict the voltage stability margin. This method fully considers the impact of factors such as grid topology, load changes, and energy storage systems on voltage stability, and can more accurately predict the voltage stability state of the future grid.

[0269] In order to further better understand and implement the present invention, the following provides an embodiment 2 of a specific application scenario of the present invention: The urban power grid of a certain city covers an area of ​​about 500 square kilometers, and the power supply load mainly includes three categories of users: industry, commerce and residents. The power grid adopts a 220kV high-voltage transmission network and supplies power to users through 110kV and 10kV substations. In order to improve the reliability of power supply, a large number of distributed photovoltaic power generation and energy storage systems have been deployed in the power grid.

[0270] According to data provided by the grid operator, the basic topology of the city's urban power grid has a total of 40 nodes, including 5 220kV power supply nodes, 10 110kV power generation nodes, 20 10kV load nodes and 5 distributed energy storage nodes. These nodes are interconnected through 60 transmission lines.

[0271] In order to analyze the voltage stability of this power grid, it is necessary to obtain the specific parameter information of each node. Through the power flow calculation of the power grid modeling software and the field measurement data, the data shown in Table 1 were obtained.

[0272] Table 1. Grid node parameters

[0273] Node Type Number of nodes Voltage(kV) Active power(MW) Reactive power (Mvar) Energy storage capacity (MWh) Power-on Node 5 220±5％ - - - Power generation node 10 110±5％ 0～500 -150～150 - Load Node 20 10±10％ 0～50 -20～20 - Energy storage node 5 10±5％ -20～20 -10～10 10～40

[0274] Next, according to the method proposed by the present invention, the voltage stability of the city power grid is analyzed and predicted.

[0275] Step S10: Establishing a city power grid topology map including distributed energy storage

[0276] according to Figure 1Based on the information in Table 1, a topological diagram of an urban power grid including distributed energy storage can be established. The diagram reflects the connection relationship between various nodes in the power grid and identifies the power-in nodes, power generation nodes, load nodes, and energy storage nodes. At the same time, the relevant parameters of each node are also obtained, including voltage amplitude, active power, reactive power, load power, and energy storage capacity.

[0277] Step S20: Establishing the energy storage grid voltage influence equation group according to the node parameters

[0278] Based on the urban power grid topology information and node parameters obtained in step S10, the following mathematical equations describing the impact of distributed energy storage on voltage stability are established:

[0279] 1) Power balance equation:

[0280] Taking node i as an example, its active power P i and reactive power Q i Should meet:

[0281]

[0282] Among them, V i and V k are the complex voltages at nodes i and k, respectively, ik is the element of the node admittance matrix, P L,i and Q L,i is the active and reactive power of the load at node i, P S,i and Q S,i Active and reactive power injected into the energy storage system of node i.

[0283] 2) Node voltage equation:

[0284] The voltage at node i can be expressed as Where |V i | is the voltage amplitude, θ i is the voltage phase angle. The rate of change of voltage amplitude and phase angle can be expressed as:

[0285]

[0286] 3) Energy storage state equation:

[0287] Energy state E of energy storage system at node i S,i The changes can be expressed as:

[0288]

[0289] in, and are the charging and discharging power of the energy storage system, η c,i and η d,iis the charge and discharge efficiency, P loss,i is the self-discharge loss.

[0290] 4) Flow equation:

[0291] Active power P of node i i and reactive power Q i It can be calculated by the following formula:

[0292]

[0293] Among them, G ik and B ik are the real and imaginary parts of the node admittance matrix, θ ik is the phase angle difference between nodes i and k.

[0294] 5) Voltage stability index equation:

[0295] Using voltage stability index VSI i To characterize the voltage stability of node i, it is defined as:

[0296]

[0297] Voltage Stability Margin VSM i It can be expressed as:

[0298] VSM i =VSI i -VSI i,crit

[0299] Among them, VSI i,crit is the critical stability point index value of node i.

[0300] By establishing such a set of mathematical equations, the impact of distributed energy storage systems on the voltage stability of each node in the power grid can be analyzed.

[0301] Step S30: Calculate the stability range of node parameters according to the energy storage grid voltage influence equation group

[0302] Based on the above voltage influence equations, combined with the initial value range of the node parameters given in Table 1, the node parameter stability range that ensures the stability of the grid voltage can be calculated, as follows:

[0303] 1) Power-on node:

[0304] Voltage Amplitude|V i | ±5% of 220kV, i.e. 209kV to 231kV; Phase angle θ i is 0°; the short-circuit capacity is determined according to the actual system parameters.

[0305] 2) Power generation node:

[0306] Active power P i Between 0MW and 500MW; reactive power Q i Between -150Mvar and 150Mvar; voltage amplitude |V i |±5% of 110kV, i.e. 104.5kV to 115.5kV.

[0307] 3) Load Node:

[0308] Active load P L,i and reactive load Q L,i Determined based on historical data and load forecast, the voltage amplitude |V i | The range of ±10% of 10kV is allowed, i.e. 9kV to 11kV.

[0309] 4) Energy storage nodes:

[0310] Charge and discharge power P S,i No more than ±20MW, energy storage capacity Between 10MWh and 40MWh, the voltage amplitude is |V i | ±5% of 10kV.

[0311] Through the above steps, the range of each node parameter that meets the voltage stability constraint is obtained. These results will serve as the basis for subsequent voltage stability analysis and prediction.

[0312] Step S40: Calculate the voltage stability margin based on the grid topology and node parameter stability range

[0313] Using the node parameter stability range obtained in step S30, the voltage influence equations are solved by numerical calculation to obtain the node voltage state of the urban power grid at a certain moment, including the voltage amplitude |V i | and phase angle θ i .

[0314] Then, these voltage state parameters are substituted into the voltage stability index equation to calculate the voltage stability index VSI of each node. i Then, by continuously increasing the system load until the critical point where the power flow calculation does not converge or the voltage collapses, the voltage stability index value VSI at the critical point is obtained. i,crit .

[0315] Finally, based on the node's current VSI i and critical point VSI i,crit The voltage stability margin VSM of each node is calculated i =VSI i -VSI i,critThe following Table 2 lists the calculation results of the voltage stability margin of each node in the urban power grid at a certain moment:

[0316] Table 2. Voltage stability margin calculation results

[0317] node <![CDATA[VSI i ]]> <![CDATA[VSI i,crit ]]> <![CDATA[VSM i ]]> 1 0.952 0.890 0.062 2 0.970 0.910 0.060 ... ... ... ... 40 0.925 0.860 0.065

[0318] As can be seen from the table, the voltage stability margin of most nodes in the power grid is around 0.06, indicating that the overall voltage of the power grid is relatively stable. However, there are also some nodes with lower stability margins, such as the VSM at node 40. i It is only 0.065, which means that the voltage of this node is close to the critical point of instability and needs special attention.

[0319] Step S50: Convert the grid voltage stability margin information into a matrix representation

[0320] Based on the grid node voltage stability margin calculated in step S40, it is converted into the form of a grid voltage stability margin matrix VSM. The number of rows and columns of this matrix is ​​40, representing 40 nodes in the grid. The matrix element VSM ij Represents the voltage stability margin value between node i and node j. If there is no direct connection between the two nodes, then VSM ij = 0. Diagonal elements VSM ii It represents the voltage stability margin of node i itself. The voltage stability margin matrix of the city power grid at a certain moment is given below:

[0321]

[0322] This matrix representation not only includes the voltage stability of each node itself, but also reflects the mutual influence relationship between nodes. The subsequent voltage stability prediction analysis will be based on this voltage stability margin matrix.

[0323] Step S60: Repeat S40-S50 to obtain the grid voltage stability margin matrix at multiple consecutive acquisition moments.

[0324] In order to analyze the law of grid voltage stability over time, the city grid is sampled every 15 minutes, and steps S40-S50 are repeated to obtain a voltage stability margin matrix sequence for 96 consecutive moments (one day).

[0325] According to the temporal variation trend of the collected voltage stability margin matrix, it can be judged that the voltage stability margin of different nodes in the power grid shows obvious dynamic changes over time. During the morning peak period, the voltage stability margin of some load nodes decreases, while at night when the load is low, the stability margin of most nodes increases. This time series feature provides an important basis for subsequent voltage stability prediction.

[0326] At the same time, it is also observed that the voltage stability between nodes affects each other. For example, when the energy storage system of a node changes in charge and discharge, the voltage stability margin of other nodes connected to it will also fluctuate. This spatial correlation feature will also be taken into consideration in the prediction model.

[0327] Step S70: Predict using the pre-trained grid voltage stability margin prediction model

[0328] In step S60, the voltage stability margin matrix sequence of the city power grid at 96 consecutive moments is obtained. Combined with the forecast data of the load and energy storage system, the pre-trained grid voltage stability margin prediction model will be used to predict the voltage stability margin of each node in the future.

[0329] 1) Structure of the grid voltage stability margin prediction model

[0330] The prediction model adopts a multi-task learning structure, which mainly includes the following four sub-networks:

[0331] Time series feature extraction subnetwork: This is a recurrent neural network based on the long short-term memory network (LSTM), which is used to capture the changing law of the voltage stability margin time series. The input of this subnetwork is the grid voltage stability margin matrix sequence {VSM(t-95), VSM(t-94), …, VSM(t)} obtained in step S60 for 96 consecutive moments, and the output is the time series feature vector h t .

[0332] Spatial feature extraction subnetwork: This is a subnetwork based on graph convolutional neural network (GCN) to extract spatial correlation in the power grid topology. The input of this subnetwork is the grid voltage stability margin matrix VSM(t) at each moment, and the output is the spatial feature vector h s .

[0333] Change factor processing subnetwork: This is a multi-layer feedforward neural network used to consider the impact of load and energy storage system changes on voltage stability. The input of this subnetwork is load forecast data {P L,i ,Q L,i} and energy storage output forecast data The output is the characteristic vector h of the change factor c .

[0334] Fusion sub-network: This is also a multi-layer feedforward neural network used to integrate the temporal features h t , spatial features h s and the changing factor characteristics h c, and the final voltage stability margin prediction is performed. The output of this subnetwork is the voltage stability margin prediction value of each node in the next m time steps.

[0335] 2) Training of prediction models

[0336] The following steps are used to train the grid voltage stability margin prediction model:

[0337] Step 1: Initialize model parameters

[0338] Step 2: Use the training set data for forward propagation and calculate the predicted value The loss function between the actual value VSM

[0339] Step 3: Update the model parameters to minimize the loss function through the back propagation algorithm

[0340] Step 4: Evaluate model performance on the validation set, such as prediction accuracy, mean absolute error, etc.

[0341] Step 5: Repeat steps 2-4 until the preset number of training rounds or performance indicators are reached

[0342] Step 6: Final evaluation using the test set

[0343] 3) Acquisition of training data set

[0344] Using the method of steps S10-S60, the voltage stability margin matrix sequence {VSM(t-95), VSM(t-94), ..., VSM(t)} of the urban power grid in the past 96 time steps is obtained as input-output samples.

[0345] At the same time, the load forecast data for the corresponding time period {P L,i ,Q L,i} and energy storage output forecast data These forecast data can be obtained through time series forecasting models (such as ARIMA, Prophet or LSTM networks), taking into account the impact of seasonality, cyclicality and external factors (such as weather, holidays, etc.).

[0346] Finally, the historical grid voltage stability margin matrix sequence and the load-storage forecast data of the corresponding time are used as training samples as input features; the voltage stability margin matrix of the next m time steps is used as the prediction target output.

[0347] 4) Prediction process

[0348] After completing the model training, the prediction model can be used to predict the future grid voltage stability. The specific steps are as follows:

[0349] Step 1: Input the grid voltage stability margin matrix sequence {VSM(t-95), VSM(t-94), …, VSM(t)} of the most recent 96 time steps obtained in step S60 into the pre-trained prediction model

[0350] Step 2: Other data input into the model include: load forecast data for the corresponding time period {P L,i ,Q L,i}、Energy storage output forecast data And the grid topology information A

[0351] Step 3: The model outputs the predicted value of the voltage stability margin of each node in the next m time steps Form a three-dimensional tensor

[0352] Through the above steps, the pre-trained grid voltage stability margin prediction model is used to output the voltage stability margin prediction value of each node in the future. These prediction results can provide important reference for grid dispatchers and help to timely discover and prevent potential risks of voltage instability.

[0353] The following Table 3 is the prediction result of the voltage stability margin of the city power grid at a certain time point:

[0354] Table 3. Voltage stability margin prediction results

[0355]

[0356] From the prediction results, it can be seen that the overall voltage stability margin of the power grid will show a downward trend in the next three hours. In particular, the voltage stability margin of node 40 has dropped from the current 0.065 to 0.054, which is very close to the critical point of instability. This indicates that there may be a risk of voltage instability at this node in the future, and the power grid dispatchers need to take necessary control measures, such as adjusting the node load and increasing the output of the energy storage system, to maintain the safe and stable operation of the power grid.

[0357] In summary, in Example 2, the distributed energy storage urban power grid voltage stability prediction method proposed by the present invention can not only analyze the current power grid stability state, but also predict the voltage stability margin change trend of each node in the future. This provides important decision support for power grid dispatchers, helps to take targeted control measures, and improves the safety and reliability of the power grid.

[0358] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A distributed energy storage grid-connected voltage stability prediction method, characterized in that: The following steps are involved: S10, establishing a topological diagram of an urban power grid including distributed energy storage, wherein the topological diagram includes nodes and edges, wherein the nodes include power-input nodes, power generation nodes, load nodes, and energy storage nodes, and the edges represent power transmission paths between nodes, and obtaining the value of a node parameter of each node; S20, establishing a group of energy storage grid-connected voltage influence equations according to the node parameters, including a power balance equation, a node voltage equation, an energy storage state equation, a power flow equation, and a voltage stability index equation; S30, calculating the node parameter stability range of each node according to the energy storage grid-connected voltage influence equation group and the initial values ​​and constraints of each node parameter, wherein the node parameter stability range refers to the range of node parameter values ​​that ensures the node voltage stability; S40, based on the urban power grid topology, the stability range of each node parameter and the current voltage state of the node acquired in real time, calculating the voltage stability margin of each node, wherein the voltage stability margin represents the distance between the current voltage state of the node and the critical point of voltage instability; S50, setting the value of each node of the urban power grid topology map as the voltage stability margin of the node to form a power grid voltage stability margin topology map, and converting the power grid voltage stability margin topology map into a power grid voltage stability margin matrix using an adjacency matrix method; S60, repeatedly executing S40-S50 to obtain a grid voltage stability margin matrix at a plurality of consecutive acquisition moments; S70, input the obtained grid voltage stability margin matrix of multiple consecutive acquisition moments into a pre-trained grid voltage stability margin prediction model, and output the voltage stability margin prediction value of each node in the future period of time as the distributed energy storage grid-connected voltage stability prediction result.

2. A distributed energy storage grid-connected voltage stability prediction method according to claim 1, characterized in that: The node parameters include power-in node parameters, power generation node parameters, load node parameters and energy storage node parameters. The power-in node parameters include voltage amplitude, phase angle and short-circuit capacity. The power generation node parameters include active power, reactive power and voltage amplitude. The load node parameters include active load, reactive load and voltage amplitude. The energy storage node parameters include charging and discharging power, energy storage capacity and voltage amplitude.

3. A distributed energy storage grid-connected voltage stability prediction method according to claim 1, characterized in that: The initial values ​​and constraints of the parameters of each node are as follows: the voltage amplitude of the power-in node is ±5% of the rated value, the phase angle is 0°, and the short-circuit capacity is determined according to the actual system parameters; the active power of the power generation node is between 0 and the rated power, the reactive power is between -30% and +30% of the rated power, and the voltage amplitude is ±5% of the rated value; The active load and reactive load of the load node are determined based on historical data and load forecasts, and the voltage amplitude is allowed to be between 90% and 110% of the rated value; the charging and discharging power of the energy storage node does not exceed the rated power, the energy storage capacity is between 20% and 80%, and the voltage amplitude is ±5% of the rated value.

4. A distributed energy storage grid-connected voltage stability prediction method according to claim 1, characterized in that: The voltage state refers to a complex number representation consisting of the magnitude and phase angle of the node voltage.

5. A distributed energy storage grid-connected voltage stability prediction method according to claim 1, characterized in that: The method for obtaining the voltage instability critical point is: through continuous power flow calculation and PV curve analysis, gradually increasing the system load until the power flow calculation does not converge or the voltage collapse point appears, and the non-convergence or voltage collapse point is the voltage instability critical point.

6. A distributed energy storage grid-connected voltage stability prediction method according to claim 1, characterized in that: The grid voltage stability margin prediction model adopts a multi-task learning structure, including a time series feature extraction subnetwork, a spatial feature extraction subnetwork, a change factor processing subnetwork and a fusion subnetwork.

7. A distributed energy storage grid-connected voltage stability prediction method according to claim 6, characterized in that: The input of the time series feature extraction subnetwork is the grid voltage stability margin matrix at multiple consecutive acquisition moments, and the output is a time series feature vector, which is used to capture the time variation law of the voltage stability margin. Its structure is a recurrent neural network based on a long short-term memory network; The input of the spatial feature extraction subnetwork is the grid voltage stability margin matrix at each acquisition moment, and the output is a spatial feature vector, which is used to extract the spatial correlation in the grid topology structure, and its structure is based on a graph convolutional neural network; The input of the change factor processing subnetwork is load forecast and energy storage processing forecast data, and the output is a load-energy storage forecast feature vector, which is used to consider the impact of load changes on voltage stability. Its structure is a multi-layer feedforward neural network; The input of the fusion subnetwork is the concatenation of the time series feature vector, the spatial feature vector and the load-energy storage prediction feature vector, and the output is the predicted value of the voltage stability margin of each node in the future period, which is used to integrate various features for final prediction. Its structure is a multi-layer feedforward neural network.

8. A distributed energy storage grid-connected voltage stability prediction method according to claim 7, characterized in that: The frequency of repeating S40-S50 is once every 15 minutes.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, which, when executed, are used to execute a distributed energy storage grid-connected voltage stability prediction method according to any one of claims 1 to 8.

10. A distributed energy storage grid-connected voltage stability prediction system, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images