Distribution network line fault tripping prediction method, equipment, medium and product
By constructing a fusion prediction method based on the Akaike information content criterion and the limit gradient improvement model, the problem of low prediction accuracy of distribution network line fault tripping was solved, and efficient prediction of the monthly fault tripping frequency was achieved, thereby improving power supply reliability.
Patent Information
- Application Number
- CN202511815490.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-04
- Publication Date
- 2026-03-20
AI Technical Summary
Existing methods for predicting fault tripping in distribution network lines are unable to effectively handle multi-dimensional nonlinear influencing factors, resulting in low prediction accuracy and an inability to provide timely and accurate fault warnings, thus affecting power supply reliability.
By constructing a tripping prediction model and a limit gradient boosting model based on the Akaike information content criterion, and combining them with the grid search method, a fusion model is constructed for weighted fusion. The optimal prediction result is selected, and the final prediction model is used to predict the monthly number of fault trips, which is updated regularly.
It enables efficient prediction of the monthly number of fault trips in distribution network lines, improving the reliability and early warning accuracy of distribution network operation and maintenance.
Smart Images
Figure CN121705944A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of power distribution network operation and maintenance, and in particular to a power distribution network line fault tripping prediction method, device, medium and product. BACKGROUND
[0002] As the end of the power system, the power distribution network is directly related to user power consumption. Line fault tripping will seriously affect power supply continuity. However, the power distribution network line is complex and covers a wide area. It is easily disturbed by climate factors such as thunderstorms and strong winds, as well as internal factors such as equipment aging and load fluctuations during operation. Fault tripping shows significant seasonal periodicity and random volatility, and traditional prediction methods relying on manual inspection or single device fault models are difficult to adapt.
[0003] In existing prediction methods, a single time series model such as the Autoregressive Integrated Moving Average (ARIMA) model can capture the timing rules, but it cannot effectively handle multi-dimensional nonlinear influencing factors such as climate and equipment status. Machine learning models such as Extreme Gradient Boosting (XGBoost) are good at mining nonlinear features, but they lack long-term timing trends in fault tripping. Moreover, most methods do not screen for correlations between influencing factors, which can easily include low-correlation redundant variables. Linear interpolation is used to fill in missing values, which can introduce bias in uneven data distribution scenarios, further reducing prediction accuracy.
[0004] With the improvement of power distribution network automation, a large amount of operation data has been accumulated. However, existing technologies do not fully integrate the timing characteristics and nonlinear features of multi-source data, nor do they optimize data preprocessing and variable screening processes for monthly prediction scenarios. This results in low data value utilization and inability to provide timely and accurate fault warnings for power distribution network operation and maintenance. Therefore, based on the above problems, it is necessary to provide a power distribution network line fault tripping prediction method to efficiently predict the monthly fault tripping number of the power distribution network line and improve the reliability of power distribution network operation and maintenance. SUMMARY
[0005] The purpose of the present application is to provide a power distribution network line fault tripping prediction method, device, medium and product that can efficiently predict the monthly fault tripping number of the power distribution network line and improve the reliability of power distribution network operation and maintenance.
[0006] To achieve the above purpose, the present application provides the following solutions: In a first aspect, the present application provides a power distribution network line fault tripping prediction method, comprising: Obtaining fault data of a distribution network line; and constructing a fault data set according to the fault data; the fault data including: historical fault tripping times and influence factor data; According to the fault data set, constructing a tripping prediction model based on the Akaike information criterion; and determining exogenous variables corresponding to the tripping prediction model; the tripping prediction model is used to predict monthly fault tripping times; the exogenous variables are selected from the influence factor data; According to the exogenous variables and the historical fault tripping times, constructing a limit gradient boosting model based on a grid search method; the limit gradient boosting model is used to predict monthly fault tripping times; According to the tripping prediction model and the limit gradient boosting model, constructing a fusion model; the fusion model is used to weight and fuse the monthly fault tripping times predicted by the tripping prediction model and the monthly fault tripping times predicted by the limit gradient boosting model to obtain a fusion monthly fault tripping time; Respectively determining prediction errors of the tripping prediction model and the fusion model, and determining a final prediction model according to the prediction errors; the prediction errors include: mean absolute error, root mean square error, and symmetric average absolute percentage error; Using the final prediction model to determine fault tripping times of a to-be-predicted month; and periodically obtaining fault data of the distribution network line to periodically update the final prediction model.
[0007] Optionally, the influence factor data includes: lightning times, lightning current amplitude, lightning number, positive lightning number, negative lightning number, average current, positive average current, negative average current, back stroke number, positive average back stroke, negative average back stroke, lightning hour, lightning day, ground flash density, monthly average temperature, monthly average humidity, monthly average precipitation, and monthly average wind speed.
[0008] Optionally, according to the fault data set, the tripping prediction model is constructed based on the Akaike information criterion; specifically including: Determining a Pearson correlation coefficient between the historical fault tripping times and the influence factor data; Determining a candidate exogenous variable according to the Pearson correlation coefficient; According to the candidate exogenous variable, constructing multiple groups of initial tripping prediction models based on the Akaike information criterion; Determining a prediction error of each initial tripping prediction model; and taking the initial tripping prediction model with the minimum prediction error as the tripping prediction model.
[0009] Optionally, the Pearson correlation coefficient between the historical fault tripping times and the influence factor data is determined, specifically including: Using the formula determining the Pearson correlation coefficient between the historical fault tripping times and the influence factor data ; wherein, is the total number of periods, is the period number, is the influence factor data of the period, is the mean value of the influence factor data, is the actual value of the historical fault tripping number of the period, is the mean value of the historical fault tripping number of the period.
[0010] Optionally, the determining of the prediction error of each initial tripping prediction model specifically comprises: determining the prediction error of each initial tripping prediction model by using formula , formula and formula ; wherein, is the mean absolute error, is the root mean square error, is the symmetric average absolute percentage error, is the total number of periods, is the period number, is the actual value of the historical fault tripping number of the period, is the prediction value of the initial tripping prediction model of the period.
[0011] Optionally, the constructing of the fusion model according to the tripping prediction model and the limit gradient boosting model specifically comprises: determining the fusion model by using formula ; wherein, is the fusion monthly fault tripping number, is the weight coefficient of the tripping prediction model, is the prediction value obtained by the tripping prediction model, is the weight coefficient of the limit gradient boosting model, is the prediction value obtained by the limit gradient boosting model.
[0012] Optionally, the determining of the prediction error of the tripping prediction model and the fusion model respectively and the determining of the final prediction model according to the prediction error specifically comprises: if the prediction errors of the fusion model are all smaller than the prediction error of the tripping prediction model, the fusion model is taken as the final prediction model; otherwise, the tripping prediction model is taken as the final prediction model.
[0013] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described distribution network line fault tripping prediction method.
[0014] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described distribution network line fault tripping prediction method.
[0015] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described distribution network line fault tripping prediction method.
[0016] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method, device, medium, and product for predicting fault tripping in distribution network lines. By focusing specifically on fault data of distribution network lines, a tripping prediction model and a Extreme Gradient Boosting (XGBoost) model are constructed sequentially. The fusion model is further determined, and the prediction error is compared to select the optimal prediction result. This application solves the problems of redundancy of exogenous variables and low prediction accuracy in traditional models, and achieves efficient prediction of the monthly fault tripping frequency of distribution network lines. It can be widely used in fault early warning by distribution network management departments at all levels, providing accurate basis for the operation and maintenance of distribution network lines and improving the reliability of power supply. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart illustrating a method for predicting fault tripping in a distribution network line according to an embodiment of this application. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0021] In one exemplary embodiment, such as Figure 1 As shown, a method for predicting fault tripping in distribution network lines is provided, including the following steps S1 to S6. Wherein: S1: Obtain fault data of distribution network lines; and construct a fault dataset based on the fault data.
[0022] The fault data includes historical fault trip counts and influencing factor data. Specifically, this application obtains at least three years of historical fault trip counts and multi-dimensional influencing factor data for distribution network lines, and divides them into monthly data.
[0023] The influencing factors data include, but are not limited to: lightning strike frequency, lightning current amplitude, number of lightning strikes, number of positive lightning strikes, number of negative lightning strikes, average current, positive average current, negative average current, number of return strokes, positive average return stroke, negative average return stroke, lightning hour, lightning day, ground flash density, monthly average temperature, monthly average humidity, monthly average precipitation, and monthly average wind speed. Historical fault trip counts are obtained from actual distribution network fault data or distribution network automation systems. Lightning-related indicators in the influencing factors data are obtained from the meteorological department's lightning monitoring database, and climate-related indicators are obtained from the meteorological department's publicly available climate database.
[0024] After obtaining the fault data of the distribution network lines, the missing data in the fault data is filled in using the mean imputation method (calculating the arithmetic mean of the non-missing data of each indicator and replacing the missing values of that indicator; the indicators are the historical fault trip count and influencing factor data). Guidelines (data exceeds) The range was determined to be abnormal. The mean of the data. Outliers in the fault data are removed (using the standard deviation as a factor); finally, the fault data is normalized to 0-1, and a fault dataset is constructed based on the normalized fault data. The normalization formula is as follows: ; in, The fault data is normalized. For fault data, The minimum value in the fault data. This represents the maximum value in the fault data.
[0025] The fault dataset is divided into a validation set and a test set. The validation set includes fault data from year 1 to year N-1, and the test set is fault data from year N, where N is the total number of fault data for each year.
[0026] S2: Based on the fault dataset, construct a tripping prediction model based on the Akaike information content criterion; and determine the exogenous variables corresponding to the tripping prediction model.
[0027] S2 specifically includes: S21: Determine the Pearson correlation coefficient between the number of historical fault trips and the data of influencing factors.
[0028] The formula for calculating the Pearson correlation coefficient is as follows: ; in, The Pearson correlation coefficient is the relationship between the number of historical fault trips and the data on influencing factors. This represents the total number of statistical periods. For periodic numbering, For the first Data on factors influencing the cycle The mean of the influencing factor data, For the first Actual number of historical fault trips during the cycle. This represents the average number of trips due to historical faults.
[0029] The period refers to the statistical sampling granularity or observation time unit, which can be selected from year, quarter, month, week, and day according to the actual situation. In this application, the period refers to month.
[0030] S22: Determine candidate exogenous variables based on Pearson correlation coefficients.
[0031] Specifically, data on influencing factors with a Pearson correlation coefficient greater than or equal to 0.6 were selected as candidate exogenous variables and sorted from largest to smallest according to the absolute value of the Pearson correlation coefficient; data on influencing factors with a Pearson correlation coefficient less than 0.6 were not included.
[0032] S23: Based on the candidate exogenous variables, construct multiple initial trip prediction models according to the Akaike information content criterion.
[0033] Let the total number of candidate exogenous variables in this application be n. Starting from a=1 (where a is the number of selected candidate exogenous variables), the number is incremented sequentially to a=n. The first a candidate exogenous variables (a≤n) are selected sequentially. The selected candidate exogenous variables and time series factors (assigned values 1-12 for January-December respectively) are used as inputs, and the monthly historical fault trip count is used as output. The order (p, d, q) of the initial trip prediction model is determined by the AIC criterion (AIC=2k-2ln(L), where L is the likelihood function value and k is the total number of parameters in the initial trip prediction model), where p is the AR order, q is the MA order, and d is the difference order. In this application, multiple initial trip prediction models are constructed, specifically using the ARIMAX model as the initial trip prediction model.
[0034] S24: Determine the prediction error of each initial trip prediction model; and take the initial trip prediction model with the smallest prediction error as the trip prediction model.
[0035] Specifically, the prediction error of each initial trip prediction model is determined. The prediction error includes Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Symmetric Mean Absolute Percentage Error (sMAPE), calculated as follows: ; ; ; in, This represents the total number of statistical periods. For periodic numbering, For the first Actual number of historical fault trips during the cycle. For the first The predicted value of the initial trip prediction model for the cycle.
[0036] The initial tripping prediction model with the smallest prediction error was selected as the tripping prediction model, namely the ARIMAX_best model. The output of the tripping prediction model is the monthly number of fault trips on the test set, which is the predicted value of the tripping prediction model. The candidate exogenous variables corresponding to the trip prediction model are used as exogenous variables.
[0037] S3: Based on exogenous variables and the number of historical fault trips, construct a limit gradient boosting model using the grid search method.
[0038] Specifically, the exogenous variables corresponding to the tripping prediction model and the previous month's historical fault tripping count (the previous month's historical fault tripping count) are used as input features, with the monthly historical fault tripping count as the label. The model iterates through hyperparameter combinations of learning rate (0.05-0.3), tree depth (3-10), and number of leaf nodes (10-100), selecting the combination with the highest prediction accuracy on the validation set as the final hyperparameters. The model is trained using the validation set, and its nonlinear prediction of the monthly fault tripping count on the test set is output. .
[0039] S4: Construct a fusion model based on the trip prediction model and the limit gradient boosting model.
[0040] The fusion model is used to weight and fuse the monthly fault trip count predicted by the trip prediction model with the monthly fault trip count predicted by the limit gradient boosting model to obtain the fused monthly fault trip count.
[0041] Specifically, the RMSE of the trip prediction model and the limit gradient boosting model are determined separately, and the weight coefficients of the trip prediction model and the limit gradient boosting model are further determined (the smaller the error, the larger the weight coefficient). The calculation formula is as follows: ; ; in, These are the weighting coefficients for the trip prediction model. The weights of the limiting gradient boosting model are used to determine the weights. For the trip prediction model's RMSE, This is the RMSE of the extreme gradient boosting model.
[0042] A fusion model is constructed based on the weight coefficients of the trip prediction model and the extreme gradient boosting model. The calculation formula is as follows: ; in, To integrate the number of monthly fault trips, The predicted value is obtained from the trip prediction model. This is the predicted value obtained from the limiting gradient boosting model.
[0043] S5: Determine the prediction errors of the trip prediction model and the fusion model respectively, and determine the final prediction model based on the prediction errors.
[0044] Specifically, the MAE, RMSE, and sMAPE values of the trip prediction model and the fusion model are calculated on the test set. If the prediction error of the fusion model is smaller than that of the trip prediction model, the fusion model is used as the final prediction model, and the monthly fault trip count obtained by the fusion model is used as the final prediction result. Conversely, if the prediction error of the fusion model is smaller, the trip prediction model is used as the final prediction model, and the monthly fault trip count obtained by the trip prediction model is used as the final prediction result. That is, if the error of the fusion model is smaller, it is selected. The final monthly forecast result is used as the basis for prediction; otherwise, the forecast result of the ARIMAX_best model is retained.
[0045] S6: Use the final prediction model to determine the number of fault trips in the month to be predicted; and regularly obtain fault data of distribution network lines to update the final prediction model periodically.
[0046] Specifically, new and complete fault data (including the number of fault trips and multi-dimensional influencing factors data, divided into monthly data) are collected every 12 months. The above steps are repeated to update the candidate exogenous variables, the trip prediction model, the limit gradient improvement model, and the final prediction model to ensure the prediction accuracy of the monthly fault trip count.
[0047] In an exemplary embodiment, 90 10kV distribution network lines under the jurisdiction of a county-level distribution network management department are taken as the research object. The specific steps of the network line fault tripping prediction method are as follows: S1: Obtain fault data of distribution network lines.
[0048] Historical fault trip counts for 5 years and 60 months from January 2019 to December 2023 were obtained (extracted from actual distribution network line fault trip records or distribution network dispatch automation systems). At the same time, multi-dimensional influencing factor data for the region during the same period were collected: lightning-related indicators (number of lightning strikes, lightning current amplitude, number of lightning strikes, number of positive lightning strikes, number of negative lightning strikes, average current, positive average current, negative average current, number of return strokes, positive average return strokes, negative average return strokes, lightning hours, lightning days, and ground flash density) were obtained from the lightning monitoring database of the local meteorological department, and climate-related indicators (monthly average temperature, monthly average humidity, monthly average precipitation, and monthly average wind speed) were obtained from the publicly available climate database of the meteorological department.
[0049] Two missing values were found in the monthly average humidity data. The arithmetic mean of the 58 months of data with no missing values was calculated, and this mean was used to fill in the missing values. All data were normalized to 0-1. For example, the original value of the lightning density for a certain month was 8 lightning strikes / (km²). 2 (year), the minimum value of this indicator is 2 times / (km) 2 (year), maximum value is 15 times / (km) 2 (year), then the normalized value is The fault dataset was divided according to the rule that "the validation set consists of complete monthly data from year 1 to year N-1, and the test set consists of complete monthly data from the last year": the validation set consists of 48 months of data from January 2019 to December 2022, and the test set consists of 12 months of data from January to December 2023.
[0050] S2: Based on the fault dataset, construct a tripping prediction model based on the Akaike information content criterion.
[0051] Pearson correlation coefficients were calculated between 18 influencing factors and the number of historical fault trips. Five candidate exogenous variables with Pearson correlation coefficients ≥ 0.6 were selected and sorted from largest to smallest by absolute value of Pearson correlation coefficient: lightning density (0.85), number of lightning strikes (0.82), lightning current amplitude (0.76), monthly average precipitation (0.71), and lightning days (0.63). Indicators with Pearson correlation coefficients < 0.6, such as the number of positive lightning strikes and negative average current, were not included in the candidate exogenous variables.
[0052] Starting with a=1, the first a candidate exogenous variables are selected sequentially to construct the initial trip prediction model, i.e., the ARIMAX model, and the order of each ARIMAX model is determined by the AIC criterion: a=1 (exogenous variable: ground flash density): ARIMAX model order is ARIMAX(1, 0, 1), AIC value is 305.2; a=2 (exogenous variables: lightning density, number of lightning strikes): The ARIMAX model order is ARIMAX(1, 0, 2), and the AIC value is 299.7; a=3 (exogenous variables: lightning density, number of lightning strikes, lightning current amplitude): ARIMAX model order is ARIMAX(2, 0, 1), AIC value is 294.5; a=4 (exogenous variables: lightning density, number of lightning strikes, lightning current amplitude, monthly average precipitation): The ARIMAX model order is ARIMAX(2, 0, 2), and the AIC value is 291.3; a=5 (exogenous variables: lightning density, number of lightning strikes, lightning current amplitude, monthly average precipitation, number of lightning days): ARIMAX model order is ARIMAX(2, 0, 2), AIC value is 292.8.
[0053] Calculate the MAE, RMSE, and sMAPE values of each ARIMAX model on the validation set: When a=1: MAE=2.35, RMSE=2.98, sMAPE=10.5%; When a=2: MAE=2.01, RMSE=2.62, sMAPE=9.1%; When a=3: MAE=1.85, RMSE=2.41, sMAPE=8.3%; When a=4: MAE=1.78, RMSE=2.25, sMAPE=7.9%; When a=5: MAE=1.82, RMSE=2.33, sMAPE=8.1%.
[0054] The ARIMAX model with a=4 exhibits the smallest errors in all three terms; therefore, it is selected as the tripping prediction model, i.e., the ARIMAX_best model. The ARIMAX_best model has an order of ARIMAX(2, 0, 2), and its exogenous variables are ground flash density, lightning strike frequency, lightning current amplitude, and monthly average precipitation. The predicted monthly fault tripping frequency of the tripping prediction model on the test set is... The range is 7-34 times.
[0055] S3: Based on exogenous variables and the number of historical fault trips, construct a limit gradient boosting model using the grid search method.
[0056] Using four exogenous variables (flash density, lightning strike count, lightning current amplitude, and monthly average precipitation) selected for the tripping prediction model, along with the previous month's historical fault trip count (the previous month's historical fault trip count) as input features, and the monthly historical fault trip count as the label, a bounded gradient boosting model is constructed. A grid search method is used to traverse hyperparameter combinations with learning rates (0.05, 0.1, 0.2), tree depths (4, 6, 8), and leaf node counts (20, 50, 80) to calculate the monthly fault trip count prediction accuracy of each combination of bounded gradient boosting models on the validation set. Learning rate = 0.05, tree depth = 4, number of leaf nodes = 20: accuracy 86.3%; Learning rate = 0.1, tree depth = 6, number of leaf nodes = 50: accuracy 91.2%; Learning rate = 0.2, tree depth = 8, number of leaf nodes = 80: accuracy 88.7%.
[0057] The parameters with the highest accuracy (learning rate = 0.1, tree depth = 6, number of leaf nodes = 50) were selected as the final hyperparameters. A bounded gradient boosting model was trained on the validation set. The monthly fault trip count predicted by the bounded gradient boosting model on the test set was then used. The range is 6-32, RMSE=1.85.
[0058] S4: Construct a fusion model based on the trip prediction model and the limit gradient boosting model.
[0059] The corresponding weight coefficients are determined based on the RMSE of the trip prediction model and the extreme gradient boosting model on the validation set. , : ; .
[0060] The formula for calculating the monthly fault trip count, i.e., the predicted value of the fusion model, is as follows: .
[0061] S5: Determine the prediction errors of the trip prediction model and the fusion model respectively, and determine the final prediction model based on the prediction errors.
[0062] Calculate the prediction error between the trip prediction model and the fusion model on the test set: Trip prediction model: MAE=2.08, RMSE=2.59, sMAPE=9.2%; Fusion model: MAE=1.65, RMSE=2.03, sMAPE=7.5%.
[0063] Since the prediction errors of the fusion model are all smaller than those of the trip prediction model, we choose... This serves as the prediction result for the final monthly number of outage trips. For example, in July 2023 (the peak season for thunderstorms), the predicted value from the trip prediction model is... Next, the predicted values of the limiting gradient boosting model. Next, the predicted values of the fusion model The actual number of fault trips in that month was 30, with a relative error of only 0.3%, and the prediction accuracy met the engineering requirements.
[0064] S6: Use the final prediction model to determine the number of fault trips in the month to be predicted; and regularly obtain fault data of distribution network lines to update the final prediction model periodically.
[0065] In December 2024, new data from January to December 2024 (including the number of fault trips and multi-dimensional influencing factors, divided by month) will be collected, and S1-S5 will be repeated to update the final prediction model: Update of candidate exogenous variables: After adding data, there are still 5 candidate exogenous variables. The order of Pearson correlation coefficients has not changed. The exogenous variables are still lightning density, number of lightning strikes, lightning current amplitude, and monthly average precipitation. The trip prediction model was updated: the order was changed to ARIMAX(2, 0, 1), and the validation set sMAPE=8.0%; Update the extreme gradient boosting model: fine-tuned hyperparameters to "learning rate = 0.12, tree depth = 6, number of leaf nodes = 50", validation set sMAPE = 6.5%; Final prediction model update: The fusion model test set sMAPE=7.3%, which is still better than the trip prediction model (sMAPE=8.0%). Therefore, the fusion model is retained as the prediction model for the number of monthly fault trips in 2025.
[0066] When periodically updating the final prediction model, the embodiment only calculates the sMAPE of the trip prediction model, the extreme gradient boosting model, and the fusion model to determine the final prediction model. In practical applications, the MAE or RMSE of the trip prediction model, the extreme gradient boosting model, and the fusion model can also be calculated separately and compared; or the MAE, RMSE, and sMAPE of the trip prediction model, the extreme gradient boosting model, and the fusion model can be calculated simultaneously and compared comprehensively. That is, this application can adaptively select the prediction error according to the actual situation.
[0067] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores distribution network line fault tripping prediction data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a distribution network line fault tripping prediction method.
[0068] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0069] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0070] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0071] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0073] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0074] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0075] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for predicting fault tripping in distribution network lines, characterized in that, The method for predicting fault tripping in distribution network lines includes: Obtain fault data of distribution network lines; and construct a fault dataset based on the fault data; the fault data includes: historical fault trip counts and influencing factor data; Based on the fault dataset, a tripping prediction model is constructed using the Akaike information content criterion; and the exogenous variables corresponding to the tripping prediction model are determined; the tripping prediction model is used to predict the monthly number of fault trips; the exogenous variables are selected from the influencing factor data; Based on the exogenous variables and the historical fault trip count, a limit gradient boosting model is constructed using the grid search method; the limit gradient boosting model is used to predict the monthly fault trip count. A fusion model is constructed based on the trip prediction model and the limit gradient boosting model. The fusion model is used to weight and fuse the monthly fault trip count predicted by the trip prediction model and the monthly fault trip count predicted by the limit gradient boosting model to obtain the fused monthly fault trip count. The prediction errors of the trip prediction model and the fusion model are determined separately, and the final prediction model is determined based on the prediction errors; the prediction errors include: mean absolute error, root mean square error and symmetric mean absolute percentage error. The final prediction model is used to determine the number of fault trips in the month to be predicted; and the fault data of the distribution network lines are obtained periodically to update the final prediction model.
2. The method for predicting fault tripping in distribution network lines according to claim 1, characterized in that, The influencing factors data include: number of lightning strikes, lightning current amplitude, number of lightning strikes, number of positive lightning strikes, number of negative lightning strikes, average current, positive average current, negative average current, number of return strokes, positive average return stroke, negative average return stroke, lightning hour, lightning day, ground flash density, monthly average temperature, monthly average humidity, monthly average precipitation, and monthly average wind speed.
3. The method for predicting fault tripping in distribution network lines according to claim 1, characterized in that, The step of constructing a tripping prediction model based on the fault dataset and the Akaike information content criterion specifically includes: Determine the Pearson correlation coefficient between the number of historical fault trips and the data on influencing factors; Candidate exogenous variables are determined based on the Pearson correlation coefficient; Based on the candidate exogenous variables, multiple initial trip prediction models are constructed using the Akaike information content criterion. Determine the prediction error of each initial trip prediction model; and select the initial trip prediction model with the smallest prediction error as the trip prediction model.
4. The method for predicting fault tripping in distribution network lines according to claim 3, characterized in that, The determination of the Pearson correlation coefficient between the number of historical fault trips and the influencing factor data specifically includes: Using formula Determine the Pearson correlation coefficient between the number of historical fault trips and influencing factor data. ; in, This represents the total number of statistical periods. For periodic numbering, For the first Data on factors influencing the cycle The mean of the influencing factor data, For the first Actual number of historical fault trips during the cycle. The average number of historical fault trips over a given period.
5. The method for predicting fault tripping in distribution network lines according to claim 3, characterized in that, The determination of the prediction error for each initial trip prediction model specifically includes: Using formula ,formula and formula Determine the prediction error of each initial trip prediction model; in, The mean absolute error, The root mean square error, For symmetrical mean absolute percentage error, This represents the total number of statistical periods. For periodic numbering, For the first Actual number of historical fault trips during the cycle. For the first The predicted value of the initial trip prediction model for the cycle.
6. The method for predicting fault tripping in distribution network lines according to claim 1, characterized in that, The construction of the fusion model based on the tripping prediction model and the limit gradient improvement model specifically includes: Using formula Determine the fusion model; in, To integrate the number of monthly fault trips, These are the weighting coefficients for the trip prediction model. The predicted value is obtained from the trip prediction model. The weights of the limiting gradient boosting model are used to determine the weights. This is the predicted value obtained from the limiting gradient boosting model.
7. The method for predicting fault tripping in distribution network lines according to claim 1, characterized in that, The process of determining the prediction errors of the tripping prediction model and the fusion model respectively, and determining the final prediction model based on the prediction errors, specifically includes: If the prediction error of the fusion model is smaller than that of the trip prediction model, the fusion model will be used as the final prediction model; otherwise, the trip prediction model will be used as the final prediction model.
8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the distribution network line fault tripping prediction method according to any one of claims 1-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the distribution network line fault tripping prediction method as described in any one of claims 1-7.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the distribution network line fault tripping prediction method as described in any one of claims 1-7.