A Risk Prevention and Control Method and System for the Electricity Spot Market with Multi-Energy Access

Abstract

The present invention discloses a risk prevention and control method and system for a multi - energy access power spot market, which relates to the technical field of risk prevention and control, and includes the following steps: obtaining multi - dimensional data of a region to be measured, obtaining the risk conduction intensity between nodes based on the power grid topological structure, and constructing a dynamic risk topological network model; extracting the node operation characteristics according to the risk topological network model, synchronously obtaining the evaluation data of the classification model, and performing model switching regulation according to the evaluation data, where the model switching regulation further includes outputting the risk spillover probability; the present application solves the problems of inaccurate risk prediction and untimely response in the multi - energy access power spot market through a method based on a dynamic risk topological network and multi - model adaptive switching.

Description

Technical Field

[0001] The present invention relates to the technical field of risk prevention and control, and more specifically, to a risk prevention and control method and system for a power spot market with multi-energy access. Background Art

[0002] With the transformation of the global energy structure and the progress of technology, the power system is developing towards diversification and intelligence. The traditional power system mainly relies on non-renewable energy sources, including coal, oil, and natural gas. However, in recent years, the rapid rise of renewable energy sources such as solar energy and wind energy has brought new vitality and challenges to the power system. Especially when these renewable energy sources are connected to the power grid in the form of distributed generation, the complexity and uncertainty of the power system increase significantly.

[0003] The power system with multi-energy access has become an inevitable trend for future development. The so-called multi-energy access means that the power system can simultaneously accept different types of energy sources, including renewable energy sources, energy storage devices, traditional fossil energy sources, etc., and achieve the complementarity and optimal allocation of these energy sources. This power system with multi-energy access can not only improve the energy utilization efficiency but also enhance the flexibility and reliability of the power system. However, multi-energy access also brings the problem of risk prevention and control in the power spot market.

[0004] For example, a method and device for evaluating the attack risk of a load frequency control system disclosed in the invention patent with the publication number of CN114967647A belong to the field of power cyber-physical systems. Among them, the method includes: establishing a discretized simulation model of a load frequency control system including multiple regions; starting from any initial moment, continuously superimposing private incentives that meet the set distribution on the control instruction input of the load frequency control system; under the set detection window, by obtaining the detection index of the load frequency control system after superimposing the private incentives, updating the confidence level of the sensors corresponding to each region in the system; according to the confidence level sorting result, obtaining the risk assessment result of the load frequency control system. The present invention can be applied to the abnormal warning and risk assessment of sensor data in a multi-region load frequency control system, can quickly locate abnormal sensor data without adding costs, realize attack identification and location, and improve the safety and reliability of the load frequency control system.

[0005] For example, the inference method and device of the performance-fault relationship graph based on the graph neural network disclosed in the invention patent with the announcement number of CN116360388B. The method includes: constructing a performance-fault relationship graph of the spacecraft control system according to FMEA; calculating the intimacy of each relationship in the performance-fault relationship graph based on historical fault cases; based on the intimacy of each relationship in the performance-fault relationship graph, using the graph neural network to infer the fault cause corresponding to each fault symptom; for each fault symptom, using the Bayesian network to fuse the fault cause corresponding to the current fault symptom to obtain the final inference result of the current fault symptom. The present invention can improve the accuracy of fault inference in the spacecraft control system.

[0006] In the above disclosed technical solution, there are at least the following technical problems:

[0007] Traditional methods focus on the monitoring of the overall risk level and lack an intuitive display of the risk propagation path, resulting in inaccurate risk prediction and untimely response. In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0008] In order to overcome the above defects of the prior art, an embodiment of the present invention provides a risk prevention and control method and system for a multi-energy access power spot market. By means of a method based on a dynamic risk topology network and multi-model adaptive switching, the problems of inaccurate risk prediction and untimely response in the multi-energy access power spot market are solved, and an innovative breakthrough in real-time dynamic regulation and multi-level risk visualization is achieved.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A risk prevention and control method and system for a multi-energy access power spot market, including the following steps: obtaining multi-dimensional data of the area to be measured, obtaining the risk conduction intensity between nodes based on the power grid topology structure, and constructing a dynamic risk topology network model; extracting the node operation characteristics according to the risk topology network model, synchronously obtaining the evaluation data of the classification model, and performing model switching regulation according to the evaluation data. The model switching regulation also includes outputting the risk spillover probability.

[0011] In a preferred embodiment, the obtaining of the multi-dimensional data of the area to be measured, obtaining the risk conduction intensity between nodes based on the power grid topology structure, and constructing a dynamic risk topology network model are specifically as follows: regarding the nodes of the wind farm, photovoltaic power station, load center, and energy storage station as network nodes, and outputting the risk conduction intensity between nodes through Granger causality analysis and impulse response function; constructing a dynamic risk topology network according to the risk conduction intensity between nodes.

[0012] In a preferred embodiment, constructing a dynamic risk topology network according to the risk conduction intensity between nodes is as follows: Using the power grid topology diagram as the basis and the risk conduction intensity as the weight of the edge, construct a risk-weighted network; Obtain the first data of the risk-weighted network, analyze the key nodes and key lines according to the first data, and identify high-risk conduction paths; Simulate the risk propagation process in the power grid through a dynamic propagation model, analyze the risk diffusion behavior under different control strategies, and construct a dynamic risk topology network.

[0013] In a preferred embodiment, analyzing the key nodes and key lines according to the first data and identifying high-risk conduction paths is as follows: According to the first data, output the second data, and screen the node risk propagation paths according to the risk conduction intensity and the second data to obtain the key nodes and key paths; Construct a risk propagation chain according to the key nodes and key paths, and identify high-risk conduction paths according to the risk propagation chain.

[0014] In a preferred embodiment, extracting the node operation characteristics according to the risk topology network model, synchronously obtaining the evaluation data of the classification model, and performing model switching regulation according to the evaluation data, and the model switching regulation also includes outputting the risk spillover probability, which is as follows: Extract the node operation characteristics, and the node operation characteristics include the first data set, the second data set, and the third data set; Perform reliability evaluation on the classification model through a sliding time window to obtain the reliability evaluation data of the classification model, and the reliability evaluation data includes prediction error and confidence level; According to the reliability evaluation data of the classification model, switch the classification model, and the corresponding node operation characteristics of the data, and output the risk spillover probability. The classification models include LSTM, Bayesian network, and Markov chain; Generate a risk heat map according to the risk spillover probability for risk prevention and control.

[0015] In a preferred embodiment, switching the classification model according to the reliability evaluation data of the classification model is as follows: Switch the classification model according to the reliability evaluation data of the classification model, specifically as follows: Under the conditions of small prediction error and high confidence level, switch to the LSTM model; Under the conditions of high prediction error and high confidence level, switch to the Bayesian network to quickly locate abnormal nodes and fault causes; Under the conditions of high confidence level and state transition rate higher than the threshold, switch to the Markov chain.

[0016] In a preferred embodiment, the confidence level includes a Markov chain confidence level, and the steps for obtaining the Markov chain confidence level are as follows: Obtain the historical state change data of the node to be measured, perform a fine-grained division of the node state through a clustering algorithm, obtain the number of transitions between different states of the node, and construct an initial state transition matrix; Predict the current state according to the state at the previous moment through the Markov chain model, compare the actual state with the predicted state, and output the state residual; Based on the state transition probability, state residual, and historical prediction accuracy in the initial state transition matrix, output the confidence level at the current moment.

[0017] In a preferred embodiment, the steps for obtaining the confidence penalty factor of the state transition matrix are as follows: Obtain the historical state residual data of the Markov chain model; Select several historical windows of different lengths according to the historical state residual data, obtain the average value of all state residuals within the historical window, and construct a state residual data set; Predict the state residuals through the Markov chain model, and dynamically adjust the confidence penalty factor by selecting the optimal window length according to the state residual data set.

[0018] In a preferred embodiment, generating a risk heat map based on the risk spillover probability for risk prevention and control is as follows: Obtain the scenario data of each node, perform scenario division on the nodes, perform dynamic model switching according to the scenario division, and predict the risk spillover probability; Construct a probability matrix based on the risk spillover probability prediction data; Map the probability matrix to the power grid geographical map, perform spatial interpolation on the discrete node probabilities, and generate a risk heat map; Perform risk prevention and control according to the risk heat map.

[0019] A system for a risk prevention and control method of a multi-energy access power spot market includes a risk topology network module and a risk spillover probability output module, and there is a connection between the modules; The risk topology network module is used to obtain multi-dimensional data of the area to be measured, obtain the risk conduction intensity between nodes based on the power grid topology structure, and construct a dynamic risk topology network model; The risk spillover probability output module is used to extract the node operation characteristics according to the risk topology network model, synchronously obtain the evaluation data of the classification model, and perform model switching regulation according to the evaluation data, and the model switching regulation includes outputting the risk spillover probability.

[0020] The technical effects and advantages of the risk prevention and control method and system for a multi-energy access power spot market of the present invention:

[0021] 1. By constructing a dynamic risk topology network model based on the risk conduction intensity of the power grid topology structure, the present invention can accurately simulate the risk propagation paths between different nodes in the power grid. Through Granger causality analysis and impulse response functions, the risk transfer relationship between nodes can be quantitatively analyzed, providing accurate basic data for subsequent risk prediction.

[0022] 2. By analyzing the risk topology network model, the present invention can extract the operating characteristics of power grid nodes, and then evaluate the reliability of the classification model through a sliding time window to obtain the prediction error and confidence level. The node operating characteristics include indicators such as voltage offset, load rate, and short-term fluctuation amplitude. These characteristics can help judge the stability or abnormality of the current state of the power grid, and based on this, the nodes are divided into different risk scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic flow chart of a risk prevention and control method for a multi-energy access power spot market according to the present invention.

[0024] Figure 2 It is a schematic structural diagram of a system for a risk prevention and control method for a multi-energy access power spot market according to the present invention.

[0025] Figure 3 It is a risk spillover probability curve graph. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0027] Embodiment 1 Figure 1 A risk prevention and control method for a multi-energy access power spot market according to the present invention is given, and the specific steps are as follows:

[0028] S1. Obtain multi-dimensional data of the area to be measured, obtain the risk conduction intensity between nodes based on the power grid topology structure, and construct a dynamic risk topology network model.

[0029] In this embodiment, obtaining multi-dimensional data of the area to be measured, obtaining the risk conduction intensity between nodes based on the power grid topology structure, and constructing a dynamic risk topology network model according to the risk conduction intensity are as follows:

[0030] Obtain data such as power generation output, load curve, voltage frequency, energy storage charge and discharge status, and electricity price through smart meters and energy storage management system devices, repair missing data using the KNN interpolation method, and remove outliers using the 3σ principle;

[0031] Use the Z-score standardization method to uniformly process data with different dimensions, and use a unified time stamp for the processed data to ensure the timeliness and consistency of the data;

[0032] Regarding the nodes of wind farms, photovoltaic power stations, load centers, and energy storage stations as network nodes, the risk conduction intensity between the nodes is output through Granger causality analysis and impulse response functions;

[0033] A risk topology network is constructed based on the risk conduction intensity between the nodes, and the edge weights between the nodes change dynamically to reflect the risk conduction relationship in real time.

[0034] In this embodiment, a dynamic risk topology network is constructed according to the risk conduction intensity between the nodes, specifically as follows:

[0035] Taking the power grid topology diagram as the basis and using the risk conduction intensity as the edge weight, a risk-weighted network is constructed;

[0036] Obtain the first data of the risk-weighted network, where the first data includes the clustering coefficient and average path length of the network;

[0037] Analyze the key nodes and key lines based on the first data to identify high-risk conduction paths;

[0038] Simulate the risk propagation process in the power grid through a dynamic propagation model, analyze the risk diffusion behavior under different control strategies, and construct a dynamic risk topology network.

[0039] The calculation formula of the risk dynamic propagation model is specifically as follows:

[0040]

[0041] In the formula: is the risk state of node i at the next moment t + 1, is the risk value of node i at time t, is the risk value of node j at time t, is the set of adjacent nodes of node i, representing all nodes directly connected to node i, is the risk conduction intensity transmitted from node j to node i, is the total risk received by node i from its neighbor node j.

[0042] It should be noted that is often used to represent the normalized risk value, 0 represents no risk, and 1 represents complete failure, is determined by the current risk state of the node and the risk conduction amount of adjacent nodes to it and depends on the topology structure of the power grid, represents the risk propagation ability from node j to node i, It reflects the process of gradual spread of faults or risks in the power grid through the topological structure. If the risk value of node j is high and is also high, then the increase in the risk of node i will be more significant.

[0043] In this embodiment, key nodes and key lines are analyzed based on the first data to identify high-risk conduction paths, specifically as follows:

[0044] Output the second data according to the first data, and the second data includes betweenness centrality and degree centrality;

[0045] Screen the node risk propagation paths according to the risk conduction intensity and the second data to obtain key nodes and key paths;

[0046] Construct a risk propagation chain based on the key nodes and key paths, and identify high-risk conduction paths according to the risk propagation chain.

[0047] The specific calculation formula of betweenness centrality is as follows:

[0048]

[0049] The specific calculation formula of the clustering coefficient is as follows:

[0050]

[0051] The specific calculation formula of degree centrality is as follows:

[0052]

[0053] In the formula: is the betweenness centrality, is the number of shortest paths passing through node i, is the number of shortest paths from node s to node t, is the degree centrality, is the adjacency matrix, is the degree of node i, is the clustering coefficient, is the actual number of connections between adjacent nodes of node i.

[0054] It should be noted that nodes with high betweenness centrality are usually "hub" nodes in the power grid. They connect multiple subnets. Once risks spread in these nodes, it will exacerbate the risk diffusion of the entire power grid; nodes with high degree centrality are connected to more power grid devices and directly affect the risk propagation of multiple nodes. Once the risks of these nodes increase, it may affect a larger area of the power grid; the clustering coefficient reflects the risk propagation characteristics in a local area. If the clustering coefficient of a certain area is high, the risk may spread rapidly in this local area without affecting the remote area.

[0055] S2. Extract the node operation characteristics according to the risk topology network model, synchronously obtain the evaluation data of the classification model, and perform model switching regulation according to the evaluation data. The model switching regulation also includes outputting the risk spillover probability.

[0056] In this embodiment, extracting the node operation characteristics according to the risk topology network model, synchronously obtaining the evaluation data of the classification model, and performing model switching regulation according to the evaluation data. The model switching regulation also includes outputting the risk spillover probability, which is specifically as follows:

[0057] Extract the node operation characteristics, where the extraction of node operation characteristics includes a first data set, a second data set, and a third data set;

[0058] The first data set includes voltage offset, load rate, and short-term fluctuation amplitude, and is used to input the LSTM (Long Short-Term Memory Network);

[0059] The second data set includes neighborhood aggregation features, risk conduction intensity, and probability distribution of the current node state, and is used to input the Bayesian network;

[0060] The third data set includes state transition rate, state transition probability matrix, and trend change rate, and is used to input the Markov chain;

[0061] Perform reliability evaluation on the classification model through a sliding time window to obtain the reliability evaluation data of the classification model. The reliability evaluation data includes prediction error and confidence;

[0062] According to the reliability evaluation data of the classification model, switch the classification model, and the corresponding node operation characteristics of the data, and output the risk spillover probability. The classification models include LSTM (Long Short-Term Memory Network), Bayesian network, and Markov chain;

[0063] Generate a risk heat map according to the risk spillover probability for risk prevention and control.

[0064] The specific calculation formula of the prediction error is as follows:

[0065]

[0066] The specific calculation formula of the confidence of the Bayesian network is as follows:

[0067]

[0068] In the formula: is the prediction error, is the number of prediction moments within the current window, is the true value at the i-th moment, is the model prediction value at the i-th moment, is the confidence of the Bayesian network, is the system state, is the observed input data, is the posterior probability of the state.

[0069] In this embodiment, the classification model is switched according to the reliability evaluation data of the classification model, specifically as follows:

[0070] Under the conditions of small prediction error and high confidence, switch to the LSTM model;

[0071] Under the conditions of high prediction error and high confidence, switch to the Bayesian network to quickly locate abnormal nodes and fault causes;

[0072] Under the conditions of high confidence and state transition rate higher than the threshold, switch to the Markov chain.

[0073] Furthermore, explain the reasons for selecting the model according to the scenario division:

[0074] Select the LSTM model: LSTM performs excellently in processing long-term sequence data and can effectively capture the temporal characteristics of the node operation state. In a stable scenario, the changes of parameters such as voltage and load usually have strong regularity. LSTM is good at extracting trends and periodic features from historical data. Since the system state is stable and there is less noise, the LSTM model can provide lower prediction error and higher confidence, and can achieve long-term window prediction to identify potential risk spillover tendencies in advance;

[0075] Select the Bayesian network: The Bayesian network represents the causal relationship between nodes through a probabilistic graph model. When an anomaly is detected, it can quickly trace the cause of the anomaly. Using Bayesian inference, the occurrence probability of an abnormal event can be quantified when the data is incomplete or there is noise. It has the ability to quickly respond to sudden abnormal events such as load surges and equipment failures, and can provide the risk spillover probability and the trend of abnormal development. High confidence indicates that the abnormal detection result is reliable, and the cause of the anomaly can be further inferred;

[0076] Select the Markov chain model: The Markov chain model is based on the state transition matrix and can effectively capture the transition law of the node between different states. In a state transition scenario, the operation state of the node may change rapidly. The Markov chain is good at predicting the state evolution in a short time. In the scenario where the node state frequently switches, the computational complexity of the model is small, and the state transition matrix can be quickly updated to achieve real-time prediction. The Markov chain is good at modeling the state transition process and is suitable for analyzing the trend of node state changes. If the confidence is high and the state transition rate is fast, the power grid state may be in a transition stage.

[0077] In this embodiment, the steps for obtaining the LSTM confidence are as follows:

[0078] Obtain the historical time series data of the node to be measured, detect and mark abnormal data, use anomaly detection algorithms based on IQR (Interquartile Range) or LOF (Local Outlier Factor), remove outliers or repair data using interpolation methods, and remove noise through smoothing filtering or wavelet transform to ensure the quality of the input data;

[0079] Capture the temporal features of the historical time series data by combining LSTM and multi-scale convolutional layers, use the multi-step prediction method to predict the node states at different time granularities, further enhance the model's ability to capture temporal dependencies, obtain the residuals between the predicted values and the actual values, and construct a residual sequence;

[0080] Decompose the residuals into trend terms, seasonal terms, and random terms through decomposition techniques (such as STL decomposition), analyze the errors from different sources respectively, and for abnormal residuals, combine expert rules or Bayesian inference to determine whether they are caused by data anomalies, reducing the misjudgment rate of errors;

[0081] Perform statistical analysis on the residual sequence through the sliding window method, and output the mean and standard deviation of the residuals;

[0082] Based on the assumption of normal distribution according to the mean and standard deviation of the residuals, output the confidence of the current prediction;

[0083] Identify abnormal jumps in the confidence based on the changing trend of the historical prediction confidence. If an anomaly occurs, trigger the confidence correction mechanism and re-evaluate the model performance;

[0084] If the residual is close to the mean and within the threshold range, the confidence is high.

[0085] The specific formula for calculating the LSTM confidence is as follows:

[0086]

[0087] In the formula: is the confidence of LSTM, ranging from 0 to 1, is the residual at the current moment, that is, the gap between the true value and the predicted value, is the mean of the residuals within the sliding window, used to measure the normal error level of the system, is the standard deviation of the residuals within the sliding window, reflecting the fluctuation range of the errors.

[0088] It should be noted that 0 and 1 in the formula for calculating the LSTM confidence are the lower and upper bounds of the confidence respectively. When the prediction error is greater than When it indicates that the error of the current prediction has seriously deviated from the historical error distribution, the confidence level is determined to be 0; when the prediction error is close to 0, that is, when the current predicted value is very close to the actual value, the confidence level is determined to be 1; if the value is less than 0, it means that the prediction error is too large and the confidence level is 0.

[0089] In this embodiment, the steps for obtaining the Markov chain confidence level are specifically as follows:

[0090] Obtain the historical state change data of the node to be measured, perform a fine-grained division of the node state through a clustering algorithm, divide the continuous state into a finer discrete state space, count the number of transitions between different states of the node, and construct an initial state transition matrix;

[0091] Predict the current state according to the state at the previous moment through the Markov chain model, compare the actual state and the predicted state, output the state residual, and use an anomaly detection algorithm (such as Kalman filtering or Bayesian anomaly detection) to identify abnormal states to prevent abnormal states from misleading the confidence level calculation;

[0092] Based on the state transition probability, state residual, and historical prediction accuracy in the initial state transition matrix, output the confidence level at the current moment;

[0093] If the state prediction is correct and the transition probability is high, the confidence level is high;

[0094] If the state residual is large or the prediction is inaccurate, the confidence level is reduced.

[0095] The specific formula for the Markov chain confidence level is as follows:

[0096]

[0097]

[0098] In the formula: is the Markov chain confidence level, is the state transition probability, that is, the probability of transitioning from the state at the previous moment to the state at the current moment, is the state residual, which is 1 if the state prediction is incorrect and 0 otherwise, is the confidence level penalty factor of the state transition matrix, is the state residual at the i-th moment, is the length of the historical window.

[0099] In this embodiment, the steps for obtaining the confidence level penalty factor of the state transition matrix are specifically as follows:

[0100] Obtain the historical state residual data of the Markov chain model;

[0101] Select several historical windows with different lengths according to the historical state residual data, obtain the average value of all state residuals within the historical window, which represents the accuracy of historical state prediction, and construct a state residual data set;

[0102] If most of the predictions within the window are correct, the average value is close to 0, indicating good prediction performance;

[0103] If most of the predictions within the window are incorrect, the average value is close to 1, indicating poor prediction performance;

[0104] Predict the state residuals through a Markov chain model, and dynamically adjust the confidence penalty factor according to the state residual data set by selecting the optimal window length.

[0105] S3. Generate a risk heat map according to the risk spillover probability for risk prevention and control.

[0106] In this embodiment, generating a risk heat map according to the risk spillover probability for risk prevention and control is as follows:

[0107] Obtain the scenario data of each node, divide the scenarios of the nodes, perform dynamic model switching according to the scenario division, and predict the risk spillover probability;

[0108] Construct a probability matrix according to the risk spillover probability prediction data;

[0109] Map the probability matrix to the power grid geographical map, and use the kernel density estimation (KDE) algorithm to perform spatial interpolation on the discrete node probabilities to generate a risk heat map;

[0110] Perform risk prevention and control according to the risk heat map.

[0111] Embodiment 2 Figure 2 A system for a risk prevention and control method of a multi - energy access power spot market according to the present invention is given, including a risk topology network module and a risk spillover probability output module, and there is a connection between the modules;

[0112] The risk topology network module is used to obtain multi - dimensional data of the area to be measured, obtain the risk conduction intensity between nodes based on the power grid topology structure, and construct a dynamic risk topology network model;

[0113] The risk spillover probability output module is used to extract the node operation characteristics according to the risk topology network model, synchronously obtain the evaluation data of the classification model, and perform model switching regulation according to the evaluation data. The model switching regulation includes outputting the risk spillover probability.

[0114] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0115] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0116] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0117] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0118] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0119] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A risk prevention and control method for the electricity spot market with multi - energy access, characterized in that, It includes the following steps: Obtain multi-dimensional data of the area to be measured, obtain the risk conduction intensity between nodes based on the power grid topology structure, and construct a dynamic risk topology network model, specifically as follows: Take the power grid topology diagram as the basis, and use the risk conduction intensity as the edge weight to construct a risk-weighted network; Obtain the first data of the risk-weighted network, and obtain the second data according to the first data; According to the risk conduction intensity and the second data, screen the node risk propagation paths to obtain key nodes and key paths; Construct a risk propagation chain according to the key nodes and key paths, and identify high-risk conduction paths according to the risk propagation chain; Simulate the risk propagation process in the power grid through a dynamic propagation model, analyze the risk diffusion behavior under different control strategies, and construct a dynamic risk topology network; Extract the node operation characteristics according to the risk topology network model, synchronously obtain the evaluation data of the classification model, and perform model switching control according to the evaluation data. The model switching control also includes outputting the risk overflow probability; Switch the classification model according to the reliability evaluation data of the classification model, specifically as follows: Under the conditions of small prediction error and high confidence, switch to the LSTM model; Under the conditions of high prediction error and high confidence, switch to the Bayesian network to quickly locate abnormal nodes and fault causes; Under the conditions of high confidence and state transition rate higher than the threshold, switch to the Markov chain.

2. The risk prevention and control method for the electricity spot market with multi - energy access according to claim 1, wherein, The obtaining of multi-dimensional data of the area to be measured, obtaining the risk conduction intensity between nodes based on the power grid topology structure, and constructing a dynamic risk topology network model, specifically as follows: Regard the nodes of wind farms, photovoltaic power stations, load centers, and energy storage stations as network nodes, and output the risk conduction intensity between nodes through Granger causality analysis and impulse response function; Construct a dynamic risk topology network according to the risk conduction intensity between nodes.

3. The risk prevention and control method for the electricity spot market with multi - energy access according to claim 2, characterized in that, The extracting of node operation characteristics according to the risk topology network model, synchronously obtaining the evaluation data of the classification model, and performing model switching control according to the evaluation data. The model switching control also includes outputting the risk overflow probability, specifically as follows: Extract node operation characteristics, and the node operation characteristics include the first data set, the second data set, and the third data set; Perform reliability evaluation on the classification model through a sliding time window to obtain the reliability evaluation data of the classification model. The reliability evaluation data includes prediction error and confidence; Switch the classification model according to the reliability evaluation data of the classification model, and the corresponding node operation characteristics of the data, and output the risk overflow probability. The classification models include LSTM, Bayesian network, and Markov chain; Generate a risk heat map according to the risk overflow probability for risk prevention and control.

4. The risk prevention and control method for the electricity spot market with multi - energy access according to claim 3, characterized in that, The confidence includes the Markov chain confidence. The obtaining steps of the Markov chain confidence are specifically as follows: Obtain the historical state change data of the node to be measured, perform fine-grained division of the node state through a clustering algorithm, obtain the transfer times between different states of the node, and construct an initial state transition matrix; Predict the current state according to the previous state through the Markov chain model, compare the actual state and the predicted state, and output the state residual; Output the confidence level at the current moment based on the state transition probability, state residual, and historical prediction accuracy in the initial state transition matrix.

5. The risk prevention and control method for the electricity spot market with multi - energy access according to claim 4, characterized in that, The steps for obtaining the confidence penalty factor of the state transition matrix are specifically as follows: Obtain the historical state residual data of the Markov chain model; Select several historical windows of different lengths according to the historical state residual data, obtain the average value of all state residuals within the historical window, and construct a state residual data set; Predict the state residuals through the Markov chain model, and dynamically adjust the confidence penalty factor by selecting the optimal window length according to the state residual data set.

6. The risk prevention and control method for the electricity spot market with multi - energy access according to claim 5, characterized in that, Generating a risk heat map based on the risk spillover probability for risk prevention and control is specifically as follows: Obtain the scenario data of each node, divide the scenarios of the nodes, perform dynamic model switching according to the scenario division, and predict the risk spillover probability; Construct a probability matrix based on the risk spillover probability prediction data; Map the probability matrix to the power grid geographical map, perform spatial interpolation on the discrete node probabilities, and generate a risk heat map; Perform risk prevention and control according to the risk heat map.

7. A system for a risk prevention and control method of a power spot market using multi-energy access as described in any one of claims 1-6, characterized in that It includes a risk topology network module and a risk spillover probability output module, and there is a connection between the modules; The risk topology network module is used to obtain the multi-dimensional data of the area to be measured, obtain the risk conduction intensity between nodes based on the power grid topology structure, and construct a dynamic risk topology network model; The risk spillover probability output module is used to extract the node operation characteristics according to the risk topology network model, synchronously obtain the evaluation data of the classification model, and perform model switching regulation according to the evaluation data. The model switching regulation includes outputting the risk spillover probability.

Images