Power grid power failure area prediction method and system under typhoon disaster, storage medium and electronic equipment

Through machine learning models based on historical data to predict power grid power outage areas under typhoon disasters, the problem of lack of prediction mechanisms in the existing technology is solved, and higher prediction accuracy and operation and maintenance guarantees are achieved.

CN119005398APending Publication Date: 2024-11-22ELECTRIC POWER RES INST OF STATE GRID ZHEJIANG ELECTRIC POWER COMAPNY +3
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411012362.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing technology lacks an effective power outage prediction mechanism for power grids under typhoon disasters, which makes it impossible for operation and maintenance personnel to take measures in advance, increasing the losses and impact of the disaster on the power grid and personnel.

Method used

A sample set is generated based on the state data related to historical typhoon disasters, and different dimensionality reduction methods and machine learning algorithm models are used to predict whether the power grid area to be predicted is outage by pre-processing the data and selecting the optimal model.

Benefits of technology

It improves the accuracy and accuracy of power grid power outage areas forecasting during typhoon disasters, and ensures that operation and maintenance personnel can take measures in advance to reduce losses and impacts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119005398A_ABST
    Figure CN119005398A_ABST
Patent Text Reader

Abstract

The invention provides a power grid power failure area prediction method and system under a typhoon disaster, a storage medium and electronic equipment, and the method comprises the following steps: generating a sample set based on the historical typhoon disaster related state data of a to-be-predicted power grid area; respectively processing the sample set by adopting different dimension reduction modes to generate a corresponding training set; training different machine learning algorithm models by adopting different training sets to obtain a plurality of different prediction models; determining an optimal prediction model according to the evaluation index of each prediction model; and inputting the current typhoon disaster related state data of the to-be-predicted power grid region into the optimal prediction model to obtain a prediction result of whether the to-be-predicted power grid region is powered off or not. The method can effectively predict whether power failure occurs in a power grid area under typhoon disasters, thereby effectively guiding operation and maintenance first-aid repair personnel to carry out maintenance and guarantee work.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power systems, and specifically relates to a method, system, storage medium and electronic equipment for predicting power outage areas in power grids under typhoon disasters. Background Art

[0002] Typhoons are a type of extreme weather with extremely strong destructive power. In recent years, their frequency and intensity have been increasing. The power grid is damaged to varying degrees by typhoons almost every year, and the frequency of disaster-causing typhoons is on the rise. During typhoons, power system failures frequently occur, causing serious economic losses and having a serious impact on daily life and production.

[0003] The power system has complex structures and diverse equipment types, which makes it very easy to have large-scale, cascading failures after extreme disasters, and it is difficult to evaluate them. There is a lack of prediction mechanisms on the market to determine whether the power grid will be out of power after a typhoon disaster, which makes it impossible for operation and maintenance personnel to take effective measures in advance to reduce the losses and impacts of typhoons on the power grid and its users. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the above-mentioned background technology and to provide a method, system, storage medium and electronic device for predicting power outage areas in power grids under typhoon disasters, which can effectively predict whether power outages will occur in power grid areas under typhoon disasters, thereby effectively guiding operation and maintenance repair personnel to carry out maintenance and guarantee work.

[0005] The technical solution adopted by the present invention is: a method for predicting power outage areas of power grids under typhoon disasters, comprising the following steps:

[0006] Based on the historical typhoon disaster related status data of the power grid area to be predicted, a sample set is generated; the input of the sample set is the typhoon disaster related status data of the power grid area to be predicted, and the training label is the classification conclusion of power outage or no power outage;

[0007] Use different dimensionality reduction methods to process sample sets respectively and generate corresponding training sets;

[0008] Use different training sets to train different machine learning algorithm models to obtain several different prediction models;

[0009] Determine the optimal prediction model based on the evaluation indicators of each prediction model;

[0010] The current typhoon disaster related status data of the power grid area to be predicted is input into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

[0011] In the above technical solution, the typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

[0012] In the above technical solution, the meteorological information includes the maximum wind speed, wind direction, first set wind speed duration, second set wind speed duration, rainfall, temperature and humidity in the forecast area.

[0013] In the above technical solution, the geographic information includes longitude, latitude, presence or absence of users, altitude, slope direction, underlying surface type and surface type.

[0014] In the above technical solution, the electrical equipment information includes: the number of users, the number of transformers, the number of box transformers, the number of poles and towers, the number of guy wires, the number of non-guy wires and the line length.

[0015] In the above technical scheme, during the training and use of the prediction model, the typhoon disaster-related status data of the power grid area to be predicted is preprocessed and used as the input of the prediction model; the preprocessing process includes: using unique hot encoding to process the character categorical variables in the typhoon disaster-related status data; and normalizing all typhoon disaster-related status data.

[0016] In the above technical solution, the dimension reduction method includes feature selection and principal component analysis.

[0017] In the above technical solution, the machine learning algorithm models include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model.

[0018] In the above technical solution, the evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0019] In the above technical solution, the value of each evaluation index of any prediction model is calculated respectively, and the average value of all evaluation indexes of the prediction model is obtained as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model.

[0020] In the above technical solution, during the training process of any prediction model, different hyperparameter combinations are generated for the prediction model; based on the corresponding training set, the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models; and the initial model with the highest evaluation index score is selected as the prediction model.

[0021] In the above technical solution, the evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0022] In the above technical solution, during the training process of any prediction model, it also includes: calculating the value of each evaluation index for any trained initial model, and obtaining the mean value of all evaluation indexes of the initial model as the evaluation index score of the initial model.

[0023] The present invention also provides a system for predicting power outage areas in power grids under typhoon disasters, including a sample set generation model, a training set generation module, a model training module, a model selection module, and a prediction execution module; wherein,

[0024] The sample set generation model is used to generate a sample set based on the historical typhoon disaster related status data of the power grid area to be predicted;

[0025] The training set generation module is used to process the sample sets respectively using different dimensionality reduction methods to generate corresponding training sets;

[0026] The model training module is used to train different machine learning algorithm models using different training sets to obtain several different prediction models;

[0027] The model selection module is used to determine the optimal prediction model according to the evaluation index of each prediction model;

[0028] The prediction execution module is used to input the current typhoon disaster-related status data of the power grid area to be predicted into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

[0029] In the above technical solution, the typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

[0030] In the above technical solution, the meteorological information includes the maximum wind speed, wind direction, first set wind speed duration, second set wind speed duration, rainfall, temperature and humidity in the forecast area.

[0031] In the above technical solution, the geographic information includes longitude, latitude, presence or absence of users, altitude, slope direction, underlying surface type and surface type.

[0032] In the above technical solution, the electrical equipment information includes: the number of users, the number of transformers, the number of box transformers, the number of poles and towers, the number of guy wires, the number of non-guy wires and the line length.

[0033] In the above technical scheme, the model training module pre-processes the typhoon disaster-related status data of the power grid area to be predicted during the training process of the prediction model, and the prediction execution module pre-processes the typhoon disaster-related status data as the input of the prediction model during the use of the prediction model; the pre-processing process includes: using unique hot encoding to process the character classification variables in the typhoon disaster-related status data; and normalizing all typhoon disaster-related status data.

[0034] In the above technical solution, the dimensionality reduction method adopted by the training set generation module includes feature selection and principal component analysis.

[0035] In the above technical solution, the machine learning algorithm models used in the model training module include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model.

[0036] In the above technical solution, the evaluation indicators used by the model selection module include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0037] In the above technical solution, the model selection module calculates the value of each evaluation index for any prediction model, and obtains the average value of all evaluation indexes of the prediction model as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model.

[0038] In the above technical solution, during the training process of any prediction model, the model training module generates different hyperparameter combinations for the prediction model; based on the corresponding training set, the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models; and the initial model with the highest evaluation index score is selected as the prediction model.

[0039] In the above technical solution, the evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0040] In the above technical solution, the model training module also includes, during the training process of any prediction model: calculating the value of each evaluation index for any initial model that has been trained, and obtaining the mean value of all evaluation indexes of the initial model as the evaluation index score of the initial model.

[0041] The present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for predicting power outage areas in a power grid under a typhoon disaster described in the above technical solution is implemented.

[0042] The present invention also provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for predicting power outage areas in power grids under typhoon disasters as described in the above technical solution by executing the computer instructions.

[0043] The beneficial effects of the present invention are as follows: the present invention uses the historical data of the power grid area associated with the typhoon disaster as a sample set to train the prediction model, ensuring that the prediction model can effectively reflect the correlation between the state of the power grid area and whether there is a power outage, and effectively ensuring the prediction accuracy. The present invention constructs several different prediction models through different dimensionality reduction methods and machine learning models, and selects the optimal model for subsequent application based on the evaluation index, fully considering the generalization of different typhoon conditions and different power grid states, and ensuring the prediction accuracy of the model under different conditions.

[0044] Furthermore, the present invention uses the meteorological information, geographic information and electrical equipment information of the power grid area to be predicted as the input of the prediction model, and can fully consider the data associated with the typhoon disaster in the prediction result to ensure the prediction accuracy. At the same time, the specific meteorological information, geographic information and electrical equipment information selected by the present invention can effectively highlight the characteristics of the typhoon disaster and its impact range and capacity, and further improve the prediction accuracy of whether there will be a power outage.

[0045] Furthermore, the present invention uses one-hot encoding and normalization to preprocess the model input data, which effectively improves the efficiency of the model in processing data and improves the computational cost and accuracy of the model.

[0046] Furthermore, the dimension reduction method selected by the present invention includes feature selection and principal component analysis, which can effectively handle data redundancy and improve the computational efficiency of the prediction model.

[0047] Furthermore, the machine learning algorithm models selected by the present invention include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, support vector machine model, which can meet the needs of different data characteristics, ensure the breadth of the selection range of the optimal prediction model, and further improve the accuracy of the prediction model.

[0048] Furthermore, the evaluation indicators selected by the present invention include the area under the receiver operating characteristic curve, accuracy and F1 score; and can effectively select the optimal hyperparameter combination and the optimal model to ensure the prediction accuracy of the model.

[0049] Furthermore, the present invention selects the optimal hyperparameter combination for each model through evaluation indicators, thereby effectively improving the prediction accuracy of the model.

[0050] Furthermore, the present invention fully considers the diverse characteristics of the model by calculating the mean of the three evaluation indicators as the evaluation criteria for the hyperparameter combination and the prediction model, ensuring that the final selected optimal model can effectively reflect the characteristics of the area to be tested and further improve the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0052] Figure 2 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments to facilitate a clear understanding of the present invention, but they do not constitute a limitation on the present invention.

[0054] Example 1

[0055] like Figure 1 As shown, the present invention provides a method for predicting power outage areas in a power grid under a typhoon disaster, comprising the following steps:

[0056] Based on the historical typhoon disaster related status data of the power grid area to be predicted, a sample set is generated; the input of the sample set is the typhoon disaster related status data of the power grid area to be predicted, and the training label is the classification conclusion of power outage or no power outage;

[0057] Use different dimensionality reduction methods to process sample sets respectively and generate corresponding training sets;

[0058] Use different training sets to train different machine learning algorithm models to obtain several different prediction models;

[0059] Determine the optimal prediction model based on the evaluation indicators of each prediction model;

[0060] The current typhoon disaster related status data of the power grid area to be predicted is input into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

[0061] Specifically, the method further includes the following steps: using Geographic Information System (GIS) technology to visualize the prediction results of the optimal prediction model.

[0062] Specifically, the typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

[0063] Preferably, the meteorological information includes the maximum wind speed, wind direction, duration of a wind speed of 20 m / s, duration of a wind speed of 30 m / s, rainfall, temperature and humidity in the forecast area.

[0064] Preferably, the geographic information includes longitude, latitude, presence or absence of users, altitude, slope aspect, underlying surface type and surface type.

[0065] Preferably, the electrical equipment information includes: the number of users, the number of station transformers, the number of box transformers, the number of pole towers, the number of guy wires, the number of non-guy wires and the line length.

[0066] Specifically, during the training and use of the prediction model, the typhoon disaster-related status data of the power grid area to be predicted is preprocessed and used as the input of the prediction model; the preprocessing process includes: using one-hot encoding to process the character-type categorical variables in the typhoon disaster-related status data; and normalizing all typhoon disaster-related status data.

[0067] If a character-type categorical variable is a three-category variable, a three-digit code value is formed, and the coding format is as follows:

[0068] Table 1 Coding of three-category character variables

[0069]

[0070] Data normalization includes centering and scaling to fix the data within [0,1]. The specific calculation formula is:

[0071]

[0072] Among them: X' is the normalized data, X is the original sample data, X min is the minimum value of the feature data, X max is the maximum value of the feature data.

[0073] Specifically, the dimensionality reduction method includes Feature Selection (FS) and Principal Component Analysis (PCA). Several features are used as inputs of the model in the sample set. In order to streamline the sample set, the dimensionality reduction method is used to reduce the number of model input features.

[0074] FS uses feature contribution scores to determine the value and importance of each feature to the model: Based on the random forest model, the contribution score of each feature is determined 100 times, and the first several features with the highest feature contribution scores are selected as subsequent model inputs.

[0075] Principal Component Analysis (PCA) is a linear transformation technique for data dimensionality reduction. It reduces the dimension of the data by projecting the original data into a new coordinate system while retaining the main information of the data as much as possible:

[0076] PCA is sensitive to the scale of the data, so the data must first be standardized so that each feature has a distribution with a mean of 0 and a standard deviation of 1.

[0077] Then the covariance matrix is ​​calculated, which is used to measure the linear relationship between features.

[0078] Then calculate the covariance matrix, which is used to measure the linear relationship between features. Sort by the size of the eigenvalues ​​and select the eigenvectors corresponding to the first k eigenvalues. These eigenvectors are the new coordinate axes, i.e., the principal components. The larger the selected eigenvalue, the greater the variance in that direction, and the more data variability is explained.

[0079] Finally, the original data is projected onto the selected principal components to obtain the reduced-dimensional data.

[0080] Specifically, the machine learning algorithm models include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, support vector machine model, hereinafter referred to as KNN, GBDT, ET, RF and SVM respectively.

[0081] Preferably, in the training process of any prediction model (i.e., any one of the K-nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model), different hyperparameter combinations are generated for the prediction model; based on the corresponding training set (i.e., one of the training sets generated by using FS and PCA as dimensionality reduction methods), the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models;

[0082] For any initial model that has been trained, the values ​​of its various evaluation indicators are calculated respectively, and the mean of all evaluation indicators of the initial model is obtained as the evaluation indicator score of the initial model;

[0083] The initial model with the highest evaluation index score is selected as the prediction model. The evaluation index includes the area under the receiver operating characteristic curve, accuracy and F1 score.

[0084] In this specific embodiment, 10 prediction models can be obtained through 5 machine learning algorithm models and 2 dimensionality reduction methods.

[0085] Then, for any prediction model, the values ​​of each evaluation index are calculated respectively, and the mean of all evaluation indexes of the prediction model is taken as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model. The evaluation indexes include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0086] Preferably, this specific embodiment uses grid search to tune the hyperparameters of each algorithm. By presetting multiple groups of hyperparameter candidate values, the Cartesian product of different candidate combinations is generated to form a hyperparameter combination grid. Each hyperparameter combination is trained and evaluated, and the hyperparameter with the largest area under the receiver operating characteristic curve, accuracy ACC, and F-value mean is selected as the final combination.

[0087] The candidate values ​​of each hyperparameter and the final combination of the obtained model are shown in Table 2. The hyperparameters not marked in the table are the default values. Except for KNN, the other four algorithms all involve random sampling. Their random_state is set to 0, and their sampling method is fixed to reduce the impact of randomness on the algorithm. The first column is the name of each algorithm; the second column is the hyperparameter candidate value group. Parameter tuning will traverse the set combination and select the optimal model; the third and fourth columns are the optimal combination values ​​of hyperparameters after two dimensionality reduction methods.

[0088] Table 2 The final combination of candidate values ​​of each hyperparameter and the obtained prediction model

[0089]

[0090]

[0091] Preferably, this specific embodiment uses three evaluation indicators, namely AUC, ACC and F value, as the basis for judgment. By changing the boundary of the classifier to divide samples into positive examples (1) or negative examples (0), multiple groups of false positive examples and true positive values ​​are calculated, and the receiver operating characteristic curve is plotted with false positive examples as the horizontal axis and true positive examples as the vertical axis. AUC is the area enclosed by the curve and the x-axis; ACC is the ratio of the number of samples correctly predicted by the model to the total number of samples; F value is the weighted mean of precision and recall. In typhoon disasters, decision makers have a low tolerance for misjudging power outages as no power outages, so the weight is 3. Calculate the mean of AUC, ACC, and F value of each model, compare the mean of each algorithm, and the model with the largest mean is the optimal prediction combination. The calculation method for judgment is as follows:

[0092]

[0093]

[0094] Where: P AUC is AUC; N FP is a false positive example; N TP is a true example;ACC is the accuracy rate; N FN is a false counterexample; N TN is a true counterexample; P F is the F value; P R is the accuracy; P RE is the recall rate; β is the weighting coefficient, which is 3.

[0095] Example 2

[0096] like Figure 2 As shown, the present invention provides a system for predicting power outage areas in power grids under typhoon disasters, including a sample set generation model, a training set generation module, a model training module, a model selection module, and a prediction execution module; wherein,

[0097] The sample set generation model is used to generate a sample set based on the historical typhoon disaster related status data of the power grid area to be predicted; the input of the sample set is the typhoon disaster related status data of the power grid area to be predicted, and the training label is the classification conclusion of power outage or no power outage;

[0098] The training set generation module is used to process the sample sets respectively using different dimensionality reduction methods to generate corresponding training sets;

[0099] The model training module is used to train different machine learning algorithm models using different training sets to obtain several different prediction models;

[0100] The model selection module is used to determine the optimal prediction model according to the evaluation index of each prediction model;

[0101] The prediction execution module is used to input the current typhoon disaster-related status data of the power grid area to be predicted into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

[0102] Specifically, it also includes a visualization module for visualizing the prediction results of the optimal prediction model using Geographic Information System (GIS) technology.

[0103] Specifically, the typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

[0104] Specifically, the meteorological information includes the maximum wind speed, wind direction, duration of 20 m / s wind speed, duration of 30 m / s wind speed, rainfall, temperature and humidity in the forecast area.

[0105] Specifically, the geographic information includes longitude, latitude, presence or absence of users, altitude, slope direction, underlying surface type and surface type.

[0106] Specifically, the electrical equipment information includes: the number of users, the number of transformers, the number of box transformers, the number of poles and towers, the number of guy wires, the number of non-guy wires and the line length.

[0107] Specifically, during the training process of the prediction model, the model training module and the prediction execution module both pre-process the typhoon disaster-related status data of the power grid area to be predicted and use them as input of the prediction model; the pre-processing process includes: using unique hot encoding to process the character categorical variables in the typhoon disaster-related status data; and normalizing all typhoon disaster-related status data.

[0108] Specifically, the dimensionality reduction methods used by the training set generation module include feature selection and principal component analysis.

[0109] Specifically, the machine learning algorithm models used in the model training module include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model.

[0110] Specifically, the evaluation indicators used by the model selection module include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0111] Specifically, the model selection module calculates the value of each evaluation index for any prediction model, and obtains the mean value of all evaluation indexes of the prediction model as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model.

[0112] Specifically, during the training process of any prediction model, the model training module generates different hyperparameter combinations for the prediction model; based on the corresponding training set, the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models; and the initial model with the highest evaluation index score is selected as the prediction model.

[0113] Specifically, the evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

[0114] Specifically, during the training process of any prediction model, the model training module further includes: calculating the value of each evaluation index for any initial model that has been trained, and obtaining the mean value of all evaluation indexes of the initial model as the evaluation index score of the initial model.

[0115] Preferably, the sample set generation model can integrate the meteorological information, geographic information and electrical equipment information collected by the microcontroller; use a digital signal processor to perform unique hot encoding on the text data; and use an embedded system board to normalize the sorted data.

[0116] The training set generation module can use an embedded control unit to reduce the dimension of the sample set.

[0117] The model training module can use a graphics processor to establish an initial model of the power outage area of ​​the power system under typhoon disasters by using KNN, GBDT, ET, RF and SVM; and use a learning curve generation device to draw a learning curve and a grid optimization method to determine the optimal hyperparameter combination.

[0118] The model selection module can use a single-board computer to compare the established prediction models and select the optimal model; the prediction execution module uses a single-board computer to perform subsequent predictions. The visualization module uses a single-board computer to perform visualization and allows users to observe the prediction results through a monitor.

[0119] Example 3

[0120] The present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for predicting power outage areas in a power grid under a typhoon disaster described in the above technical solution is implemented.

[0121] Example 4

[0122] The present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for predicting power outage areas in power grids under typhoon disasters described in the above technical solution by executing the computer instructions.

[0123] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0124] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0125] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0127] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

[0128] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

Claims

1. A method for predicting power outage areas in a typhoon disaster, characterized in that: The following steps are involved: Generate a sample set based on historical typhoon disaster-related data in the power grid area to be predicted; Use different dimensionality reduction methods to process sample sets respectively and generate corresponding training sets; Use different training sets to train different machine learning algorithm models to obtain several different prediction models; Determine the optimal prediction model based on the evaluation indicators of each prediction model; The current typhoon disaster-related typhoon disaster phases of the power grid area to be predicted are input into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

2. The method for predicting power outage areas in a typhoon disaster according to claim 1, characterized in that: The typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

3. The method for predicting power outage areas in a typhoon disaster according to claim 2, characterized in that: The meteorological information includes the maximum wind speed, wind direction, first set wind speed duration, second set wind speed duration, rainfall, temperature and humidity in the forecast area.

4. The method for predicting power outage areas in a typhoon disaster according to claim 2, characterized in that: The geographic information includes longitude, latitude, presence or absence of users, altitude, slope direction, underlying surface type and surface type.

5. The method for predicting power outage areas in a typhoon disaster according to claim 2, characterized in that: The electrical equipment information includes: the number of users, the number of transformers, the number of box transformers, the number of poles and towers, the number of guy wires, the number of non-guy wires and the line length.

6. The method for predicting power outage areas in a typhoon disaster according to claim 1, characterized in that: During the training and use of the prediction model, the typhoon disaster-related status data of the power grid area to be predicted is preprocessed and used as the input of the prediction model; The preprocessing process includes: using one-hot encoding to process character-type categorical variables in typhoon disaster-related status data; All typhoon disaster-related status data are normalized.

7. The method for predicting power outage areas in a typhoon disaster according to claim 1, characterized in that: The dimension reduction method includes feature selection and principal component analysis.

8. The method for predicting power outage areas in a typhoon disaster according to claim 1, characterized in that: The machine learning algorithm models include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model.

9. The method for predicting power outage areas in a typhoon disaster according to claim 1, characterized in that: The evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

10. The method for predicting power outage areas in a typhoon disaster according to claim 9, characterized in that: For any prediction model, the value of each evaluation index is calculated respectively, and the mean value of all evaluation indexes of the prediction model is taken as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model.

11. The method for predicting power outage areas in a typhoon disaster according to claim 9, characterized in that: During the training process of any prediction model, different hyperparameter combinations are generated for the prediction model; based on the corresponding training set, the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models; and the initial model with the highest evaluation index score is selected as the prediction model.

12. The method for predicting power outage areas in a typhoon disaster according to claim 11, characterized in that: The evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

13. The method for predicting power outage areas in a typhoon disaster according to claim 12, characterized in that: The training process of any prediction model also includes: calculating the value of each evaluation index for any initial model that has been trained, and obtaining the average value of all evaluation indexes of the initial model as the evaluation index score of the initial model.

14. A system for predicting power outage areas in power grids under typhoon disasters, characterized in that: It includes a sample set generation model, a training set generation module, a model training module, a model selection module, and a prediction execution module; among them, The sample set generation model is used to generate a sample set based on the historical typhoon disaster related status data of the power grid area to be predicted; The training set generation module is used to process the sample sets respectively using different dimensionality reduction methods to generate corresponding training sets; The model training module is used to train different machine learning algorithm models using different training sets to obtain several different prediction models; The model selection module is used to determine the optimal prediction model according to the evaluation index of each prediction model; The prediction execution module is used to input the current typhoon disaster-related status data of the power grid area to be predicted into the optimal prediction model to obtain the prediction result of whether the power grid area to be predicted will have a power outage.

15. The system for predicting power outage areas in power grids under typhoon disasters according to claim 14, characterized in that: The typhoon disaster related status data of the power grid area to be predicted includes: meteorological information, geographical information and electrical equipment information of the power grid area to be predicted.

16. The system for predicting power outage areas in power grids under typhoon disasters according to claim 15, characterized in that: The meteorological information includes the maximum wind speed, wind direction, first set wind speed duration, second set wind speed duration, rainfall, temperature and humidity in the forecast area.

17. The system for predicting power outage areas in power grids under typhoon disasters according to claim 15, characterized in that: The geographic information includes longitude, latitude, presence or absence of users, altitude, slope direction, underlying surface type and surface type.

18. The system for predicting power outage areas in power grids under typhoon disasters according to claim 15, characterized in that: The electrical equipment information includes: the number of users, the number of transformers, the number of box transformers, the number of poles and towers, the number of guy wires, the number of non-guy wires and the line length.

19. The method for predicting power outage areas in a typhoon disaster according to claim 14, characterized in that: During the training of the prediction model, the model training module and the prediction execution module both pre-process the typhoon disaster-related status data of the power grid area to be predicted and use them as the input of the prediction model. The preprocessing process includes: using one-hot encoding to process character-type categorical variables in typhoon disaster-related status data; All typhoon disaster-related status data are normalized.

20. The method for predicting power outage areas in a typhoon disaster according to claim 14, characterized in that: The dimensionality reduction methods used in the training set generation module include feature selection and principal component analysis.

21. The method for predicting power outage areas in a typhoon disaster according to claim 14, characterized in that: The machine learning algorithm models used in the model training module include: K nearest neighbor algorithm model, gradient boosting decision tree model, extreme random tree model, random forest model, and support vector machine model.

22. The method for predicting power outage areas in a typhoon disaster according to claim 14, characterized in that: The evaluation indicators used in the model selection module include the area under the receiver operating characteristic curve, accuracy and F1 score.

23. A method for predicting power outage areas in a power grid under a typhoon disaster according to claim 22, characterized in that: The model selection module calculates the value of each evaluation index for any prediction model, and obtains the mean value of all evaluation indexes of the prediction model as the score of the prediction model; the prediction model with the highest score is taken as the optimal prediction model.

24. The method for predicting power outage areas in a typhoon disaster according to claim 22, characterized in that: During the training process of any prediction model, the model training module generates different hyperparameter combinations for the prediction model; based on the corresponding training set, the machine learning algorithm model under each hyperparameter combination is trained to obtain different initial models; and the initial model with the highest evaluation index score is selected as the prediction model.

25. A method for predicting power outage areas in a power grid under a typhoon disaster according to claim 24, characterized in that: The evaluation indicators include the area under the receiver operating characteristic curve, accuracy and F1 score.

26. A method for predicting power outage areas in a typhoon disaster according to claim 25, characterized in that: During the training process of any prediction model, the model training module also includes: calculating the value of each evaluation index for any initial model that has been trained, and obtaining the mean value of all evaluation indexes of the initial model as the evaluation index score of the initial model.

27. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting power outage areas in a power grid under a typhoon disaster described in any one of claims 1 to 10 is implemented.

28. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for predicting power outage areas in power grids under typhoon disasters as described in any one of claims 1 to 10 by executing the computer instructions.