New energy power system power flow prediction method considering source-network-load uncertainty

Abstract

The invention relates to a new energy power system power flow prediction method considering source-network-load uncertainty, and the method comprises the specific steps: carrying out the normalization processing of the historical parameters of a power grid, extracting multi-dimensional track features through variational mode decomposition, and constructing an operation trend set; a cellular automaton model is combined to simulate a spatio-temporal evolution law, and a fault feature recognition mechanism is optimized; constructing a trend state classification database based on the residual transformation rate; dynamically distributing state contribution weights by using a multi-constraint regression model, and establishing a prediction model; and iteratively optimizing the model through error feedback, and constructing a vulnerability evaluation system based on sensitivity and interference analysis. According to the method, multi-source data and dynamic weight distribution are fused, the problems that traditional prediction precision is low and fault response is lagged are solved through a bidirectional feedback optimization framework, the accuracy of power grid trend prediction and vulnerability assessment can be improved, and support is provided for security defense.

Description

Technical Field

[0001] The present invention relates to the technical field of power system control, and particularly relates to a new energy power system power flow prediction method considering the uncertainties of power sources, grids, and loads. Background Art

[0002] With the large-scale development of renewable clean energy and the commissioning of the "Three-North China" UHV synchronized power grid and the UHV DC transmission projects of large energy bases, the operation and dispatching of the power grid face huge challenges. The number of levels of interconnected large power grids increases, the electrical connection is enhanced, the operating characteristics of the power grid become more complex, and the difficulty of power grid dispatching and control increases significantly. The access of large-scale new energy sources such as wind and light, distributed generation on the demand side, and the rapid development of electric vehicles bring great uncertainties, and the uncertainties of power grid dispatching and operation control are significantly enhanced. In addition, external extreme disaster weather, such as natural disasters like ice storms, typhoons, earthquakes, and floods, seriously affects the safe and stable operation of the power grid. Summary of the Invention

[0003] In order to cope with the uncertainties brought by the high-penetration access of renewable energy to the power grid and random demand-side responses to the power grid operation. By means of artificial intelligence algorithms such as machine learning and deep learning, combined with regulation big data, through the learning of a large amount of training sample data, predicting the future operation trend of the power grid power flow has important practical significance for ensuring the safe and stable operation of the power grid. Based on this, the present invention provides a new energy power system power flow prediction method considering the uncertainties of power sources, grids, and loads, including the following steps: Preprocess the historical parameters of the power grid, extract multi-dimensional trajectory features through variational mode decomposition, and construct a power grid operation trend set; Combine the cellular automaton model to simulate the spatio-temporal evolution law, iteratively optimize the power grid operation trend model, and obtain the power grid operation trend evolution mechanism and mathematical criterion under different fault states; Construct a state recognition database for the power grid operation trend based on the residual transformation rate; Use a multi-constraint regression model to dynamically allocate state contribution weights and establish a power grid operation trend prediction model; Iteratively optimize the model through error feedback, and construct a vulnerability evaluation system based on sensitivity and interference analysis to complete the quantitative analysis of the power grid vulnerability.

[0004] Specifically, the preprocessing of the historical parameters of the power grid, extracting multi-dimensional trajectory features through variational mode decomposition, and constructing a power grid operation trend set specifically includes: Normalize the historical operation parameter trajectory of the power grid, combine it with the uncertainty analysis model of the power grid operation parameter trajectory, analyze the specificity between the electrical quantities of the power grid operation and the external parameters affecting the power grid operation state under different power grid operation states, classify the power grid parameters using the mathematical statistics model and clustering algorithm, and determine the power grid operation parameter trajectory clusters composed of the electrical quantities and external parameters of the power grid operation under different categories; based on the variational mode decomposition algorithm, extract the frequency domain characteristics of the power grid operation parameter trajectory at different central frequencies, form a set of power grid operation trajectory characteristic information, define the power grid operation parameter trajectory clusters and their multi-dimensional characteristic sets as the power grid operation trend, and fuse data from different sources, different time scales, and different qualities to form a set of power grid operation trends.

[0005] Specifically, the set of power grid operation trends has the following mathematical expression: ; In the formula, refers to the power grid operation parameter trajectory cluster, refers to the n th node at time t The set expression of the power grid operation parameter trajectory cluster at the place, where refers to the q th power grid operation parameter at the n th node in t The value at time T is the power grid operation trend analysis period; , refers to the set of power grid operation trajectory cluster characteristics, refers to a power grid operation trend analysis period T The set of power grid operation parameter characteristics of the n th node within, where refers to a power grid operation trend analysis period T The n th node within q The j th characteristic of the is the set of power grid operation trends.

[0006] Specifically, the combination of the cellular automaton model to simulate the spatio-temporal evolution law, iteratively optimize the power grid operation trend model, and obtain the power grid operation trend evolution mechanism and mathematical criterion under different fault states specifically includes: According to the calculated set of power grid operation trends, combined with the basic laws of the spatio-temporal evolution of power grid operation trends, the Agent-Cellular Automata model is adopted. Constrained by the spatial distribution and time evolution law of power grid operation trends, it simulates the dynamic spatio-temporal distribution process of power grid operation trends, depicts the law of the change of power grid operation trends with time in the same spatial region, analyzes the time-domain change characteristics of power grid operation trends, and depicts the spatial distribution law of power grid operation trends in different spatial regions at the same moment, analyzes the spatial correlation characteristics of power grid operation trends. Combining the differences in the evolution models under normal and fault conditions, the power grid operation trend model is iteratively optimized to accurately identify the set of characteristic parameters of power grid operation trends in the fault state, and finally the evolution mechanism and mathematical criteria of power grid operation trends under different fault states are obtained.

[0007] Specifically, the Agent-Cellular Automata model integrates the statistical laws of power grid operation trends at the macroscopic level and the interaction behaviors among various Agents at the microscopic level. The power grid spatial region is discretized into a two-dimensional lattice grid to form the following spatial pattern: ; Among them, represents the cell in the \(i\)-th row and \(j\)-th column. Each cell represents a geographical area in the power grid. The model uses a discrete time step for dynamic evolution; Each cell contains one or more agents, defined as: ; Among them, represents the \(k\)-th Agent in the cell and represents key power grid components such as load, generator, and bus in this area; k Each Agent has a state variable at time t to represent the characteristic quantity of power grid operation trends. The state transition rule is: ; Among them: represents the external influence factors; is the state transition function constructed based on the physical characteristics of the power grid and the behavior mechanism of the Agent; is the state set of the cell and its neighborhood, defined as the Moore neighborhood: ; The time evolution of power grid operation trends depends on the interaction of the surrounding neighborhood and external disturbances. The specific model expression is: ; Among them: α and β are the weight coefficients of neighborhood interaction and external perturbation respectively; is the spatial influence weight of the cell pair in the neighborhood ( m , n ) on the central cell ; At any moment t , the spatial distribution of the power grid operation trend is constrained by the spatial autocorrelation characteristics, and the following spatial variogram is defined: ; Among them: is the normalized spatial weight, reflecting the influence intensity between different neighborhood units; The model defines two types of state transition functions respectively: the normal operating condition evolution function and the fault operating condition evolution function ; The difference between the two is expressed as: ; The model is iteratively optimized through historical operation data to minimize the deviation between the prediction and the observed data: ; Among them, is the historical observed data, is the set of parameters to be optimized; After the optimization is completed, the characteristic parameter set under the fault operating condition is extracted, and based on this, the evolution mechanism and mathematical criterion of the power grid under different fault types are formed: ; Among them, and are the trend mutation threshold and the spatial instability threshold respectively, which are used to accurately identify power grid faults.

[0008] Specifically, the state recognition database for constructing the power grid operation trend based on the residual transformation rate specifically includes: according to the mathematical criterion of the power grid operation trend, combined with the characteristic parameter set of the power grid operation trend, a trend-state space discrimination index is defined, and considering the state evolution characteristics of the power grid in the cluster, the power grid operation trend characteristics in different spatial regions, and the trend-state space discrimination index, a power grid operation trend state classification model is constructed; arranged in descending order of the discrimination index, with the minimum residual transformation rate divided by the power grid operation trend state as the optimization goal, the power grid operation trend state set with the minimum residual transformation rate is determined, and finally the state recognition database of the power grid operation trend is formed.

[0009] Specifically, the defined trend-state space discrimination index is as follows: ; where: : The trend feature of the i,j area at time t; : The time mean of the trend of the entire regional power grid; : The i , j spatial dispersion of the regional trend; : The abnormal deviation in the trend evolution of the i,j area; : The weight factor of each characteristic index; Based on the set of discrimination indices of all cells , sort according to the discrimination index from large to small: ; where is the k largest discrimination index; According to the sorting result, divide different power grid operation trend state clusters , where L is the number of initially divided state categories; Introduce the residual transformation rate as the optimization objective; define as follows: ; where: is the mean value of the discrimination index within the l th trend state cluster; is the mean value of the global discrimination index; the goal is to find a trend state division scheme that minimizes η , denoted as: ; Among them, the division process uses the dynamic partitioning method, combines hierarchical clustering with trend-state space characteristics for iterative partitioning, and satisfies ; After completing the optimal trend state division, obtain the trend state database: ; where is a set of characteristic data containing various power grid operation trend states.

[0010] Specifically, the method of dynamically allocating state contribution weights using a multi-constraint regression model to establish a power grid operation trend prediction model specifically includes: according to the power grid operation trend prediction results, calculating the error sequence of the power grid operation trend prediction, defining the prediction error trend factor, and if the prediction error shows an increasing trend, updating the power grid operation trend evolution model; if the prediction error shows a decreasing or normal fluctuation trend, maintaining the current power grid operation trend evolution model unchanged at the next prediction moment; further, based on the error sequence, analyzing the response characteristics of the power grid to small disturbances under different operation trend states, proposing a sensitivity evaluation index for the power grid operation trend trajectory, and combining the influence of internal and external system disturbances, establishing a power grid vulnerability evaluation index, using the entropy weight method to weight each trajectory, constructing a power grid vulnerability evaluation model, and finally completing the quantitative analysis of the power grid vulnerability.

[0011] Specifically, the power grid vulnerability evaluation model is defined as follows: ; where is the sensitivity of the power grid operation parameter trajectory cluster , is the sensitivity of the characteristics of the power grid operation parameter trajectory cluster , is the actual value of the disturbance parameter trajectory, is the average value of the disturbance parameter trajectory; ; where is the power grid vulnerability index, is the weight assigned by the entropy weight method for this parameter, is the incidence rate of the instability event of this parameter in the face of disturbances, is the upper bound of the safe operation of the trajectory, is the average value of the trajectory when the power grid operates stably.

[0012] Advantages of the present invention: The present invention integrates multi-source data and dynamic weight allocation, and through a two-way feedback optimization framework, solves the problems of low traditional prediction accuracy and lagging fault response, can improve the accuracy of power grid trend prediction and vulnerability assessment, and provides support for security defense. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention has the following drawings: Figure 1 is a method for power flow prediction of a new energy power system considering the uncertainties of the source-grid-load; Figure 2 is a schematic diagram of the structure of the power grid operation trend set. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] To make the objectives, technical solutions, and advantages of this application more clear and understandable, the following will be described in conjunction with specific embodiments.

[0015] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the embodiments of this application should have the ordinary meanings understood by those with ordinary skills in the field to which this application belongs. The "first", "second", and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0016] In view of this, as Figure 1 shown, the embodiments of the present invention application propose a power flow prediction method for a new energy power system considering the uncertainties of the source-grid-load, and the implementation steps are as follows: Step A. Normalize the historical operation parameter trajectories of the power grid, combine the uncertainty analysis model of the power grid operation parameter trajectories, study and analyze the specificities between the electrical quantities of the power grid operation and the external parameters affecting the power grid operation state under different power grid operation states (normal state and fault state), classify the power grid parameters using a mathematical statistics model and a clustering algorithm, and determine the power grid operation parameter trajectory clusters composed of the electrical quantities of the power grid operation and the external parameters under different categories. Based on the variational mode decomposition algorithm, extract the frequency-domain characteristics of the power grid operation parameter trajectories at different central frequencies, form a set of power grid operation trajectory characteristic information, define the power grid operation parameter trajectory clusters and their multi-dimensional characteristic sets as the power grid operation trend, and fuse data from different sources, different time scales, and different qualities to form a set of power grid operation trends, laying a foundation for the prediction of the power grid operation trend.

[0017] Step B. Based on the calculated set of power grid operation trends, combined with the basic spatio-temporal evolution law of power grid operation trends, adopt the Agent-Cellular Automata model. Constrained by the spatio-temporal distribution and time evolution law of power grid operation trends, simulate the dynamic spatio-temporal distribution process of power grid operation trends, depict the law of change of power grid operation trends over time in the same spatial area, analyze the time-domain change characteristics of power grid operation trends, and depict the spatial distribution law of power grid operation trends in different spatial areas at the same moment, analyze the spatial correlation characteristics of power grid operation trends. Combine the differences in the evolution models under normal and fault conditions, iteratively optimize the power grid operation trend model, accurately identify the set of characteristic parameters of power grid operation trends in the fault state, and finally obtain the evolution mechanism and mathematical criterion of power grid operation trends under different fault states.

[0018] Step C. According to the mathematical criterion of power grid operation trends, combined with the set of characteristic parameters of power grid operation trends, define the trend-state space discrimination index. Considering comprehensively the state evolution characteristics of the power grid in the cluster, the power grid operation trend characteristics in different spatial areas, and the trend-state space discrimination index, construct a power grid operation trend state classification model. Arrange in descending order of the discrimination index, and take the minimum residual transformation rate of the power grid operation trend state division as the optimization goal to determine the set of power grid operation trend states that minimize the residual transformation rate, and finally form a state recognition database of power grid operation trends.

[0019] Step D. According to the power grid operation trend state recognition database, with the condition of the contribution weight constraint of the trend state to the power grid, establish a multiple linear regression model, calculate the sum of squared residuals between the overall power grid operation trend and the weighted output characteristics of the power grid operation trend in each state, iteratively solve the optimal contribution weight to minimize the sum of squared residuals, obtain the contribution weight matrix of the power grid operation trend state, and combine the dynamic change characteristics of the power grid operation trend to update the contribution weight matrix in real time. Finally, establish a power grid operation trend prediction model with the power grid operation trend state characteristics as the input and the overall power grid operation trend as the output.

[0020] Step E. According to the power grid operation trend prediction results, calculate the error sequence of the power grid operation trend prediction, define the prediction error trend factor. If the prediction error shows an increasing trend, return to Step B to update the power grid operation trend evolution model; if the prediction error shows a decreasing or normal fluctuation trend, keep the current power grid operation trend evolution model unchanged at the next prediction moment. Further, based on the error sequence, analyze the response characteristics of the power grid to small disturbances in different operation trend states, propose a sensitivity evaluation index for the power grid operation trend trajectory, and combine the influence of internal and external disturbances of the system to establish a power grid vulnerability evaluation index, use the entropy weight method to weight each trajectory, and construct a power grid vulnerability evaluation model to finally complete the quantitative analysis of the power grid vulnerability.

[0021] As Figure 2As shown, in a preferred embodiment, in step A, the power grid operation trend set has the following mathematical expression: ; In the formula, refers to the power grid operation parameter trajectory cluster, refers to the n th node at time t of the set expression of the power grid operation parameter trajectory cluster, where refers to the q th power grid operation parameter at the n th node at t time T is the power grid operation trend analysis period; refers to the power grid operation trajectory cluster feature set, refers to a power grid operation trend analysis period T within the n th node's power grid operation parameter feature set, where refers to a power grid operation trend analysis period T within the n th node's q th power grid operation parameter's j th feature. is the power grid operation trend set.

[0022] In a preferred embodiment, in step B, the Agent-Cellular Automata (Agent-CA) model combines the statistical laws of power grid operation trends at the macroscopic level with the interaction behaviors among various Agents at the microscopic level, discretizes the power grid spatial region into a two-dimensional lattice grid, and forms the following spatial pattern: ; where represents the cell in the ii-th row and jj-th column, and each cell represents a geographical area in the power grid. The model uses a discrete time step for dynamic evolution.

[0023] Each cell contains one or more agents (Agents), defined as: ; where represents the th Agent in cell k and represents key power grid components such as loads, generators, and buses in this area.

[0024] Each Agent has a state variable at time t ​ , characteristic quantities for representing the operation trend of the power grid (such as frequency deviation, voltage level, or load change, etc.). The state transition rule is: ; Where: represents external influencing factors (such as fluctuations in renewable energy output, load disturbances, etc.); is a state transition function constructed based on the physical characteristics of the power grid and the Agent behavior mechanism; is the state set of the cell and its neighborhood, defined as the Moore neighborhood: ; The time evolution of the power grid operation trend depends on the interaction of the surrounding neighborhood and external disturbances. The specific model expression is: ; Where: α and β are the weight coefficients of neighborhood interaction and external disturbance respectively. is the spatial influence weight of the cell within the neighborhood ( m , n ) on the central cell , reflecting the characteristics of the power grid topology structure.

[0025] At any moment t , the spatial distribution of the power grid operation trend is constrained by the spatial autocorrelation characteristics. The following spatial variogram is defined: ; Where: is the normalized spatial weight, reflecting the influence intensity between different neighborhood cells; The smaller the value, the stronger the spatial aggregation of the trend, and the spatial distribution shows obvious local consistency.

[0026] To distinguish the power grid evolution characteristics under normal and fault conditions, the model defines two types of state transition functions respectively: the normal condition evolution function and the fault condition evolution function

[0027] The difference between the two is expressed as: ; This difference can reflect the mutation characteristics of the power grid operation trend under fault conditions (such as a large drop in frequency, voltage collapse, etc.).

[0028] The model is iteratively optimized through historical operation data to minimize the deviation between the prediction and the observed data: ; Among them, is historical observation data, is the set of parameters to be optimized. Optimization methods can adopt swarm intelligence optimization means such as genetic algorithms and particle swarm algorithms.

[0029] After the optimization is completed, extract the characteristic parameter set under fault conditions , and based on this, form the evolution mechanism and mathematical criterion of the power grid under different fault types: ; Among them, and are the trend mutation threshold and the spatial instability threshold respectively, which are used to accurately identify power grid faults.

[0030] In a preferred embodiment, in step C, define the trend-state space discrimination index as follows: ; Among them: : The trend characteristics of the i, j area at time t; : The time mean of the trend of the entire regional power grid; : The i , j spatial dispersion of the regional trend; : The abnormal deviation in the trend evolution of the i, j area; : Is the weight factor of each characteristic index.

[0031] Based on the set of discrimination indices of all cells, sort according to the discrimination index from large to small: ; Among them is the k th largest discrimination index.

[0032] According to the sorting results, divide different power grid operation trend state clusters , among which L is the number of initially divided state categories. To optimize the division effect of the trend state, introduce the residual transformation rate η as the optimization goal. Define as follows: ; Among them: is the mean value of the discrimination index within the l th trend state cluster; is the mean value of the global discrimination index; The smaller it is, the better the discrimination of the current state division, and the more consistent the trends within the cluster. The goal is to find aη The smallest trend state partitioning scheme, denoted as: ; Among them, the partitioning process adopts a dynamic partitioning method, and combines hierarchical clustering with trend-state space characteristics for iterative partitioning, satisfying ( is the convergence threshold of the residual transformation rate).

[0033] After completing the optimal trend state partitioning, a trend state database is obtained: ; Among them, is a feature data set containing various power grid operation trend states, and is used for subsequent applications such as fault diagnosis, trend prediction, and situation awareness.

[0034] In a preferred embodiment, in step D, based on the established power grid operation trend state recognition database , assuming that the overall power grid operation trend Y ( t ) is obtained by weighted combination of multiple trend state features, the following weighted model is established: ; Among them: is the weighted output feature of the l-th trend state cluster at time t; is the contribution weight of the l -th state cluster to the overall trend; is the fitting residual; L is the number of partitioned trend states.

[0035] Construct a prediction model with trend state features as input and overall operation trend as output: ; Among them: is the predicted trend at time t ; is extracted in real time through the state recognition database ; the weight is updated in real time to ensure the adaptive ability of the model to the dynamic characteristics of the power grid operation trend.

[0036] In a preferred embodiment, in step E, the power grid vulnerability evaluation model is defined as follows: ; Among them, is the sensitivity of the power grid operation parameter trajectory cluster , is the sensitivity of the characteristics of the power grid operation parameter trajectory cluster , is the actual value of the disturbance parameter trajectory, is the average value of the disturbance parameter trajectory.

[0037] ; where, is the power grid vulnerability index, is the weight assigned by the entropy weight method for this parameter, is the occurrence rate of the instability event faced by this parameter due to disturbances, is the upper bound for the safe operation of the trajectory, is the trajectory mean value during the stable operation of the power grid.

Claims

1. A power flow prediction method for a new energy power system considering the uncertainties of power sources, power grids, and loads, characterized in that It includes the following steps: Preprocess the historical parameters of the power grid, extract multi-dimensional trajectory features through variational mode decomposition, and construct a set of power grid operation trends; Combine the cellular automaton model to simulate the spatio-temporal evolution law, iteratively optimize the power grid operation trend model, and obtain the evolution mechanism and mathematical criterion of the power grid operation trend under different fault states; Construct a state recognition database for the power grid operation trend based on the residual transformation rate; Use a multi-constraint regression model to dynamically allocate state contribution weights and establish a power grid operation trend prediction model; Iteratively optimize the model through error feedback, and construct a vulnerability evaluation system based on sensitivity and interference analysis to complete the quantitative analysis of the power grid vulnerability.

2. The method according to claim 1, wherein The preprocessing of the historical parameters of the power grid, extracting multi-dimensional trajectory features through variational mode decomposition, and constructing a set of power grid operation trends specifically include: Normalize the trajectory of the historical operation parameters of the power grid, combine the uncertainty analysis model of the power grid operation parameter trajectory, analyze the specificity between the electrical quantities of the power grid operation and the external parameters affecting the power grid operation state under different power grid operation states, classify the power grid parameters using mathematical statistics models and clustering algorithms, and determine the power grid operation parameter trajectory clusters composed of the electrical quantities of the power grid operation and external parameters under different categories; based on the variational mode decomposition algorithm, extract the frequency domain features of the power grid operation parameter trajectory at different central frequencies to form a set of power grid operation trajectory feature information, define the power grid operation parameter trajectory clusters and their multi-dimensional feature sets as the power grid operation trends, and fuse data from different sources, different time scales, and different qualities to form a set of power grid operation trends.

3. The method according to claim 2, wherein The set of power grid operation trends has the following mathematical expression: ； In the formula, refers to the set of power grid operation parameter trajectories, refers to the n -th node's set expression of the power grid operation parameter trajectory cluster at time t , where refers to the value of the q -th power grid operation parameter at the n -th node at time t ; T is the power grid operation trend analysis period; , refers to the set of power grid operation trajectory cluster characteristics, refers to a power grid operation trend analysis period T within the n -th node's set of power grid operation parameter characteristics, where refers to the T -th node's n -th power grid operation parameter's q -th characteristic within a power grid operation trend analysis period j ; is the power grid operation trend set.

4. The method according to claim 1, wherein The combination of the cellular automaton model to simulate the spatio-temporal evolution law, iteratively optimize the power grid operation trend model, and obtain the evolution mechanism and mathematical criterion of the power grid operation trend under different fault states specifically include: According to the calculated set of power grid operation trends, combined with the basic spatio-temporal evolution law of the power grid operation trend, use the Agent-cellular automaton model, with the spatial distribution and time evolution law of the power grid operation trend as constraints, simulate the dynamic spatio-temporal distribution process of the power grid operation trend, depict the law of the power grid operation trend changing with time within the same spatial region, analyze the time-domain change characteristics of the power grid operation trend, and depict the spatial distribution law of the power grid operation trend in different spatial regions at the same moment, analyze the spatial correlation characteristics of the power grid operation trend, combine the differences in the evolution models under normal and fault conditions, iteratively optimize the power grid operation trend model, accurately identify the set of characteristic parameters of the power grid operation trend under fault states, and finally obtain the evolution mechanism and mathematical criterion of the power grid operation trend under different fault states.

5. The method according to claim 4, wherein The Agent-cellular automaton model combines the statistical laws of the power grid operation trend at the macroscopic level with the interaction behaviors between various Agents at the microscopic level, discretizes the power grid spatial region into a two-dimensional lattice grid, and forms the following spatial pattern: ； Among them, represents the cell in the \(i\)-th row and \(j\)-th column. Each cell represents a geographical area in the power grid, and the model uses discrete time steps for dynamic evolution; Each cell contains one or more agents, defined as: ； Among them, represents the th k Agent within the cell, representing key power grid components such as loads, generators, and busbars in this area; Each Agent at time t has a state variable , which is a characteristic quantity representing the operation trend of the power grid. The state transition rule is as follows: ； Wherein: represents external influencing factors; is a state transition function constructed based on the physical characteristics of the power grid and the Agent behavior mechanism; is the state set of the cell and its neighborhood, defined as the Moore neighborhood: ； The time evolution of the power grid operation trend depends on the interaction of the surrounding neighborhood and external disturbances. The specific model expression is: ； Wherein: α and β are the weight coefficients of neighborhood interaction and external perturbation, respectively; is the spatial influence weight of the cell pair within the neighborhood ( m , n ) on the central cell ; At any moment t , the spatial distribution of the power grid operation trend is constrained by the spatial autocorrelation characteristics, and the following spatial variation function is defined : ； Wherein: is the normalized spatial weight, reflecting the influence intensity between different neighborhood units; The model defines two types of state transition functions respectively: the normal operating condition evolution function and the fault condition evolution function ; The difference between the two is expressed as: ； Iteratively optimize the model through historical operation data to minimize the deviation between the prediction and the observed data: ； Among them, is historical observation data, is the set of parameters to be optimized; After the optimization is completed, extract the characteristic parameter set under fault conditions , and based on this, form the evolution mechanism and mathematical criterion of the power grid under different fault types: ； Among them, and are the trend mutation threshold and the spatial instability threshold respectively, which are used to accurately identify power grid faults.

6. The method according to claim 1, wherein The state recognition database for constructing the power grid operation trend based on the residual transformation rate specifically includes: according to the mathematical criterion of the power grid operation trend, combined with the set of characteristic parameters of the power grid operation trend, defining the trend-state space discrimination index, comprehensively considering the state evolution characteristics of the power grid in the cluster, the power grid operation trend characteristics in different spatial regions, and the trend-state space discrimination index, constructing a power grid operation trend state classification model; arranging in descending order according to the discrimination index, taking the minimum residual transformation rate divided by the power grid operation trend state as the optimization goal, determining the set of power grid operation trend states that minimize the residual transformation rate, and finally forming the state recognition database of the power grid operation trend.

7. The method according to claim 6, characterized in that, The defined trend-state space discrimination index is as follows: ； Wherein: : The trend feature of the i,j region at time t; : The time average of the trend of the entire regional power grid; : The i , j Spatial dispersion of the regional trend; : The abnormal deviation in the trend evolution of the i,j region; : The weight factor of each characteristic index; Based on the set of discrimination indices of all cells sort them in descending order according to the discrimination indices: ； Among them is the k largest discrimination index; Divide different clusters of power grid operation trend states according to the sorting results , where L is the number of initially divided state categories; Introduce the residual transformation rate As the optimization objective, it is defined as follows: ； Wherein: is the mean of the discrimination indices within the l th trend state cluster; is the mean of the global discrimination indices; the goal is to find a trend state partitioning scheme that minimizes η , denoted as: ； Among them, the partitioning process uses a dynamic partitioning method, combines hierarchical clustering with trend-state space features for iterative partitioning, and satisfies ; After completing the optimal trend state division, the trend state database is obtained: ； Among them, is a characteristic data set containing various power grid operation trend states.

8. The method according to claim 1, wherein The method of dynamically allocating state contribution weights using a multi-constraint regression model to establish a power grid operation trend prediction model specifically includes: according to the power grid operation trend prediction results, calculating the error sequence of the power grid operation trend prediction, defining the prediction error trend factor, if the prediction error shows an increasing trend, then updating the power grid operation trend evolution model; if the prediction error shows a decreasing or normal fluctuation trend, then maintaining the current power grid operation trend evolution model unchanged at the next prediction moment; further, based on the error sequence, analyzing the response characteristics of the power grid to small disturbances under different operation trend states, proposing a sensitivity evaluation index for the power grid operation trend trajectory, and combining the influence of internal and external disturbances of the system, establishing a power grid vulnerability evaluation index, using the entropy weight method to weight each trajectory, constructing a power grid vulnerability evaluation model, and finally completing the quantitative analysis of the power grid vulnerability.

9. The method according to claim 8, characterized in that The power grid vulnerability evaluation model is defined as follows: ； Among them, is the sensitivity of the power grid operation parameter trajectory cluster , is the sensitivity of the characteristics of the power grid operation parameter trajectory cluster , is the actual value of the disturbance parameter trajectory is the average value of the disturbance parameter trajectory; ； Among them, is the power grid vulnerability index, is the weight assigned by the entropy weight method of this parameter, is the occurrence rate of the instability event faced by this parameter, is the upper bound of the safe operation of the trajectory, is the trajectory mean value when the power grid operates stably.

Images