A method, system, equipment, and medium for risk assessment and early warning of power system operation status under extreme weather conditions.

Abstract

This application discloses a method for risk assessment and early warning of power system operation status under extreme weather conditions, relating to the field of power system safe operation and risk early warning technology. The method includes inputting global meteorological data into a load forecasting architecture and outputting load forecasting results; constructing a power system fault probability model based on the load forecasting results; determining the initial fault operation scenario results of the power system under extreme weather conditions based on the power system fault probability model; constructing an operation assessment scheme based on the N-k cascading fault propagation path and the initial fault operation scenario results of the power system under extreme weather conditions; determining the high-risk operation scenario under the current extreme weather conditions based on the operation assessment scheme, and issuing an early warning based on the high-risk operation scenario under extreme weather conditions. This improves the ability to quantitatively assess and warn of power system operation status risks, and enhances the accuracy of power system operation status risk assessment and early warning under extreme weather conditions.

Description

Technical Field

[0001] This application relates to the field of power system safe operation and risk early warning technology, and in particular to a method, system, equipment and medium for risk assessment and early warning of power system operation status under extreme weather conditions. Background Technology

[0002] In modern society, the stability of the power system is crucial for ensuring social operation and economic development. Extreme weather events, such as typhoons, torrential rains, and extreme heat, have an increasingly significant impact on the power system, leading to damage to power equipment, surges in load demand, or reductions in power supply capacity, triggering cascading failures and seriously threatening the safe operation of the power system. Therefore, accurately assessing the impact of extreme weather on the system and providing timely warnings and responses have become urgent problems that the power system must address.

[0003] Traditional power system risk assessment methods often rely on historical data and experience-based judgments, which are difficult to adapt to the complex and ever-changing operating environment brought about by extreme weather.

[0004] To address these challenges, power system risk assessment and early warning technologies need further improvement and optimization. Existing research and practice show that data-driven methods combined with advanced machine learning algorithms can provide more accurate and real-time predictions of power system operating conditions, thereby providing decision support for the safe operation of power systems. However, these methods are not accurate enough in handling power system operating condition risk assessment and early warning under extreme weather conditions. They often suffer from insufficient consideration of the impact of extreme weather, limited ability to identify cascading fault propagation paths, and insufficient quantitative assessment of power system operating condition risks, resulting in inaccuracies in handling power system operating condition risk assessment and early warning under extreme weather conditions. Summary of the Invention

[0005] This application provides a method, system, equipment, and medium for risk assessment and early warning of the operating status of a power system under extreme weather conditions, in order to solve the problem of low accuracy in risk assessment and early warning of the operating status of a power system caused by the impact of extreme weather on the operating status of the power system.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] Firstly, this application provides a method for risk assessment and early warning of power system operation status under extreme weather conditions, including:

[0008] Global meteorological data is input into a load forecasting architecture, which outputs load forecasting results. This architecture is built upon a graph neural network and includes a feature extraction layer, a feature transformation layer, and a prediction output layer connected sequentially. The feature extraction layer extracts first high-dimensional meteorological features from the global meteorological data within the power supply area and mines the short-term and long-term time-series features implicit in the load data of the global meteorological data to obtain second and third high-dimensional meteorological features. The feature transformation layer performs linear transformations on the first, second, and third high-dimensional meteorological features to obtain fourth, fifth, and sixth high-dimensional meteorological features. The prediction output layer weightedly fuses the fifth and sixth high-dimensional meteorological features to output load forecasting results under extreme weather conditions.

[0009] A power system fault probability model is constructed based on the load forecast results.

[0010] Based on the power system fault probability model, the initial fault operation scenario of the power system under extreme weather conditions is determined.

[0011] Based on the Nk cascading fault propagation path, an operation evaluation scheme is constructed according to the results of the initial fault operation scenario of the power system under extreme weather conditions; the operation evaluation scheme includes an operation evaluation scheme for the cascading fault propagation stage and an operation evaluation scheme for the optimized scheduling stage.

[0012] The operational assessment scheme determines the high-risk operational scenarios under current extreme weather conditions and issues early warnings based on these high-risk scenarios.

[0013] Secondly, this application provides a power system operation status risk assessment and early warning system under extreme weather conditions, including:

[0014] The load forecasting module is used to input global meteorological data into the load forecasting architecture and output load forecasting results.

[0015] The power system initial fault operation scenario result output module is used to determine the power system initial fault operation scenario result under extreme weather conditions based on the power system fault probability model.

[0016] The high-risk early warning module is used to determine the high-risk operation scenario under the current extreme weather conditions based on the operation assessment scheme, and to issue an early warning based on the high-risk operation scenario under the extreme weather conditions.

[0017] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of risk assessment and early warning of the operating status of the power system under extreme weather conditions as described above.

[0018] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of risk assessment and early warning of the operating status of the power system under extreme weather conditions as described above.

[0019] According to the specific embodiments provided in this application, this application has the following technical effects:

[0020] This application provides a method for risk assessment and early warning of power system operation status under extreme weather conditions. Global meteorological data is input into a load forecasting architecture, which outputs load forecasting results. The load forecasting architecture is constructed based on a graph neural network. The architecture includes a feature extraction layer, a feature transformation layer, and a prediction output layer connected sequentially. The feature extraction layer extracts first high-dimensional meteorological features from the global meteorological data within the power supply area and mines the short-term and long-term time-series features implicit in the load data of the global meteorological data to obtain second and third high-dimensional meteorological features. The feature transformation layer performs linear transformations on the first, second, and third high-dimensional meteorological features to obtain fourth, fifth, and sixth high-dimensional meteorological features. The prediction output layer converts the fifth high-dimensional meteorological feature into a second high-dimensional feature. The system performs weighted fusion of key features and the sixth meteorological feature to output load forecast results under extreme weather conditions. Based on the power system fault probability model, it determines the initial fault operation scenario results of the power system under extreme weather conditions. Based on the Nk cascading fault propagation path, it constructs an operation assessment scheme according to the initial fault operation scenario results of the power system under extreme weather conditions. The operation assessment scheme includes operation assessment schemes for the cascading fault propagation stage and operation assessment schemes for the optimized scheduling stage. Based on the operation assessment scheme, it determines the high-risk operation scenario under the current extreme weather conditions and issues early warnings based on the high-risk operation scenario under extreme weather conditions. By fully considering the impact of extreme weather and improving the ability to identify cascading fault propagation paths, the system enhances the ability to quantify and warn of power system operation status risks, thereby improving the accuracy of power system operation status risk assessment and early warning under extreme weather conditions. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram of the process for risk assessment and early warning of power system operation status under extreme weather conditions provided in this application embodiment;

[0023] Figure 2 A schematic diagram of the IEEE 39-node line topology provided for embodiments of this application;

[0024] Figure 3 A schematic diagram of the probability density distribution function of line failure rate as a function of temperature, provided for an embodiment of this application;

[0025] Figure 4 This is a schematic diagram illustrating the prediction results of load data under normal weather conditions provided in an embodiment of this application.

[0026] Figure 5 This is a schematic diagram illustrating the prediction results of load nodes under high-temperature weather conditions provided in an embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] In modern society, the stability of power systems is crucial for ensuring social operation and economic development. Extreme weather events, such as typhoons, torrential rains, and extreme heat, have an increasingly significant impact on power systems, leading to damage to power equipment, surges in load demand, or reductions in power supply capacity, triggering cascading failures and seriously threatening the safe operation of power systems. With the intensification of global climate change, the frequency and intensity of extreme weather events are increasing, further escalating the risks faced by power systems. Therefore, accurately assessing the impact of extreme weather on power systems and providing timely warnings and responses have become urgent problems to be solved in the field of power systems.

[0030] Traditional power system risk assessment methods often rely on historical data and experience-based judgments, making them ill-suited to the complex and volatile operating environments brought about by extreme weather. Furthermore, with the rapid development of new energy generation technologies, the proportion of intermittent energy sources such as wind and solar power in the power system is continuously increasing. The output of these energy sources is highly uncertain, and this uncertainty, coupled with fluctuations in power system load demand, leads to frequent imbalances in power supply and demand within the interconnected grid, causing frequency deviation problems. Extreme weather conditions further exacerbate these output fluctuations, posing an even greater challenge to the power system's power balance and frequency stability.

[0031] To address these challenges, power system risk assessment and early warning technologies require further improvement and optimization. Existing research and practice show that data-driven methods combined with advanced machine learning algorithms can provide more accurate and real-time predictions of power system operating conditions, thus providing decision support for the safe operation of power systems. However, these methods often suffer from insufficient consideration of the impact of extreme weather, limited ability to identify cascading fault propagation paths, and inadequate quantitative assessment of power system operating risks when dealing with power system operating risk assessment under extreme weather conditions. Therefore, developing a new method for early warning of high-risk power system operations to meet the needs of safe power system operation under extreme weather conditions has become a research hotspot in this field.

[0032] like Figure 1 As shown in the figure, this application provides a method for risk assessment and early warning of power system operation status under extreme weather conditions, specifically including:

[0033] Step 101: Input global meteorological data into the load forecasting architecture and output load forecasting results.

[0034] Step 102: Construct a power system fault probability model based on the load forecast results.

[0035] Step 103: Based on the power system fault probability model, determine the initial fault operation scenario results of the power system under extreme weather conditions.

[0036] Step 104: Based on the Nk cascading fault propagation path, construct an operation evaluation scheme according to the results of the initial fault operation scenario of the power system under extreme weather conditions.

[0037] Step 105: Determine the high-risk operation scenario under the current extreme weather conditions based on the operation assessment scheme, and issue an early warning based on the high-risk operation scenario under the extreme weather conditions.

[0038] This application provides a method for risk assessment and early warning of the operating status of a power system under extreme weather conditions. This method aims to improve the operational resilience of the power system under extreme weather conditions. By accurately predicting load changes, simulating power equipment failures, and assessing the risk of cascading failures, it provides decision support for the planning and scheduling of the power system, thereby reducing the losses caused by extreme weather to the power grid and enhancing the overall resilience of the power grid.

[0039] Firstly, this application constructs a load forecasting architecture based on graph neural network feature fusion, which incorporates global meteorological data to improve forecasting accuracy. This architecture enables more accurate prediction of load changes under extreme weather conditions, providing crucial input for power system operational risk assessment.

[0040] This load forecasting architecture utilizes a graph neural network (GNN) to extract meteorological information from each load node within the power system for feature extraction. The GNN updates the meteorological features of the load nodes by aggregating meteorological station information near the load nodes, and achieves deeper feature learning through multi-layer stacking.

[0041] The extracted meteorological features are combined with historical load data to form a comprehensive feature set. This step is achieved through a linear transformation module, mapping high-dimensional features to the same feature space, which facilitates the subsequent training of the load forecasting model. Based on the fused features, a load forecasting model is then trained. This model can predict the load demand at a specific future time point based on the input meteorological information and historical load data.

[0042] Secondly, this application employs a non-sequential Monte Carlo sampling simulation method to construct a set of initial fault operation scenarios for the power system under extreme weather conditions. This method simulates various fault scenarios that may occur in the power system under the influence of extreme weather, providing comprehensive initial conditions for subsequent risk assessment.

[0043] Thirdly, this application designs an operation assessment scheme based on Nk cascading failure propagation path identification to screen out high-risk scenarios. This scheme can effectively identify cascading failure propagation paths that may occur under extreme weather conditions, thereby identifying vulnerable links in the power system and providing decision support for power system operation risk management.

[0044] Based on this, an optimized scheduling model is proposed to optimize the adjustment of remaining generator nodes or load nodes after cascading failures propagate, thereby ensuring the safety of the power system under the new operating mode. This model considers the economic factors of generator output adjustment and aims to minimize the impact of extreme weather on the power grid.

[0045] Fourthly, the effectiveness of the proposed method was verified through simulation results. Simulation results show that the method can effectively identify high-risk fault scenarios under extreme weather conditions, providing decision support for reducing power grid losses and enhancing the resilience and stability of the power grid under extreme weather conditions. By implementing the above technical solution, this application can provide powerful risk warning and management strategies for the safe operation of the power system under extreme weather conditions.

[0046] By implementing the above technical solutions, this application can effectively assess and provide early warning of high-risk operation of the power system under extreme weather conditions, reduce the negative impact on power grid stability, and improve the overall resilience and operational performance of the power grid.

[0047] Thus, this application not only accurately predicts load results but also enhances the sensitivity of monitoring cascading failures caused by extreme weather while ensuring the safe operation of the power grid. This application fully considers the impact of extreme weather on the power system and the vulnerability of power equipment components under abnormal operating conditions, describing the high-risk operation warning method for the power system as a comprehensive model combining data-driven approaches and Nk cascading failure propagation path identification. Based on this, this application employs a non-sequential Monte Carlo sampling simulation method to rigorously construct a set of initial fault operation scenarios for the power system under extreme weather conditions. These criteria are applicable to complex conditions such as power constraints and uncertainties in fault probability caused by meteorological factors. Through these criteria, this application can achieve accurate early warning of high-risk operation risks of the power system under extreme weather conditions, ensuring the stability and reliability of the power grid even in the event of cascading failures caused by extreme weather. Furthermore, the early warning method proposed in this application has high adaptability and flexibility, capable of addressing the challenges of power grid operation under extreme weather conditions, especially maintaining stable grid operation when the output uncertainty of new energy sources is high. Simulation verification shows that the early warning method proposed in this application not only reduces power grid losses and improves decision-making efficiency, but also enhances the power grid's ability to cope with extreme weather and its risk management level, proving its effectiveness and superiority in actual power grid operation.

[0048] Furthermore, in an exemplary embodiment, step 101 can be replaced by the following steps.

[0049] Step 1011: The load prediction architecture is built on a graph neural network and includes a feature extraction layer, a feature transformation layer and a prediction output layer connected in sequence.

[0050] Step 1012: The feature extraction layer includes a GNN (Graph Neural Network), a TCN (Temporal Convolutional Network), and an LSTM (Long Short-Term Memory) module. The GNN module extracts the first high-dimensional meteorological feature from the global meteorological data across the entire power supply area based on meteorological station information. The TCN module, based on the first high-dimensional meteorological feature, combines multi-layer causal convolution and extended convolution to capture the change pattern of the load data within a preset time range. The LSTM module, based on the change pattern of the load data within the preset time range, mines the implicit short-term and long-term temporal features in the load data to obtain the second and third high-dimensional meteorological features. Temporal features refer to features in the data that change over time; they can be short-term (e.g., intraday variations) or long-term (e.g., seasonal variations). High-dimensional meteorological features, on the other hand, are more complex feature representations with more information obtained through feature extraction and transformation based on time-series features. In other words, high-dimensional meteorological features are an extension and deepening of time-series features, with increases in both dimension and complexity.

[0051] This layer comprises three modules: GNN (Graph Neural Network), TCN (Temporal Convolutional Network), and LSTM (Long Short-Term Memory). Since the meteorological information of each load node exhibits spatial correlation, GNN is used to extract high-dimensional meteorological features across the entire power supply area. The TCN and LSTM modules are used to mine the short-term and long-term temporal features implicit in the load data, respectively.

[0052] (1) Weather feature extraction module based on GNN (Graph Neural Network).

[0053] Let the historical records of different weather stations be V = [V1, V2, ..., VΞ ] T .in (φ = 1, 2, ..., Ξ) represents the historical meteorological data of the φ-th load node. Let E be the meteorological data vector for the m-th past time, where m = 1, 2, ..., M; and let E be the set of geographical distances between meteorological stations and load nodes. id |i=1,2,...,Ξ; d=1,2,...,N D}, e id The distance between meteorological station i and load node d; the historical load data for each load node are as follows: N D Ω represents the number of load nodes and Ω represents the length of a load segment.

[0054] In this application, the GNN (Graph Neural Network) aggregates and updates the meteorological features of the load node using information from meteorological stations near the load node, and achieves deeper feature learning through multi-layer stacking. The meteorological feature extraction process based on GNN (Graph Neural Network) is as follows:

[0055]

[0056] In the formula, z∈[1,Z] represents the number of layers in the GNN (Graph Neural Network). Let φ be the aggregation result of weather station φ at layer z, σ be the activation function, N(d) be the set of neighboring weather stations of load node d, R be the preset distance, |N(d)| be the number of neighboring weather stations, and w be the aggregation result of weather station φ at layer z. (z) b is the weight matrix; (z) For bias.

[0057] (2) Load feature extraction module based on TCN and LSTM.

[0058] Temporal Convolutional Networks (TCNs) combine multi-layer causal convolutions and extended convolutions to capture the changing patterns of workload data over short time intervals. By adjusting the size of the convolutional kernels and the stride of the extension, TCNs achieve greater sensitivity to local features. The process of extracting local temporal features using TCNs is illustrated in the following equation.

[0059]

[0060] In the formula, k∈[1,Π] is the number of layers in the TCN (Temporal Convolutional Network); This is the output of the k-th intermediate layer; C TCN The size of the convolution kernel; Φ represents the convolution kernel weights; (k) β represents the expansion rate. (k) For bias.

[0061] LSTM (Long Short-Term Memory) networks use gating mechanisms to ensure the continuous retention of hidden features during forward propagation, thereby capturing long-term features of the load.

[0062]

[0063] η t-ρ =o t-ρ ·tanh(C t-ρ ).

[0064] h LSTM =[η t-Ω ,η t-Ω+1 ,...,η1].

[0065] In the formula, f t-ρ ξ t-ρ and o t-ρ The outputs of the forget gate, input gate, and output gate are respectively; w F w I w C and w O v is the weight matrix; F v I v C and v O For bias; For intermediate memory state; C t-ρ η represents the memory state at time t-ρ. t-ρ h represents the hidden state at time t-ρ. LSTM This is the output sequence of an LSTM (Long Short-Term Memory) network.

[0066] Step 1013: The fourth high-dimensional meteorological feature is:

[0067] h' GNN =w GNN h GNN +b GNN .

[0068] The fifth high-dimensional meteorological feature is:

[0069] h' TCN =w TCN h TCN +b TCN .

[0070] The sixth high-dimensional meteorological feature is:

[0071] h' LSTM =w LSTM h LSTM +b LSTM .

[0072] Among them, h GNN As the first high-dimensional meteorological feature, h TCN As the second highest-dimensional meteorological feature, h LSTM As the third high-dimensional feature, h' GNN As the fourth higher-dimensional meteorological feature, h' TCN As the fifth high-dimensional meteorological feature, h' LSTM As the sixth higher-dimensional meteorological feature, w GNN w TCN and w LSTM Let b be the weight matrix. GNN b TCN and b LSTM For bias.

[0073] This layer contains three linear transformation modules, which respectively transform the high-dimensional features h obtained from the feature extraction layer. GNN h TCN and h LSTM Perform a linear transformation. Where h GNN After linear transformation, a linear combination of meteorological characteristics is obtained, which facilitates the prediction of the probability of extreme weather occurring at a certain point in time; h TCN and h LSTM After linear transformation, they are mapped to the same feature space, which is beneficial for their fusion.

[0074] h′ GNN =w GNN h GNN +b GNN .

[0075] h T ′ CN =w TCN h TCN +b TCN .

[0076] h L ′ STM =w LSTM h LSTM +b LSTM .

[0077] In the formula, w GNN wTCN and w LSTM Let b be the weight matrix. GNN b TCN and b LSTM For bias.

[0078] Step 1014: The prediction output layer specifically includes:

[0079] Using formula Determine the probability of extreme weather events occurring based on load data; where λ d h' represents the probability of extreme weather occurring for the d-th load data point; GNN This represents the meteorological characteristics of the fourth higher dimension.

[0080] Step 1015: Based on the probability of extreme weather occurrence from the d-th load data, use the formula η d =λ d ·h′ TCN +(1-λ d )·h′ LSTM The fifth high-dimensional meteorological feature and the sixth meteorological feature are weighted and fused to determine the fused high-dimensional meteorological feature, where η d This represents the high-dimensional meteorological characteristics after fusion; h' TCN Indicates the fifth higher-dimensional meteorological characteristics; h' LSTM This represents the meteorological characteristics of the sixth higher dimension.

[0081] Step 1016: Based on the fused high-dimensional meteorological characteristics, use formula P d,t =w Θ η d +b Θ Determine the load forecast results, where p d，t This indicates the load forecast result, w Θ Let b represent the weight matrix. Θ Indicates bias.

[0082] To achieve load forecasting under extreme weather conditions, the two time-series features h T ′ CN and h L ′ STM Weighted fusion is performed, with the weights depending on the probability of extreme weather events, which can be obtained from the transformed meteorological information features h′. GNN The result is obtained through logistic regression calculation. The specific process is as follows:

[0083] η d =λ d ·h T ′ CN +(1-λ d )·h′ LSTM .

[0084] In the formula, λ d η represents the probability of extreme weather occurring at the d-th load node; d This represents the fused time-series characteristics. Finally, the fused characteristics undergo a linear transformation to output the load prediction result.

[0085] P d,t =w Θ η d +b Θ .

[0086] In the formula, w Θ Let b be the weight matrix. Θ For bias.

[0087] Furthermore, in an exemplary embodiment, step 102 can be replaced by the following steps.

[0088] Step 1021: Construct a power system fault probability model based on the load forecast results, specifically including:

[0089] Based on the load forecast results, using the formula Determine the failure probability dependent on meteorological factors.

[0090] In the formula, f n,W Let X be the probability of failure of the nth line in the power system, where n = 1, 2, ..., N. L N L For the number of lines; ζ n,X , λ n,X Let be the mean and standard deviation of the failure probability of the i-th line under extreme weather conditions.

[0091] Step 1022: Based on the load forecast results, use the formula Determine the failure probability of road flow dependence.

[0092] In the formula, f n,0 P represents the probability of line n failing during normal operation. n,r and P n,max P represents the rated power and maximum allowable power of line n, respectively. n,s Let n be the power flow magnitude of line n during the propagation of the s-th round of cascading failures.

[0093] Step 1023: Construct a power system fault probability model based on the fault probability dependent on meteorological factors and the fault probability dependent on power flow.

[0094] 1) Failure probability dependent on meteorological factors: The cumulative log-normal distribution function is widely used to describe the vulnerability of components under abnormal operating conditions.

[0095] 2) Fault probability dependent on power flow: When the line is operating normally, the fault probability of the power line is often low; when the power flow of the line is between the rated power and the maximum allowable power, the fault probability is approximately linearly related to the power flow of the line; when the power flow of the line exceeds the maximum allowable power, the line will inevitably be overloaded and disconnected.

[0096] Furthermore, in an exemplary embodiment, step 103 can be replaced by the following steps.

[0097] Step 1031: Based on the power system fault probability model, using formula f n =min{δ n,W f n,W +δ n,F f n,F ,1} Determine the overall line fault probability under extreme weather conditions.

[0098] Step 1032: Based on the comprehensive fault probability of the line, use the non-sequential Monte Carlo sampling simulation method to determine the initial fault operation scenario of the power system under extreme weather conditions.

[0099] Under extreme weather conditions, power systems face the risk of simultaneous failures of multiple power lines. To address this, this section employs a non-sequential Monte Carlo sampling simulation scheme to generate a series of operational scenarios. Specifically, during the k-th sampling, for the i-th power line, a random number γ is randomly generated from the interval [0, 1]. i,k Thus, the state ε of power line i in a single sampling. ik satisfy:

[0100]

[0101] In the formula, 0 represents the normal state and 1 represents the fault state. For all lines in the power system, the operating state after a single sampling is...

[0102] E k ={ε 1,k ,ε 2,k ,...,ε N,k}

[0103] Repeating the sampling operation K times yields K initial fault scenario sets, denoted as {E1, E2, ..., E...}. K}

[0104] Furthermore, in an exemplary embodiment, step 104 can be replaced by the following steps.

[0105] Step 1041: Operational evaluation scheme for the cascading fault propagation stage: During the cascading fault propagation stage, the line power flow under the initial fault scenario is calculated based on the DC power flow model, and it is determined whether the remaining lines have experienced power flow exceeding the limit, thus obtaining the first judgment result.

[0106] Step 1042: If the first judgment result is that the remaining lines have exceeded the power flow limit, disconnect the faulty power system lines, update the power line topology, and propagate the cascading faults.

[0107] Step 1043: If the first judgment result is that no power flow exceeds the limit on the remaining lines, the cascading fault propagation ends.

[0108] Step 1044: Operational evaluation scheme for the optimized scheduling phase: In the optimized scheduling phase, based on the power system operating status after the completion of the cascading fault propagation phase, an optimized scheduling plan is formulated according to the simulation dispatch center.

[0109] Step 1045: Determine the load loss and engine output adjustment amount according to the optimized scheduling plan.

[0110] Design an operational assessment scheme based on Nk chain failure propagation path identification to screen out high-risk scenarios;

[0111] For the set of fault scenarios generated in step 103, it is clear that multiple lines may disconnect simultaneously under extreme weather conditions. Furthermore, the large-scale power flow shift caused by the disconnection of multiple lines may cause some lines to trip due to power flow exceeding limits, triggering cascading faults. Therefore, this section establishes a high-risk operational early warning model that takes into account the propagation of Nk cascading faults, accurately identifies the fault propagation process of the power system under extreme weather conditions, identifies vulnerable lines, and provides a basis for decision-making by planning and dispatching departments. A detailed introduction follows.

[0112] Chain reaction propagation stage:

[0113] When multiple lines in a power system fail and go out of service, it triggers power flow shifts and subsequent line overload trips, creating a cascading fault propagation. Therefore, during the fault propagation phase, the power flow under the initial fault scenario is first calculated based on a DC power flow model to determine if any remaining lines are experiencing power flow exceeding limits. If a power flow exceeding limit occurs, the corresponding faulty lines are disconnected, the power line topology is updated, and the next round of fault propagation begins; conversely, if no power flow exceeding limit occurs, the cascading fault propagation ends. After each round of cascading fault propagation, the set of faulty lines for that round is updated.

[0114]

[0115] In the formula, L s Let s be the set of faulty lines, s∈Z +Indicates the number of times the fault propagated; l n For line numbering, It is a 0-1 variable. When the value is 0, it means that line n has experienced a power flow over-limit accident and has been taken out of service during the s-th round of faults.

[0116] The following constraints must be satisfied during power flow calculation:

[0117] (1) Node power balance constraints

[0118] For any node, the power balance constraint must be satisfied during the s-th round of fault propagation, i.e.

[0119] A s P s =B G,s P G,s -C D,s P D,s .

[0120] In the formula, A s For N×N L Node-line matrix, A s middle element a in,s (i=1,2,…,N; n=1,2,…,N L A value of 1 indicates that node i is connected to line n, and the connection direction is from line n to node i; a value of -1 indicates that the connection direction is from node i to line n; and a value of 0 indicates that node i and line n are not connected. G For N×N G B-order node-generator matrix G,s element b ig,s (i=1,2,…,N; g=1,2,…,N G A value of 1 indicates that node i is connected to generator node g, while a value of 0 indicates that they are not connected; C D,s For N×N D Node-load matrix of order C D,s element c id,s (i=1,2,…,N; d=1,2,…,N D A value of 1 indicates that node i is connected to load node d, while a value of 0 indicates that they are not connected; s This is the line power flow vector during the propagation of the s-th round of cascading failures, with dimension N. L ×1 order; P G,s Let N be the generator output vector during the s-th round of cascading fault propagation. G ×1 order; P D,s Let N be the load vector during the s-th round of cascading failure propagation. D ×1 order.

[0121] For line power flow P s The calculation results were obtained using a DC model:

[0122] P s =B L,s θ s .

[0123] In the formula, B L,s Let N be the admittance matrix for the s-th round of cascading failure propagation. L ×N L Rank, B L,s element b n,s (n = 1, 2, ..., N) L Let θ be the admittance of line n. When line n is disconnected due to extreme weather or cascading fault propagation, the admittance of this line becomes 0; otherwise, the admittance remains unchanged. s The phase vector of the line has a dimension of N. L ×1 order.

[0124] When a cascading fault propagates in the s-th round, whether any line experiences a power flow over-limit trip can be determined using the following formula:

[0125]

[0126] When the cascading failure propagates in the s-th round, the set of faulty lines can be denoted as:

[0127] L s ={n|n=1,2,...,N L ;ε n,s =1}.

[0128] Furthermore, it should be noted that if the failure is due to extreme weather or during the propagation of the cascading failure in round s-1, it may be caused by... If an out-of-limit condition causes all lines connected to the generator node or load node to disconnect, then before performing the s-th round of cascading fault determination, the power of the corresponding generator node or load node must be set to 0. Furthermore, if a generator node or load node exits, causing an imbalance in the node's active power, this violates equation P. s =B L,s θ s At this time, the convergence of DC power flow calculation can be ensured by adjusting the active power of the balancing node.

[0129] After a cascading failure propagates, the simulation dispatch center often needs to optimize and adjust the remaining generator nodes or load nodes to ensure the safety of the power system under the new operating mode. Therefore, this stage involves the simulation dispatch center formulating a dispatch plan based on the grid operating status after the cascading failure propagation is complete. Ultimately, this results in the propagation path of the cascading failure and the load loss and generator output adjustment during the optimized dispatch process. For the k-th simulated fault scenario:

[0130] Chain reaction failure propagation path:

[0131] E k →L1→L2→...→L S .

[0132] Load loss:

[0133]

[0134] Power generation adjustment:

[0135]

[0136] In the formula, ΔP d,D ΔP g,G These are the load and generator adjustment values, respectively.

[0137] Simulated dispatch centers generally aim to minimize load losses to reduce the impact of extreme weather on the power grid. This application further considers the economic factors of generator output adjustment and establishes an optimal dispatch model. The objective function is the operating cost of the fault propagation phase and the optimal dispatch phase, as shown in the following equation.

[0138] J = minμ D P loss,D +μ G P G,∑ .

[0139] In the formula, J represents the power grid operation loss after extreme weather events, and μ D μ G This is the cost coefficient.

[0140] Let S be the number of rounds in the cascading failure propagation. Then the relevant constraints are:

[0141] a) Power balance constraint: Every node must satisfy the power supply and demand conservation, i.e.

[0142]

[0143] In the formula, These represent the optimized line power flow, generator output, and load vector, respectively. Clearly, there are...

[0144]

[0145] b) Line power flow constraints: After optimization and adjustment, the power flow of each line does not exceed the power flow limit and meets the DC power flow calculation conditions.

[0146]

[0147] In the formula, Let θ be the angle of attack at node n. max This is the maximum work angle.

[0148] c) Phase angle limit constraint: To ensure the stability of the power line, the power angle of any node must meet a certain range.

[0149]

[0150] e) Generator output limit constraints: There are upper and lower limits on the active power output of the generator.

[0151]

[0152] In the formula, P g,min P g,max These are the minimum and maximum allowable outputs of the g-th generator, respectively.

[0153] f) Load Adjustment Constraints: Under extreme weather conditions, the adjustable load range is from 0 to the initial load demand, i.e.

[0154]

[0155] For the optimization scheduling model, the objective function includes an absolute value term |ΔP. g,G This problem cannot be solved directly using linear programming tools. Therefore, auxiliary variables are introduced.

[0156]

[0157] This allows the optimization scheduling model to be transformed into a linear programming model, which can then be solved using equal approaches.

[0158] Furthermore, in an exemplary embodiment, step 105 can be replaced by the following steps.

[0159] Step 1051: By repeating the cascading fault propagation stage and the optimization scheduling stage K times, the cascading fault propagation path and corresponding regulation costs caused by extreme weather to the power grid under the initial fault scenario set can be obtained. Then, by sorting the regulation costs in descending order, the hazard level of the operating scenario under extreme weather can be obtained, providing the operation planning department with information to formulate relevant disaster prevention and mitigation measures.

[0160] The effectiveness of the proposed method was verified in a simulation environment. By comparing different control schemes, the superior performance of the proposed scheme under extreme high-temperature weather was demonstrated. A simulation model was built in the MATLAB / Simulink environment, and its performance was tested under extreme high-temperature weather conditions in a city in Northwest China. The line topology is as follows: Figure 2 As shown, the red numbers 1-39 represent line numbers, and the black numbers 1-39 represent node numbers. The probability density distribution function of line failure rate as a function of temperature is shown below. Figure 3 As shown in Table 1, the simulation parameters are set according to the simulation parameters in Table 1. The parameters of the simulation model are set according to the simulation parameters in Table 1, including the fault probability weight coefficient, preset distance, cost coefficient, generator output coefficient, etc.

[0161] Table 1 Simulation Parameter Settings Parameters of the Simulation Model

[0162]

[0163] To verify the predictive performance of the proposed model under extreme weather conditions, this application selects LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), and LSTM-TCN(LT) methods to predict the daily load value of a single node under normal weather, high-temperature weather, and extreme high-temperature weather conditions, respectively, and compares it with the GLT method proposed in this application. To evaluate the algorithm performance, this application uses Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) to quantify the model's prediction bias. The formulas for calculating RMSE and MAE are as follows:

[0164]

[0165] The results are compared in Tables 2 to 4. Table 2 shows the comparison of prediction results under normal weather conditions, Table 3 shows the comparison of prediction results under high temperature weather conditions, and Table 4 shows the comparison of prediction results under extreme high temperature weather conditions.

[0166] Analysis of the results in the table shows that: 1) Under normal weather conditions, user electricity consumption habits remain stable, and the long-term temporal characteristics of the load are more significant. In this case, LSTM (Long Short-Term Memory) outperforms TCN (Temporal Convolutional Network), with RMSE decreasing by 32.43% and MAE decreasing by 25.44%; 2) However, during extreme high-temperature weather, user electricity consumption behavior changes significantly, and the short-term temporal characteristics of the load are more significant. In this case, TCN (Temporal Convolutional Network) outperforms LSTM (Long Short-Term Memory), with RMSE decreasing by 47.04% and MAE decreasing by 40.87%; 3) Under high-temperature conditions, the long-term and short-term temporal characteristics of the load have a relatively balanced impact on the results. In this case, both LSTM (Long Short-Term Memory) and TCN (Temporal Convolutional Network) are more effective than LSTM (Long Short-Term Memory). The proposed GLT method incorporates weather information, allowing the prediction model to dynamically adjust the fusion ratio of long-term and short-term time-series features based on weather conditions. Compared to LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), and LT methods, the proposed method achieves RMSE reductions of over 35.36%, 32.44%, and 33.66%, respectively, and MAE reductions of over 34.82%, 26.71%, and 36.91%, respectively.

[0167] To more intuitively demonstrate the prediction performance of the proposed method, this application selects load nodes and compares the prediction results of LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), LT, and GTL methods under different weather conditions, as shown below. Figure 4 and Figure 5 .

[0168] Table 2 Comparison of forecast results under normal weather conditions

[0169] LSTM TCN LT GLT RMSE 6.73 9.96 7.84 4.35 MAE 5.57 7.47 6.91 3.63

[0170] Table 3 Comparison of forecast results under high temperature weather conditions

[0171] LSTM TCN LT GLT RMSE 10.37 8.07 6.95 4.61 MAE 8.00 7.12 5.13 3.76

[0172] Table 4 Comparison of Prediction Results under Extreme High Temperature Weather Conditions

[0173] LSTM TCN LT GLT RMSE 14.84 7.86 12.44 5.31 MAE 11.5 6.80 9.35 4.29

[0174] This application provides a power system operation status risk assessment and early warning system under extreme weather conditions, specifically including:

[0175] The load forecasting module is used to input global meteorological data into the load forecasting architecture and output load forecasting results.

[0176] The power system initial fault operation scenario result output module is used to determine the power system initial fault operation scenario result under extreme weather conditions based on the power system fault probability model.

[0177] The high-risk early warning module is used to determine the high-risk operation scenario under the current extreme weather conditions based on the operation assessment scheme, and to issue an early warning based on the high-risk operation scenario under the extreme weather conditions.

[0178] This application provides a computer device, specifically including:

[0179] The application includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power system load forecasting and fault propagation risk early warning method described in any one of the above embodiments. This application also provides a computer-readable storage medium, specifically including:

[0180] The storage enables the processor to execute computer program instructions for any of the methods described above.

[0181] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0182] This application uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. In summary, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for risk assessment and early warning of power system operation status under extreme weather conditions, characterized in that, The power system operation status risk assessment and early warning method includes: Global meteorological data is input into the load forecasting architecture, and load forecasting results are output. The global meteorological data includes meteorological observation data, meteorological product data, and internet meteorological data. The load forecasting architecture is built based on a graph neural network. The load forecasting architecture includes a feature extraction layer, a feature transformation layer, and a prediction output layer connected in sequence. The feature extraction layer is used to extract the first high-dimensional meteorological features of the global meteorological data within the power supply area, and to mine the short-term and long-term time-series features implicit in the load data of the global meteorological data to obtain the second and third high-dimensional meteorological features. The feature transformation layer is used to perform linear transformations on the first, second, and third high-dimensional meteorological features to obtain the fourth, fifth, and sixth high-dimensional meteorological features. The prediction output layer is used to perform weighted fusion of the fifth and sixth high-dimensional meteorological features to output the load forecasting results under extreme weather conditions. A power system fault probability model is constructed based on the load forecast results; Based on the power system fault probability model, the results of the initial fault operation scenario of the power system under extreme weather conditions are determined. Based on the Nk cascading fault propagation path, an operation evaluation scheme is constructed according to the initial fault operation scenario of the power system under extreme weather conditions; the operation evaluation scheme includes an operation evaluation scheme for the cascading fault propagation stage and an operation evaluation scheme for the optimized scheduling stage; N is the number of components operating normally in the power system, and k is the number of components that have failed in the system; Based on the operational assessment plan, high-risk operational scenarios under current extreme weather conditions are identified, and early warnings are issued based on these high-risk operational scenarios under extreme weather conditions. Based on the load forecast results, a power system fault probability model is constructed, specifically including: Based on the load forecast results, using the formula Determine the failure probability depending on meteorological factors; In the formula, f n,W Let X be the probability of failure of the nth line in the power system, where n = 1, 2, ..., N. L N L For the number of lines; ζ n,X , λ n,X Let be the mean and standard deviation of the failure probability of the i-th line under extreme weather conditions, respectively; Based on the load forecast results, using the formula Determine the failure probability dependent on road flow; In the formula, f n,0 P represents the probability of line n failing during normal operation. n,r and P n,max These are the rated power and maximum allowable power of line n, respectively; P n,s Let n be the power flow magnitude of line n during the s-th round of cascading failure propagation; A power system failure probability model is constructed based on the failure probability dependent on meteorological factors and the failure probability dependent on power flow. Based on the power system fault probability model, the initial fault operation scenario results of the power system under extreme weather conditions are determined, specifically including: Based on the power system fault probability model, using formula f n =min{δ n,W f n,W +δ n,F f n,F ,1} Determine the overall line fault probability; Based on the overall line fault probability, the results of the initial fault operation scenario of the power system under extreme weather conditions are determined by using the non-sequential Monte Carlo sampling simulation method.

2. The method for risk assessment and early warning of power system operation status under extreme weather conditions according to claim 1, characterized in that, The feature extraction layer includes a GNN module, a TCN module, and an LSTM module; Using the GNN module, the first high-dimensional meteorological features of the global meteorological data within the entire power supply area based on meteorological station information are extracted; Using the TCN module, based on the first high-dimensional meteorological feature, and combined with multi-layer causal convolution and extended convolution, the change pattern of the load data within a preset time range is captured; Using the LSTM module, based on the change pattern of the load data within a preset time range, the short-term and long-term time-series features hidden in the load data are mined to obtain the second high-dimensional meteorological feature and the third high-dimensional meteorological feature.

3. The method for risk assessment and early warning of power system operation status under extreme weather conditions according to claim 1, characterized in that, The fourth high-dimensional meteorological feature is: h’ GNN =w GNN h GNN +b GNN ; The fifth high-dimensional meteorological feature is: h’ TCN =w TCN h TCN +b TCN ; The sixth high-dimensional meteorological feature is: h’ LSTM =w LSTM h LSTM +b LSTM ; Among them, h GNN As the first high-dimensional meteorological feature, h TCN As the second highest-dimensional meteorological feature, h LSTM As the third high-dimensional feature, h' GNN As the fourth higher-dimensional meteorological feature, h' TCN As the fifth high-dimensional meteorological feature, h' LSTM As the sixth higher-dimensional meteorological feature, w GNN w TCN and w LSTM Let b be the weight matrix. GNN b TCN and b LSTM For bias.

4. The method for risk assessment and early warning of power system operation status under extreme weather conditions according to claim 1, characterized in that, The prediction output layer specifically includes: Using formula Determine the probability of extreme weather events occurring based on load data; where λ d h' represents the probability of extreme weather occurring for the d-th load data point; GNN This represents the meteorological characteristics of the fourth higher dimension; Based on the probability of extreme weather occurrence from the d-th load data, the formula η is used. d =λ d ·h′ TCN +(1-λ d )·h′ LSTM The fifth and sixth high-dimensional meteorological features are weighted and fused to determine the fused high-dimensional meteorological features; where η d This represents the high-dimensional meteorological characteristics after fusion; h' TCN Indicates the fifth higher-dimensional meteorological characteristics; h' LSTM This represents the meteorological characteristics of the sixth higher dimension; Based on the fused high-dimensional meteorological characteristics, using formula P d,t =w Θ η d +b Θ Determine the load forecast results; where p d，t This indicates the load forecast result; w Θ Let b represent the weight matrix. Θ Indicates bias.

5. The method for risk assessment and early warning of power system operation status under extreme weather conditions according to claim 1, characterized in that, Based on the Nk cascading fault propagation path, an operation evaluation scheme is constructed according to the results of the initial fault operation scenario of the power system under extreme weather conditions, specifically including: Operational evaluation scheme for the cascading fault propagation phase: During the cascading fault propagation phase, the line power flow under the initial fault scenario is calculated based on the DC power flow model, and it is determined whether the remaining lines have experienced power flow exceeding the limit, thus obtaining the first judgment result; If the first judgment result is that the power flow exceeds the limit of the remaining line, disconnect the faulty line of the power system, update the power line topology and carry out cascading fault propagation; If the first judgment result is that no power flow exceeds the limit on the remaining lines, the propagation of cascading faults ends. Operational evaluation scheme for the optimized scheduling phase: In the optimized scheduling phase, based on the power system operating status after the completion of the cascading fault propagation phase, an optimized scheduling plan is formulated according to the simulation dispatch center; The load loss and engine output adjustment are determined based on the optimized scheduling plan.

6. A power system operation status risk assessment and early warning system under extreme weather conditions, employing the power system operation status risk assessment and early warning method under extreme weather conditions as described in claim 1, characterized in that, include: The load forecasting module is used to input global meteorological data into the load forecasting architecture and output load forecasting results; The power system initial fault operation scenario result output module is used to determine the power system initial fault operation scenario result under extreme weather conditions based on the power system fault probability model. The high-risk early warning module is used to determine the high-risk operation scenario under the current extreme weather conditions based on the operation assessment scheme, and to issue an early warning based on the high-risk operation scenario under the extreme weather conditions.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the power system load forecasting and fault propagation risk early warning method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage enables the processor to execute computer program instructions of any one of claims 1 to 5.

Images