Intelligent power grid mountain fire early warning method, system, equipment and medium

Abstract

The invention discloses an intelligent power grid mountain fire early warning method, system and device and a medium. The method comprises the steps that related data and environment data of a power distribution network are collected; performing anomaly analysis on the power distribution network related data and the environmental data to obtain an anomaly result, and obtaining an anomaly feature according to the anomaly result; and inputting the abnormal features into a prediction model, performing prediction through the prediction model to obtain forest fire information, and generating an early warning strategy based on the forest fire information. According to the invention, through combination of multi-source data acquisition and the Bayesian algorithm, abnormal accurate identification of the operation state and environmental parameters of the power distribution network is realized, and the accuracy of forest fire occurrence probability prediction is improved by using the integrated learning advantage of the random forest algorithm. And a comprehensive early warning strategy is generated based on historical cases and real-time data, and multi-department collaborative response is synchronized, so that the timeliness and reliability of forest fire early warning can be improved, power grid faults and economic losses caused by sudden forest fire are effectively reduced, and a scientific and efficient solution is provided for intelligent power grid forest fire prevention and control.

Description

Technical Field

[0001] The present invention relates to the field of wildfire monitoring technology, and in particular to a smart grid wildfire early warning method, system, equipment and medium. Background Art

[0002] Smart grid is built on the basis of integrated, high-speed two-way communication network. Through the application of advanced sensing and measurement technology, equipment technology, control method and advanced decision support system technology, wildfire, also known as forest fire, is a kind of natural disaster. When wildfire disaster breaks out on a large scale, it often poses a destructive threat to multiple lines of the power grid. At the same time, for the current section with heavy load, the line may trigger a chain tripping accident after tripping due to wildfire, thereby causing large-scale power outage of the power grid. Or, because the transmission line is easy to swing in the wind and frequently contacts with the trees or mountain slopes around the line, it causes short circuit discharge of the transmission line and causes jungle fire. Therefore, wildfire prevention is needed in the protection of the power grid.

[0003] After searching, it was found that a typical wildfire warning system in the prior art is a line wildfire warning method and system based on monitoring information, such as publication number CN112949515A. The method obtains the predicted probability by inputting the monitoring time and monitoring area identifier into the target probability prediction model. When the predicted probability reaches the warning value and there are correlation features between the monitoring information features of the lines of each voltage level, a logical operation is performed on the predicted probability and the monitoring information features to obtain a first logical operation result; finally, if one of the first logical operation result and the second logical operation result is true, a line wildfire alarm is issued. Its main feature is to realize remote warning of line wildfires with ground-to-air linkage, thereby protecting the safe, stable and economical operation of the power grid lines. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a smart grid wildfire early warning method to solve the problem that the existing wildfire early warning system cannot predict the occurrence of wildfires based on existing data information and take corresponding prevention and control measures.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a smart grid wildfire early warning method, comprising:

[0008] Collect distribution network related data and environmental data;

[0009] Performing an abnormality analysis on the distribution network related data and the environmental data to obtain abnormal results, and obtaining abnormal characteristics based on the abnormal results;

[0010] The abnormal features are input into a prediction model, wildfire information is obtained through prediction using the prediction model, and an early warning strategy is generated based on the wildfire information.

[0011] As a preferred solution of the smart grid wildfire early warning method of the present invention, wherein: performing an abnormality analysis on the distribution network related data and environmental data to obtain abnormal results, and obtaining abnormal characteristics based on the abnormal results, includes:

[0012] The joint abnormal probability of distribution network operating parameters and environmental parameters is calculated using the Bayesian algorithm;

[0013] Comparing the combined anomaly probability with historical data, if the anomaly probability of the parameter combination exceeds a set threshold, it is determined that there is a wildfire risk anomaly and an anomaly result is generated;

[0014] Anomaly features include abnormal parameter type, abnormal degree, abnormal time and space coordinates, and historical similar case labels.

[0015] The beneficial effects of this preferred technical solution include: Using a Bayesian algorithm to calculate the joint anomaly probability, we can scientifically quantify the likelihood of wildfires occurring when distribution network operating and environmental parameters are abnormal, based on historical data and prior probabilities. This is more objective and accurate than relying solely on empirical judgment. By comparing the joint anomaly probability with historical data and setting a threshold, we can quickly and accurately determine wildfire risk anomalies, generating clear anomaly results.

[0016] As a preferred embodiment of the smart grid wildfire early warning method of the present invention, the abnormal characteristics are input into a prediction model, and wildfire information is obtained by prediction using the prediction model, including:

[0017] The abnormal feature vector is input into the random forest prediction model, and the random forest prediction model samples and generates multiple sub-training sets, and trains the decision tree in parallel;

[0018] The input abnormal feature vector is classified by splitting based on the randomly selected abnormal feature vector through each decision tree;

[0019] The wildfire probability prediction result is obtained through a voting mechanism. If the wildfire probability is greater than the set probability threshold, it is determined that a fire is about to occur and wildfire information is obtained.

[0020] The benefits of this preferred technical solution include: The random forest prediction model leverages the advantages of ensemble learning by training decision trees in parallel, effectively handling high-dimensional, complex, and anomalous feature vector data. Each decision tree performs split classification based on randomly selected feature vectors, reducing the impact of feature correlation, lowering the risk of model overfitting, and enhancing generalization. A voting mechanism is used to determine the wildfire probability prediction, integrating the judgments of multiple decision trees to make the results more robust and reliable.

[0021] As a preferred solution of the smart grid wildfire early warning method of the present invention, generating an early warning strategy based on the wildfire information includes:

[0022] Retrieve prevention and control plans with high similarity to the wildfire probability prediction results from the historical case database and obtain relevant measures;

[0023] Use crawler technology to obtain the latest prevention and control measures from the Internet of Things, and combine historical measures to generate a comprehensive early warning strategy;

[0024] Based on the impact range of wildfires and early warning strategies, specific tasks are generated for different departments and released simultaneously through multi-channel warning modules.

[0025] The benefits of this optimized technical solution include: Retrieving similar prevention and control solutions from a historical case library allows for learning from past successes, avoiding repeated attempts, and improving response efficiency. Leveraging crawler technology to access the latest prevention and control methods from the Internet of Things (IoT) keeps pace with industry trends, integrating new technologies and methods to make early warning strategies more timely and innovative. Based on the wildfire's impact area and early warning strategy, specific tasks are generated for different departments and released simultaneously through multiple channels, enabling efficient and coordinated operations.

[0026] As a preferred solution of the smart grid wildfire early warning method of the present invention, the collection of distribution network related data and environmental data includes:

[0027] The environmental data includes weather, temperature, humidity, wind speed and direction; the distribution network data includes current, voltage, resistance and power;

[0028] After obtaining the distribution network related data and environmental data, the data is preprocessed, which includes data cleaning, data normalization and data labeling.

[0029] As a preferred solution of the smart grid wildfire early warning method of the present invention, the combined abnormal probability of the distribution network operating parameters and environmental parameters is calculated by the Bayesian algorithm and expressed as:

[0030]

[0031] Among them, P(B|A) is the joint anomaly probability, P(B) is the probability of wildfire occurrence, and P(A) is the data anomaly probability.

[0032] As a preferred solution of the smart grid wildfire early warning method of the present invention, the random forest prediction model includes:

[0033] A grid search algorithm was used to systematically adjust key hyperparameters of the random forest prediction model;

[0034] The search range for the number of decision trees, the search range for the maximum depth, and the search range for the minimum number of sample splits are predefined;

[0035] Using cross-validation, the training data is divided into multiple subsets. The model is trained and validated through the subsets, and the performance indicators of the model on the validation set under different hyperparameter combinations are calculated. By traversing all preset hyperparameter combinations, the combination with the best performance is selected to configure the random forest model.

[0036] In a second aspect, the present invention provides a smart grid wildfire early warning system, comprising: a data acquisition module for collecting distribution network related data and environmental data;

[0037] A feature extraction module is used to perform anomaly analysis on the distribution network related data and environmental data, obtain anomaly results, and obtain anomaly features based on the anomaly results;

[0038] The wildfire prediction module is used to input the abnormal characteristics into the prediction model, obtain wildfire information through prediction through the prediction model, and generate an early warning strategy based on the wildfire information.

[0039] In a third aspect, the present invention provides an electronic device, comprising:

[0040] memory and processor;

[0041] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the smart grid wildfire early warning method are implemented.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the smart grid wildfire warning method.

[0043] Compared with the existing technology, the beneficial effects of the present invention are as follows: the present invention combines multi-source data collection with the Bayesian algorithm to achieve accurate identification of anomalies in the operating status and environmental parameters of the distribution network, and uses the integrated learning advantages of the random forest algorithm to improve the accuracy of the prediction of the probability of wildfire occurrence. It then generates a comprehensive early warning strategy based on historical cases and real-time data and synchronizes multi-department collaborative responses, which can improve the timeliness and reliability of wildfire warnings, effectively reduce power grid failures and economic losses caused by sudden wildfires, and at the same time, through data preprocessing and model optimization mechanisms, enhance the system's adaptability to complex environments, providing a scientific and efficient solution for smart grid wildfire prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 The figure is a schematic diagram of the overall process of a smart grid wildfire early warning method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0047] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a smart grid wildfire early warning method, comprising:

[0048] S100: collects distribution network related data and environmental data;

[0049] S102: Perform an anomaly analysis on distribution network related data and environmental data, obtain an anomaly result, and obtain an anomaly feature based on the anomaly result;

[0050] S104: Input the abnormal features into the prediction model, obtain wildfire information through prediction by the prediction model, and generate an early warning strategy based on the wildfire information.

[0051] It should be noted that first, distribution network related data and environmental data are collected through S100. These data are the basis for subsequent analysis; in S102, the collected data is analyzed for anomalies, and the Bayesian algorithm is used to calculate the joint anomaly probability and compare it with historical data. Parameter anomalies that may cause wildfires can be found, and the acquired anomaly features provide key information for prediction. The anomaly features are input into the random forest prediction model. Its unique sampling, training and classification mechanism can accurately predict wildfire information. The early warning strategy generated based on this can effectively guide relevant departments to make preparations in advance. The present invention realizes a complete process from data collection to risk judgment to early warning, which can improve the accuracy and timeliness of smart grid wildfire warnings, and provide a strong guarantee for timely prevention and control measures and reducing wildfire losses.

[0052] Example 2, reference Figure 1 , which is an embodiment of the present invention, provides a smart grid wildfire early warning method based on the above embodiment.

[0053] In the embodiment of the present invention, collecting distribution network related data and environmental data in step S100 specifically includes:

[0054] Use satellite remote sensing technology to collect relevant information about wildfires, including time, location, credibility, longitude and latitude, burned area, and interference sources;

[0055] Environmental data includes weather, temperature, humidity, wind speed and direction; distribution network data includes current, voltage, resistance and power;

[0056] After obtaining the distribution network related data and environmental data, the data is preprocessed, which includes data cleaning, data normalization and data labeling.

[0057] It should be noted that by collecting core information such as wildfire time and location through satellite remote sensing technology, combined with the comprehensive collection of environmental data and distribution network data, a multi-dimensional and three-dimensional data system can be constructed to provide a rich and accurate data foundation for wildfire warning. In the data preprocessing link, data cleaning removes invalid data to ensure data quality; data normalization unifies the data scale to facilitate efficient algorithm processing; data annotation gives data time, space and event attributes to enhance the analyzability of the data. The present invention improves the availability and reliability of data, and can also effectively reduce the errors of subsequent anomaly analysis and prediction models, so that the system can more keenly capture wildfire risk signals, laying a solid data foundation for the accuracy, timeliness and effectiveness of smart grid wildfire warnings, and helping relevant departments to take targeted prevention and control measures in advance to reduce the threat of wildfires to the power grid.

[0058] In an optional implementation, data preprocessing can also include data dimensionality reduction. This involves removing redundant features through methods such as principal component analysis or linear discriminant analysis, reducing data dimensionality and improving the efficiency of subsequent model training. For example, strongly correlated parameters such as distribution network current, voltage, and resistance can be mapped to a low-dimensional space, preserving essential information while reducing computational complexity.

[0059] In the embodiment of the present invention, the spreading direction of mountain activities is predicted by GIS, and the fire situation of the distribution network is monitored in real time by thermal imaging technology.

[0060] In an optional implementation, distribution network-related data and environmental data can be obtained through edge computing devices, and micro-computing units can be deployed at sensor terminals to filter invalid data and compress key parameters in real time, and then transmitted through the 5G network.

[0061] In another optional implementation, distribution network-related data and environmental data can be obtained through drone inspections. Drones equipped with high-precision sensors fly along preset routes, collecting real-time temperature field, vegetation status, and micro-meteorological data around the lines, and combining GPS positioning to achieve precise spatiotemporal data annotation.

[0062] In the embodiment of the present invention, step S102 performs an abnormality analysis on the distribution network related data and the environmental data to obtain abnormal results, and obtains abnormal characteristics based on the abnormal results, further comprising sub-steps A1-A3:

[0063] A1: Calculate the joint abnormal probability of distribution network operating parameters and environmental parameters using the Bayesian algorithm;

[0064] A2: Compare the combined anomaly probability with historical data. If the anomaly probability of the parameter combination exceeds the set threshold, it is determined that there is a wildfire risk anomaly and an anomaly result is generated.

[0065] A3: Anomaly features include the anomaly parameter type, anomaly degree, the spatiotemporal coordinates of the anomaly occurrence, and labels of historical similar cases.

[0066] In the embodiment of the present invention, the joint abnormal probability of the distribution network operating parameters and environmental parameters is calculated by the Bayesian algorithm and expressed as:

[0067]

[0068] Among them, P(B|A) is the joint anomaly probability, P(B) is the probability of wildfire occurrence, and P(A) is the data anomaly probability.

[0069] In an optional embodiment, the abnormal characteristics may also include the changing trend of abnormal parameters, such as the continuous rising rate of current and the frequency of wind speed mutation; multi-parameter coupling correlation, such as the probability of the coordinated occurrence of voltage anomaly and humidity anomaly, etc.

[0070] In an optional implementation, the isolation forest algorithm in machine learning can be used to obtain abnormal results. First, the distribution network operating parameters (current, voltage, and resistance) and the environmental parameters (temperature, humidity, and wind speed) are combined into a high-dimensional data vector and input into the isolation forest model. The model recursively constructs multiple isolation trees by randomly selecting features and split points.

[0071] Each tree calculates the path length from the root node to the leaf node for each data point. Since abnormal data is usually sparse in the feature space, its path length is significantly shorter than that of normal data. The model calculates the anomaly score of each data point by integrating the path length results of multiple trees. When the score exceeds the preset threshold, it is determined that there is a wildfire risk anomaly.

[0072] In another optional implementation, the ARIMA model in time series analysis can be used to obtain abnormal results. First, the distribution network parameters and environmental parameters are preprocessed separately, including noise removal and missing value filling. Then, an ARIMA model is constructed for each parameter temperature series and minute-by-minute current data. The model order is determined by automatically identifying the data's autocorrelation, differential stationarity, and moving average properties.

[0073] After the ARIMA model is trained, a rolling forecast is performed on the real-time data. The residuals between the actual observed values ​​and the predicted values ​​are calculated, and the mean and standard deviation of the residuals are calculated. If the absolute value of the residual at a certain moment exceeds a set multiple of the standard deviation, the parameter is considered abnormal.

[0074] For example, if the predicted wind speed for a certain period is 5m / s, the actual observed value is 12m / s, and the residual is 7m / s, which exceeds three times the standard deviation of the historical residual, then it is considered a wind speed anomaly. Combining the anomaly detection results of multiple parameters, a comprehensive assessment is made to determine whether there is a wildfire risk anomaly.

[0075] It should be noted that the present invention uses a Bayesian algorithm to calculate the joint anomaly probability of distribution network operating parameters and environmental parameters. This method can quantify the degree of correlation between parameter anomalies and wildfire risk based on historical data and prior knowledge, converting real-time multi-source data into measurable risk indicators. When the joint anomaly probability exceeds a threshold, a risk anomaly is determined and an anomaly signature is generated, including parameter type, degree, spatiotemporal coordinates, and historical case labels. This not only accurately identifies wildfire risk but also provides structured input information for subsequent prediction models, enabling the system to quickly locate high-risk areas and trace similar historical scenarios, thereby improving the accuracy and pertinence of early warnings.

[0076] In an embodiment of the present invention, in step S104, the abnormal characteristics are input into the prediction model, and wildfire information is obtained through prediction by the prediction model. Based on the wildfire information, an early warning strategy is generated, which also includes sub-steps B1-B3:

[0077] B1: Input the abnormal feature vector into the random forest prediction model, which samples and generates multiple sub-training sets and trains the decision trees in parallel;

[0078] B2: Classify the input abnormal feature vectors by splitting them based on the randomly selected abnormal feature vectors through each decision tree;

[0079] B3: Obtain wildfire probability prediction results through a voting mechanism. If the wildfire probability is greater than the set probability threshold, it is determined that a fire is about to occur and wildfire information is obtained.

[0080] Specifically, the abnormal feature vector set is sampled in parallel at each node, and the Bootstrap sampling method is repeatedly used to sample, and k sub-training sets are obtained. At the same time, the abnormal feature vectors that are not sampled constitute the out-of-bag data set, and finally k data sets are obtained. These out-of-bag data are used to test the prediction performance of the random forest model;

[0081] Perform decision tree modeling on each node, use the above k sub-training sets to build k decision models, collect k decision tree models to build a random forest model; vote on the predicted samples, and the one with the most votes is the final category.

[0082] In the embodiment of the present invention, the randomness of the random forest algorithm is mainly reflected in the following two aspects: the training samples of the sub-model are randomly extracted and the characteristic variables of the sub-model are also randomly extracted.

[0083] Specifically, the random forest algorithm includes sample randomness and feature randomness. Sample randomness assumes that the training data set has M objects in total, and N samples are randomly extracted from the sample data with replacement (Boostrap). Because of the replacement extraction, some data may be selected multiple times, and some data may not be selected. The samples taken out each time are not exactly the same. These samples constitute the training data set of the decision tree;

[0084] Random features means assuming that each sample data has K features, randomly selecting k (k ≤ K) features from all features, and selecting the best split attributes as nodes to establish a CART decision tree. The size of k remains unchanged during the growth of the decision tree.

[0085] Repeat the previous steps to build M CART trees. These trees are fully grown and not pruned. These trees form a random forest.

[0086] In an embodiment of the present invention, a grid search algorithm is used to systematically adjust the key hyperparameters of the random forest prediction model; the search range for the number of decision trees, the search range for the maximum depth, and the search range for the minimum number of sample splits are predefined; cross-validation is used to divide the training data into multiple subsets, and the model is trained and verified through the subsets. The performance indicators of the model on the verification set under different hyperparameter combinations are calculated, and by traversing all preset hyperparameter combinations, the combination with the best performance is selected to configure the random forest model.

[0087] It should be noted that the present invention effectively improves the model's generalization ability and anti-overfitting ability for abnormal features through the sample randomness and feature randomness mechanism of the random forest algorithm, combined with parallel sampling and out-of-bag data verification, thereby ensuring the robustness of the wildfire probability prediction; the hyperparameter optimization process of grid search and cross-validation enables the model to achieve optimal configuration in key parameters such as the number of decision trees and maximum depth, further exploring potential correlations in the data, thereby accurately outputting wildfire information.

[0088] In an optional implementation, a prediction model can be constructed and forecasted using a gradient boosting decision tree method. First, the anomaly parameter type, degree, and spatiotemporal coordinates of the anomaly feature vector are standardized, and the categorical features are one-hot encoded to form a uniformly scaled feature matrix. A weak learner is initialized and an initial value is predicted for all samples, namely, the mean wildfire probability.

[0089] The iterative training phase then begins. In each iteration, the residual between the current model's prediction and the actual wildfire occurrence label is calculated. This residual is used as the new target value to train a new decision tree, whose goal is to fit the residual. The predictions of the new tree are added to the original model at a preset learning rate, gradually reducing the residual. Sample weights are dynamically adjusted, increasing the weight of samples incorrectly predicted by the previous model. This process is repeated, ultimately resulting in an ensemble model composed of multiple decision trees. When predicting a new sample, the predictions of all trees are weighted and summed, and then converted into a wildfire probability using a logistic function.

[0090] In another optional embodiment, a prediction model can be constructed using a neural network model and predictions can be made; the normalized abnormal feature vector is input into the input layer of the neural network, multiple hidden layers are set in the middle of the network, each layer contains a number of neurons, and the input is transformed by a nonlinear activation function to extract high-order correlations between features;

[0091] The output layer uses a sigmoid activation function, outputting a wildfire probability value between 0 and 1. During model training, stochastic gradient descent or its variants are used to minimize the cross-entropy loss between the predicted probability and the true label. To prevent overfitting, a dropout layer is added to the hidden layer to randomly drop neurons, and L2 regularization is applied to the weights. Training uses a batch training approach, dividing the data into small batches for iterative updates. A validation set is used to monitor model performance, triggering early stopping when the validation set loss no longer decreases. When predicting new samples, the feature vector is input into the trained network, and the wildfire probability is directly output.

[0092] In the embodiment of the present invention, step S104 generates an early warning strategy based on the wildfire information, and further includes sub-steps C1-C3:

[0093] C1: Retrieve prevention and control plans with high similarity to the wildfire probability prediction results from the historical case database and obtain relevant measures;

[0094] C2: Use crawler technology to obtain the latest prevention and control measures from the Internet of Things and combine historical measures to generate a comprehensive early warning strategy;

[0095] C3: Generate specific tasks for different departments based on the impact range of wildfires and early warning strategies, and release them simultaneously through multi-channel warning modules.

[0096] In an optional embodiment, based on the prediction results, a crawler is used to search and crawl the prevention and control measures on the Internet of Things. After the measures are sent, the key words and phrases are extracted and concentrated, duplicates are removed, and a corresponding prevention and control plan is generated based on reasonable measures used in the past.

[0097] In an optional embodiment, multiple units such as firefighting, public security, electricity, forestry production, first aid, emergency response, and firefighting teams are warned simultaneously through multiple channels to ensure clear communication of warning information and the normal implementation of rescue work;

[0098] In an optional embodiment, after the rescue or treatment measures are completed, multiple real-time measures are recorded, and the advantages and disadvantages existing after the real-time measures are reflected and backed up, that is, the backup information provides a reference for subsequent prevention and control measures.

[0099] It should be noted that the present invention retrieves similar prevention and control plans from the historical case library, can quickly call effective measures verified by practice, avoid repeated R&D costs, and improve emergency response speed; use crawler technology to capture the latest prevention and control methods from the Internet of Things in real time, and combine them with historical measures to form a comprehensive strategy to ensure that the plan incorporates cutting-edge industry technologies and enhances adaptability to new wildfire risks; generate specific tasks for different departments based on the wildfire impact range and early warning strategies and publish them simultaneously through multiple channels, breaking down departmental information barriers, and realizing full-process coordination of monitoring-prediction-disposal, maximizing the time difference from early warning to response, and reducing the actual damage of wildfires to the power grid and surrounding environment.

[0100] In Example 3, the above is a schematic scheme of a smart grid wildfire early warning method. It should be noted that the technical solution of this smart grid wildfire early warning system and the technical solution of the smart grid wildfire early warning method described above are based on the same concept. For details not described in detail in the technical solution of the smart grid wildfire early warning system in this embodiment, please refer to the description of the technical solution of the smart grid wildfire early warning method described above.

[0101] This embodiment also provides a smart grid wildfire early warning system, including:

[0102] Data acquisition module, used to collect distribution network related data and environmental data;

[0103] Feature extraction module, used to perform anomaly analysis on distribution network related data and environmental data, obtain anomaly results, and obtain anomaly features based on the anomaly results;

[0104] The wildfire prediction module is used to input abnormal features into the prediction model, obtain wildfire information through prediction through the prediction model, and generate early warning strategies based on the wildfire information.

[0105] This embodiment also provides an electronic device suitable for smart grid wildfire warning situations, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the smart grid wildfire warning method proposed in the above embodiment.

[0106] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for realizing smart grid wildfire early warning as proposed in the above embodiment is implemented.

[0107] The storage medium proposed in this embodiment and the method for implementing smart grid wildfire warning proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, please refer to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0108] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A smart grid wildfire early warning method, characterized in that: include: Collect distribution network related data and environmental data; Performing anomaly analysis on the distribution network related data and environmental data to obtain anomaly results, and obtaining anomaly features based on the anomaly results; The abnormal features are input into a prediction model, wildfire information is obtained through prediction using the prediction model, and an early warning strategy is generated based on the wildfire information.

2. The smart grid wildfire early warning method according to claim 1, characterized in that: Performing an abnormality analysis on the distribution network related data and the environmental data to obtain abnormal results, and obtaining abnormal characteristics based on the abnormal results, including: The joint abnormal probability of distribution network operating parameters and environmental parameters is calculated using the Bayesian algorithm; Comparing the combined anomaly probability with historical data, if the anomaly probability of the parameter combination exceeds a set threshold, it is determined that there is a wildfire risk anomaly and an anomaly result is generated; Anomaly features include abnormal parameter type, abnormal degree, abnormal time and space coordinates, and historical similar case labels.

3. The smart grid wildfire early warning method according to claim 2, characterized in that: Inputting the abnormal features into a prediction model, and performing prediction using the prediction model to obtain wildfire information, including: The abnormal feature vector is input into the random forest prediction model, and the random forest prediction model samples and generates multiple sub-training sets, and trains the decision tree in parallel; The input abnormal feature vector is classified by splitting based on the randomly selected abnormal feature vector through each decision tree; The wildfire probability prediction result is obtained through a voting mechanism. If the wildfire probability is greater than the set probability threshold, it is determined that a fire is about to occur and wildfire information is obtained.

4. The smart grid wildfire early warning method according to claim 3, characterized in that: Based on the wildfire information, an early warning strategy is generated, including: Retrieve prevention and control plans with high similarity to wildfire probability prediction results from the historical case database and obtain relevant measures; Use crawler technology to obtain the latest prevention and control measures from the Internet of Things, and combine historical measures to generate a comprehensive early warning strategy; Based on the impact range of wildfires and early warning strategies, specific tasks are generated for different departments and released simultaneously through multi-channel warning modules.

5. The smart grid wildfire early warning method according to claim 4, characterized in that: Collect distribution network related data and environmental data, including: The environmental data includes weather, temperature, humidity, wind speed and direction; the distribution network data includes current, voltage, resistance and power; After obtaining the distribution network related data and environmental data, the data is preprocessed, which includes data cleaning, data normalization and data labeling.

6. The smart grid wildfire early warning method according to claim 2, characterized in that: The joint abnormal probability of distribution network operation parameters and environmental parameters calculated by Bayesian algorithm is expressed as: Among them, P(B|A) is the joint anomaly probability, P(B) is the probability of wildfire occurrence, and P(A) is the data anomaly probability.

7. The smart grid wildfire early warning method according to claim 5, characterized in that: Random forest prediction models, including: A grid search algorithm was used to systematically adjust key hyperparameters of the random forest prediction model; The search range for the number of decision trees, the search range for the maximum depth, and the search range for the minimum number of sample splits are predefined; Using cross-validation, the training data is divided into multiple subsets. The model is trained and validated through the subsets, and the performance indicators of the model on the validation set under different hyperparameter combinations are calculated. By traversing all preset hyperparameter combinations, the combination with the best performance is selected to configure the random forest model.

8. A smart grid wildfire early warning system, applying the method according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, used to collect distribution network related data and environmental data; A feature extraction module is used to perform anomaly analysis on the distribution network related data and environmental data, obtain anomaly results, and obtain anomaly features based on the anomaly results; The wildfire prediction module is used to input the abnormal characteristics into the prediction model, obtain wildfire information through prediction through the prediction model, and generate an early warning strategy based on the wildfire information.

9. An electronic device, characterized in that: include: A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the smart grid wildfire warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that It stores computer-executable instructions, which, when executed by a processor, implement the steps of the smart grid wildfire early warning method according to any one of claims 1 to 7.

Images