A power transmission line fault prediction system based on multi-algorithm fusion

Abstract

A transmission line fault prediction system based on multi-algorithm fusion includes a data acquisition module, an entropy weight-TOPSIS evaluation module, an ARIMA prediction module, a dynamic weight adjustment module, a closed-loop feedback optimization module, a risk level classification module, and a decision support and visualization platform. The system standardizes multi-source data using the entropy weight-TOPSIS evaluation model, calculates the information entropy and weight of each indicator, and derives a line health status score. The ARIMA prediction module predicts the fault probability based on the evaluation results and time-series data. The dynamic weight adjustment module dynamically adjusts the ARIMA model parameters according to the health status score. The closed-loop feedback optimization module compares the prediction results with actual fault data and continuously optimizes the model parameters. The risk level classification module outputs four-level risk warnings based on the fault probability and health status score. The decision support and visualization platform generates inspection suggestions, maintenance priority ranking, and route optimization schemes, and supports GIS map display.

Description

Technical Field

[0001] This invention belongs to the field of power system safe operation and intelligent assessment technology, specifically relating to a transmission line fault prediction system based on multi-algorithm fusion. Background Technology

[0002] As a crucial component of the power system, the safe and stable operation of transmission lines directly impacts the overall reliability of the power grid. However, transmission lines are exposed to the natural environment for extended periods, making them susceptible to lightning strikes, icing, and external damage, leading to frequent faults.

[0003] Existing technologies typically obtain fault information through power transmission equipment. For example, Chinese patent CN120831536A discloses a method, device, and medium for fault diagnosis of power transmission lines based on edge computing. In this invention, faulty equipment is located by acquiring image information of the equipment. However, such analysis methods rely on image edge computing for fault diagnosis, lacking predictive capabilities and only able to identify faults that have already occurred. The data source is singular, failing to integrate multi-source data such as meteorological and equipment status data. Furthermore, there is a lack of a comprehensive evaluation mechanism for the health status of the line and analysis of its temporal characteristics, making it impossible to provide early warnings and hindering the implementation of a prevention-oriented operation and maintenance strategy.

[0004] Therefore, there is an urgent need for a transmission line fault prediction system and method based on multi-algorithm fusion to solve the problem of difficulty in predicting and analyzing transmission line faults. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a transmission line fault prediction system based on multi-algorithm fusion. Through a multi-level architecture of data acquisition → comprehensive evaluation → dynamic prediction → risk classification → decision support, it achieves a prevention-oriented operation and maintenance strategy.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A transmission line fault prediction system based on multi-algorithm fusion includes a data acquisition module, an entropy weight-Topsis evaluation module, an ARIMA prediction module, a dynamic weight adjustment module, a closed-loop feedback optimization module, a risk level classification module, and a decision support and visualization platform. The data acquisition module is used to collect multi-source data related to the operation of transmission lines; The Entropy Weight-Topsis evaluation module standardizes the collected multi-source data, calculates the information entropy of each evaluation indicator, determines the indicator weight, constructs a weighted decision matrix, and calculates the distance between each transmission line segment and the optimal and worst solutions to obtain the line health status score. The ARIMA prediction module uses entropy weight-Topsis evaluation results and time series data to construct a differential integrated moving average autoregressive model to predict the probability of transmission line faults in future periods. The dynamic weight adjustment module dynamically adjusts the input parameters and weight coefficients of the ARIMA model based on the line health status score; The closed-loop feedback optimization module compares the prediction results with the actual fault data to optimize the parameters of the entropy weight-Topsis evaluation model and the ARIMA prediction model. The risk level classification module outputs the risk level of the transmission line based on the predicted failure probability and the evaluation of the health status score. The decision support and visualization platform is used to generate inspection suggestions, maintenance priority ranking, and route optimization schemes, and supports GIS map display.

[0007] The entropy weight-Topsis evaluation module uses the range standardization method to standardize the collected multi-source data; (1); In equation (1), The standardized index value, These are the original index values. and These are the minimum and maximum values ​​of the indicator, respectively. The Entropy Weight-Topsis evaluation module calculates the information entropy of each indicator. This reflects the information uncertainty of the indicator, and the specific formula is as follows: (2); In equation (2), , represents the normalization coefficient for entropy calculation. , representing the characteristic weight of the i-th evaluation object on the j-th indicator. n The number of objects to be evaluated; Entropy Weight - The Topsis evaluation module calculates the weight of indicators based on entropy values. The greater the difference, the higher the weight. Calculate the first... j The weights of each indicator are calculated using the following formula: (3); In equation (3), For the first j The weight of each evaluation indicator; For the first j The entropy value of the evaluation index; m The total number of evaluation indicators; The Entropy Weight-Topsis evaluation module calculates the Euclidean distance between each transmission line segment and the positive and negative ideal solutions, and derives a health status score: (4); In equation (4), For the first iThe Euclidean distance between each line segment and the ideal solution. For the first i The Euclidean distance between each line segment and the negative ideal solution.

[0008] The ARIMA prediction module uses the ARIMA algorithm for prediction. It eliminates trends through differencing, captures historical dependencies through autoregression, and smooths random fluctuations through moving averages, thereby achieving accurate prediction of future failure probabilities. The ARIMA prediction module uses the following model expression: (5); In equation (5), For the shift operator; These are the autoregressive coefficients; The autoregressive order indicates how many past time points the model uses for prediction; This is the differential order; the dynamic weight adjustment module adjusts this parameter based on the line health status score. The value observed at time t; q The moving average order; The moving average coefficient; Let be a white noise sequence at time t, which follows a normal distribution. (0, ); Based on the time-series fitting and prediction results of the ARIMA model above, the predicted fault probability P of the transmission line within the next T time period is output, and the expression is: (6); In equation (6), For the output of the ARIMA model, the future Predicted values ​​of the line health status at any given time; This refers to the critical health state threshold for line faults. For the future The probability that the line health status is lower than the fault threshold during the time period, with a value range of [0,1].

[0009] The dynamic weight adjustment module adjusts the weight based on the line health status score. Dynamically adjust the difference order of the ARIMA model d Weights of input features: when When the value is greater than 0.7, setting d=0 makes the model focus more on predicting stationary sequences. When 0.4 < When ≤0.7, set d=1, the model focuses on predicting slight trends; when When ≤0.4, set d=2, the model focuses on predicting drastic fluctuations and abrupt changes.

[0010] The closed-loop feedback optimization module dynamically optimizes and adaptively updates the model parameters through a data comparison mechanism and parameter optimization algorithm, so that the system can improve the prediction accuracy as the state of the transmission line changes and the accumulation of actual fault data. The data comparison mechanism is as follows: The closed-loop feedback optimization module first compares the fault probability P1 output by the ARIMA prediction module with the actual fault data P2 collected by the data acquisition module, and calculates the prediction error. ; At the same time, the line health status score output by the entropy weight-Topsis evaluation module will be used. The scoring error is calculated by comparing the results with the actual line condition assessment. The actual line status assessment results consist of manual inspection and verification results, drone inspection and diagnosis results, and measured data from line outage maintenance, which are synchronized to the system through the data acquisition module. Error analysis determines the degree and direction of deviation in model predictions, providing a basis for parameter optimization.

[0011] The parameter optimization algorithm in the closed-loop feedback optimization module includes an improved spectral conjugate gradient method, which iteratively adjusts model parameters, such as autoregressive coefficients. Moving average coefficient The difference order d is used to minimize the prediction error; An improved spectral conjugate gradient optimization algorithm for the ARIMA prediction model, based on prediction error. To optimize the objective, we construct an objective function that minimizes the prediction error: (7); In equation (7), This is the set of parameters to be optimized in the ARIMA model. This represents the total number of samples used for error statistics.

[0012] An attenuation factor adapted to the non-stationary time-series characteristics of transmission lines is introduced during the iterative process of the improved spectral conjugate gradient optimization algorithm. The iterative search direction is corrected, and the iterative formula for the search direction is: (8); In equation (8), For the first The search direction for the next iteration; For the first The gradient of the objective function in the next iteration; These are the spectral conjugate gradient coefficients; It is a symbolic function; It is a dynamic decay factor; Attenuation factor Its core function is to dynamically constrain the search direction, and the line health status score is output by the entropy weight-Topsis evaluation module. The calculation is obtained dynamically, and the calculation formula is as follows: (9); In equation (9), This is the upper limit of the decay factor. This is the lower limit of the attenuation factor; Rate the health status of the current line segment.

[0013] For the parameter optimization of the entropy weight-Topsis evaluation model, including outlier preprocessing optimization and dynamic optimization of the evaluation index system: The 3σ criterion is used to remove and correct outliers from the input multi-source index data. The correction formula is as follows: (10); In equation (10), These are the corrected indicator values; For the first The sample mean of each indicator; For the first The sample standard deviation of the item; Based on indicator information entropy Set the entropy threshold Eliminating information entropy To address the low discrimination index, we updated the evaluation index system and corresponding index weights to improve model stability.

[0014] The risk level classification module is based on the line health status score. Based on the predicted failure probability P, the risk of transmission lines is divided into four levels, retaining low risk, medium risk, high risk, and emergency risk, providing clear risk guidance for operation and maintenance decisions; In the risk level classification module, the rules for determining the risk level are as follows: Low risk: >0.7 and P<0.3; Medium risk: 0.4 < ≤0.7 and P<0.5; High risk: ≤0.4 and P<0.7; Emergency Risk: ≤0.4 and P≥0.7.

[0015] The decision support and visualization platform is the final output of the system. Based on the results of the risk level classification module, the decision support and visualization platform generates inspection suggestions, maintenance priority ranking, and route optimization schemes, and displays them through GIS maps to provide intuitive support for operation and maintenance decisions. The decision support and visualization platform adopts a layered architecture design, including the following functional modules: Data presentation layer: scores the health status of the lines. Predicting the probability of failure Risk levels, GIS maps, statistical charts, and alarm dashboards are displayed in a visual format; Analysis service layer: Based on risk level data, it generates decision support information such as inspection suggestions, maintenance priority ranking, and route optimization schemes; Application Interface Layer: Interfaces with external systems to enable bidirectional data flow and business collaboration.

[0016] The main beneficial effects of this invention are as follows: 1) Multi-algorithm fusion enables accurate and proactive fault prediction, significantly improving prediction reliability: The system deeply integrates the entropy-weighted Topsis comprehensive evaluation algorithm with the ARIMA time-series prediction algorithm. First, the entropy-weighted Topsis algorithm standardizes and objectively weights multi-source data to accurately quantify the health status of transmission lines. Then, based on this, the ARIMA algorithm is used to predict the time-series probability of faults, overcoming the limitations of traditional single-algorithm analysis. Simultaneously, the dynamic weight adjustment module adaptively adjusts the ARIMA model's differencing order based on the line health status score, allowing the model to accurately match prediction strategies for different line states, such as stable, slightly fluctuating, and drastic changes. This effectively improves the accuracy and foresight of fault prediction, providing sufficient time for fault prevention.

[0017] 2) A closed-loop feedback optimization mechanism enables the model to self-iterate and upgrade, adapting to the dynamic operating characteristics of the line: The closed-loop feedback optimization module in the system design calculates the error by comparing the predicted results with actual fault data. It then dynamically optimizes the parameters of the entropy-weighted Topsis and ARIMA models using an improved spectral conjugate gradient method. Furthermore, it introduces an attenuation factor dynamically calculated from the health status score to correct the iteration direction, addressing the non-stationary time-series characteristics of transmission lines. Simultaneously, it employs the 3σ criterion to correct outliers and removes low-discrimination indicators based on entropy thresholds. This mechanism allows the model to continuously self-optimize and self-update as the transmission line's operating status changes and actual fault data accumulates. This effectively solves the problem of traditional models having fixed parameters and difficulty adapting to dynamic line operation, ensuring the system maintains high prediction accuracy throughout long-term operation.

[0018] 3) Multi-dimensional risk grading provides clear quantification for operation and maintenance decisions: Based on a risk level classification module combined with two core indicators—line health status scoring and predicted fault probability—a four-level risk early warning system is established. This system uses clearly defined quantitative thresholds to classify line risks into low, medium, high, and emergency levels, avoiding the ambiguity of experience-based judgments in traditional operation and maintenance. This classification method allows maintenance personnel to intuitively grasp the risk level of each transmission line segment, accurately identify high-risk and emergency-risk lines, and achieve precise positioning and tiered implementation of maintenance work, significantly improving the scientific rigor and relevance of maintenance decisions.

[0019] 4) Full-process decision support and visualization significantly improve the efficiency of transmission line operation and maintenance: The system's decision support and visualization platform automatically generates inspection suggestions, maintenance priority ranking, and route optimization plans based on risk level results. It also uses GIS maps to spatially visualize line risks and employs a layered architecture comprising a data display layer, an analysis service layer, and an application interface layer. This architecture enables intuitive presentation of data such as health status, fault probability, and risk level, as well as bidirectional data collaboration with external systems. This design not only eliminates the tedious process of manually analyzing data and developing maintenance plans but also prioritizes and optimizes inspection and maintenance routes, significantly reducing ineffective maintenance work, lowering labor and material costs, and substantially improving the intelligence and efficiency of transmission line maintenance. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart illustrating the overall steps of an embodiment of the present invention; Figure 2 This is the output interface of the decision support platform of this invention; Figure 3 This is a schematic diagram illustrating the data collection process of this invention. Detailed Implementation

[0021] Example 1: As Figure 1 As shown, a transmission line fault prediction system based on multi-algorithm fusion includes a data acquisition module, an entropy weight-Topsis evaluation module, an ARIMA prediction module, a dynamic weight adjustment module, a closed-loop feedback optimization module, a risk level classification module, and a decision support and visualization platform. The data acquisition module is used to collect multi-source data related to the operation of transmission lines; The Entropy Weight-Topsis evaluation module standardizes the collected multi-source data, calculates the information entropy of each evaluation indicator, determines the indicator weight, constructs a weighted decision matrix, and calculates the distance between each transmission line segment and the optimal and worst solutions to obtain the line health status score. The ARIMA prediction module uses entropy weight-Topsis evaluation results and time series data to construct a differential integrated moving average autoregressive model to predict the probability of transmission line faults in future periods. The dynamic weight adjustment module dynamically adjusts the input parameters and weight coefficients of the ARIMA model based on the line health status score; The closed-loop feedback optimization module compares the prediction results with the actual fault data to optimize the parameters of the entropy weight-Topsis evaluation model and the ARIMA prediction model. The risk level classification module outputs the risk level of the transmission line based on the predicted failure probability and the evaluation of the health status score. The decision support and visualization platform is used to generate inspection suggestions, maintenance priority ranking, and route optimization schemes, and supports GIS map display.

[0022] The entropy weight-Topsis evaluation module uses the range standardization method to standardize the collected multi-source data; (1); In equation (1), The standardized index value, These are the original index values. and These are the minimum and maximum values ​​of the indicator, respectively. The Entropy Weight-Topsis evaluation module calculates the information entropy of each indicator. This reflects the information uncertainty of the indicator, and the specific formula is as follows: (2); In equation (2), , represents the normalization coefficient for entropy calculation. , representing the characteristic weight of the i-th evaluation object on the j-th indicator. n The number of objects to be evaluated; Entropy Weight - The Topsis evaluation module calculates the weight of indicators based on entropy values. The greater the difference, the higher the weight. Calculate the first... j The weights of each indicator are calculated using the following formula: (3); In equation (3), For the first j The weight of each evaluation indicator; For the first j The entropy value of the evaluation index; m The total number of evaluation indicators; The Entropy Weight-Topsis evaluation module calculates the Euclidean distance between each transmission line segment and the positive and negative ideal solutions, and derives a health status score: (4); In equation (4), For the first i The Euclidean distance between each line segment and the ideal solution. For the first i The Euclidean distance between each line segment and the negative ideal solution.

[0023] The ARIMA prediction module uses the ARIMA algorithm for prediction. It eliminates trends through differencing, captures historical dependencies through autoregression, and smooths random fluctuations through moving averages, thereby achieving accurate prediction of future failure probabilities. The ARIMA prediction module uses the following model expression: (5); In equation (5), For the shift operator; These are the autoregressive coefficients; The autoregressive order indicates how many past time points the model uses for prediction; This is the differential order; the dynamic weight adjustment module adjusts this parameter based on the line health status score. The value observed at time t; q The moving average order; The moving average coefficient; Let be a white noise sequence at time t, which follows a normal distribution. (0, ); Based on the time-series fitting and prediction results of the ARIMA model above, the predicted fault probability P of the transmission line within the next T time period is output, and the expression is: (6); In equation (6), For the output of the ARIMA model, the future Predicted values ​​of the line health status at any given time; The threshold value for the critical health state of the line fault is 0.4. For the future The probability that the line health status is lower than the fault threshold during the time period, with a value range of [0,1].

[0024] The dynamic weight adjustment module adjusts the weight based on the line health status score. Dynamically adjust the difference order of the ARIMA model d Weights of input features: when When the value is greater than 0.7, setting d=0 makes the model focus more on predicting stationary sequences. When 0.4 < When ≤0.7, set d=1, the model focuses on predicting slight trends; when When ≤0.4, set d=2, the model focuses on predicting drastic fluctuations and abrupt changes.

[0025] The closed-loop feedback optimization module dynamically optimizes and adaptively updates the model parameters through a data comparison mechanism and parameter optimization algorithm, so that the system can improve the prediction accuracy as the state of the transmission line changes and the accumulation of actual fault data. The data comparison mechanism is as follows: The closed-loop feedback optimization module first compares the fault probability P1 output by the ARIMA prediction module with the actual fault data P2 collected by the data acquisition module, and calculates the prediction error. ; At the same time, the line health status score output by the entropy weight-Topsis evaluation module will be used. The scoring error is calculated by comparing the results with the actual line condition assessment. The actual line status assessment results consist of manual inspection and verification results, drone inspection and diagnosis results, and measured data from line outage maintenance, which are synchronized to the system through the data acquisition module. Error analysis determines the degree and direction of deviation in model predictions, providing a basis for parameter optimization.

[0026] The parameter optimization algorithm in the closed-loop feedback optimization module includes an improved spectral conjugate gradient method, which iteratively adjusts model parameters, such as autoregressive coefficients. Moving average coefficient The difference order d is used to minimize the prediction error; An improved spectral conjugate gradient optimization algorithm for the ARIMA prediction model, based on prediction error. To optimize the objective, we construct an objective function that minimizes the prediction error: (7); In equation (7), This is the set of parameters to be optimized in the ARIMA model. This represents the total number of samples used for error statistics.

[0027] An attenuation factor adapted to the non-stationary time-series characteristics of transmission lines is introduced during the iterative process of the improved spectral conjugate gradient optimization algorithm. The iterative search direction is corrected, and the iterative formula for the search direction is: (8); In equation (8), For the first The search direction for the next iteration; For the first The gradient of the objective function in the next iteration; These are the spectral conjugate gradient coefficients; It is a symbolic function; It is a dynamic decay factor; Attenuation factor Its core function is to dynamically constrain the search direction, and the line health status score is output by the entropy weight-Topsis evaluation module. The calculation is obtained dynamically, and the calculation formula is as follows: (9); In equation (9), This is the upper limit of the decay factor. This is the lower limit of the attenuation factor; Rate the health status of the current line segment.

[0028] For the parameter optimization of the entropy weight-Topsis evaluation model, including outlier preprocessing optimization and dynamic optimization of the evaluation index system: The 3σ criterion is used to remove and correct outliers from the input multi-source index data. The correction formula is as follows: (10); In equation (10), These are the corrected indicator values; For the first The sample mean of each indicator; For the first The sample standard deviation of the item; Based on indicator information entropy Set the entropy threshold Eliminating information entropy To address the low discrimination index, we updated the evaluation index system and corresponding index weights to improve model stability.

[0029] The risk level classification module is based on the line health status score. Based on the predicted failure probability P, the risk of transmission lines is divided into four levels, retaining low risk, medium risk, high risk, and emergency risk, providing clear risk guidance for operation and maintenance decisions; In the risk level classification module, the rules for determining the risk level are as follows: Low risk: >0.7 and P<0.3; Medium risk: 0.4 < ≤0.7 and P<0.5; High risk: ≤0.4 and P<0.7; Emergency Risk: ≤0.4 and P≥0.7.

[0030] The decision support and visualization platform is the final output of the system. Based on the results of the risk level classification module, the decision support and visualization platform generates inspection suggestions, maintenance priority ranking, and route optimization schemes, and displays them through GIS maps to provide intuitive support for operation and maintenance decisions. The decision support and visualization platform adopts a layered architecture design, including the following functional modules: Data presentation layer: scores the health status of the lines. Predicting the probability of failure Risk levels, GIS maps, statistical charts, and alarm dashboards are displayed in a visual format; Analysis service layer: Based on risk level data, it generates decision support information such as inspection suggestions, maintenance priority ranking, and route optimization schemes; Application Interface Layer: Interfaces with external systems to enable bidirectional data flow and business collaboration.

[0031] Example 2: This example provides application data for a 500kV transmission line in a mountainous area managed by a provincial power grid company. The line is 120 kilometers long, traversing complex mountainous terrain with an elevation difference of 800 meters. Historically, it has experienced multiple failures due to icing, lightning strikes, and landslides. A sensor network installed on 156 towers collects environmental data every 5 minutes, including temperature, humidity, wind speed, wind direction, and rainfall, as well as equipment status data such as conductor temperature, insulator leakage current, and tower tilt. Weekly drone inspections acquire line image data, which is then combined with 362 historical fault records from a database. The raw data is preprocessed to remove outliers, fill in missing data, and complete standardization.

[0032] An evaluation system was constructed using 12 key indicators, including icing thickness, temperature change rate, lightning density, insulator pollution level, conductor sag, and tower tilt. The information entropy and weight of each indicator were calculated, with icing thickness being a key indicator. =0.28) and lightning density ( The entropy value is smallest and the weight is highest at (=0.31), respectively. =0.18 and =0.15; The evaluation results show that the health status scores of tower sections #35-#42 and #87-#93 of the line are relatively low. <0.4), mainly affected by icing and lightning strike history.

[0033] Based on historical fault data from the past three years, an initial ARIMA(2,1,1) model is constructed. Due to the mountainous area #35-#42 section =0.32, the system automatically adjusted the difference order to d=2 to enhance the ability to capture sudden trends; the prediction results showed that during the winter icing season, the failure probability P of section #35-#42 would rise to 0.75, and that of section #87-#93 would be 0.48. According to the risk assessment rules, section #35-#42 was classified as "emergency risk", and section #87-#93 as "high risk"; the system generated inspection suggestions: start daily manual inspection of section #35-#42 and install de-icing devices; arrange twice-weekly drone inspections of section #87-#93 to strengthen lightning protection; on the GIS map, the emergency risk section was marked with flashing red, and the optimal inspection route was planned to avoid high-risk areas of landslides; the system predicted icing failure of tower #37 on January 25th 15 days in advance, and the maintenance department deployed de-icing equipment in advance, successfully avoiding a possible line outage. After three months of actual operation, the actual data from three small-scale icing events were compared with the predicted results; the ARIMA model parameters were optimized, and the autoregressive coefficients were adjusted. The moving average coefficient was adjusted from 0.45 to 0.58. The value was adjusted from -0.32 to -0.25; after optimization, the accuracy of icing fault prediction in subsequent predictions improved from 82% to 93%.

[0034] Example 3: This example provides application data using a 220kV transmission network in a coastal city as the application object. The network covers a typhoon-prone area, is 86 kilometers long, and traverses coastal wetlands and urban built-up areas, facing multiple risks such as typhoons, salt spray corrosion, and external damage.

[0035] This study integrates typhoon warning data from the meteorological bureau, salt spray monitoring data from the environmental protection bureau, tide data from the maritime safety administration, and online monitoring data from power transmission lines. The multi-source heterogeneous data is fused in the time and frequency domain to extract key features such as typhoon path, wind speed changes, and salt spray concentration. Satellite remote sensing is used to acquire surface subsidence data in coastal areas to assess the stability risk of power tower foundations. An evaluation system comprising 15 indicators is constructed, with the addition of coastal-specific indicators such as the typhoon risk index, salt spray corrosion rate, and soil resistivity change rate. Calculation results show that the health status scores of areas #23-#31 (coastal tidal flats) and #56-#63 (densely urban areas) are significantly lower than other areas. The values ​​were 0.35 and 0.38 respectively. Through feature engineering units, the wind speed gradient change characteristics in the 72 hours before typhoon landfall were extracted as important inputs for ARIMA prediction. A seasonal ARIMA(1,1,2)(1,1,1) model was constructed to capture the seasonal patterns and short-term abrupt changes of typhoons. When a typhoon warning was issued, the system automatically detected the #23-#31 segment. The probability of failure in the #23-#31 section will be reduced to 0.31. The difference order d will be adjusted from 1 to 2 to enhance the sensitivity to sudden changes. The prediction results show that 24 hours before the typhoon makes landfall, the failure probability P of the #23-#31 section will rise rapidly from 0.25 to 0.83, and the probability of failure in the #56-#63 section will rise to 0.65.

[0036] Based on the tower locations and meteorological conditions, the system divided the power line into 500m×500m grids and generated fault probabilities for each grid according to meteorological conditions. It was found that towers #25-#27 were located in the overlapping area of ​​the typhoon path and high tide level, with a local grid fault probability as high as 0.92. The system generated differentiated inspection strategies: reinforce the tower guy wires of sections #23-#31 48 hours in advance and transfer critical loads; clear tree obstructions of sections #56-#63 to prevent external damage; and simulate the stress on the power line under different typhoon intensities using digital twin technology to predict the possible damage to the foundation of tower #26 and deploy emergency repair teams in advance.

[0037] After the typhoon, comparing the actual fault points (towers #26 and #27) with the predicted results revealed that the model's quantification of the superimposed high tide effect was insufficient. A high tide correction factor was introduced to optimize the difference parameters in the ARIMA model. The entire emergency response process, including warning time, response measures, and effect evaluation, was recorded using blockchain, providing a traceable experience base for subsequent typhoon responses. In subsequent typhoon responses, the optimized model improved the accuracy of fault point prediction to 89% and shortened the emergency response time by 40%.

Claims

1. A transmission line fault prediction system based on multi-algorithm fusion, characterized in that: It includes a data acquisition module, an entropy weight-Topsis evaluation module, an ARIMA prediction module, a dynamic weight adjustment module, a closed-loop feedback optimization module, a risk level classification module, and a decision support and visualization platform; The data acquisition module is used to collect multi-source data related to the operation of transmission lines; The Entropy Weight-Topsis evaluation module standardizes the collected multi-source data, calculates the information entropy of each evaluation indicator, determines the indicator weight, constructs a weighted decision matrix, and calculates the distance between each transmission line segment and the optimal and worst solutions to obtain the line health status score. The ARIMA prediction module uses entropy weight-Topsis evaluation results and time series data to construct a differential integrated moving average autoregressive model to predict the probability of transmission line faults in future periods. The dynamic weight adjustment module dynamically adjusts the input parameters and weight coefficients of the ARIMA model based on the line health status score; The closed-loop feedback optimization module compares the prediction results with the actual fault data to optimize the parameters of the entropy weight-Topsis evaluation model and the ARIMA prediction model. The risk level classification module outputs the risk level of the transmission line based on the predicted failure probability and the evaluation of the health status score. The decision support and visualization platform is used to generate inspection suggestions, maintenance priority ranking, and route optimization schemes, and supports GIS map display.

2. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The entropy weight-Topsis evaluation module uses the range standardization method to standardize the collected multi-source data; （1）； In equation (1), The standardized index value, These are the original index values. and These are the minimum and maximum values ​​of the indicator, respectively. The Entropy Weight-Topsis evaluation module calculates the information entropy of each indicator. This reflects the information uncertainty of the indicator, and the specific formula is as follows: （2）； In equation (2), , represents the normalization coefficient for entropy calculation. , representing the characteristic weight of the i-th evaluation object on the j-th indicator. n The number of objects to be evaluated; The Entropy Weight - Topsis evaluation module calculates the weight of indicators based on entropy values. The greater the difference, the higher the weight. Calculate the first... j The weights of each indicator are calculated using the following formula: （3）； In equation (3), For the first j The weight of each evaluation indicator; For the first j The entropy value of the evaluation index; m The total number of evaluation indicators; The Entropy Weight-Topsis evaluation module calculates the Euclidean distance between each transmission line segment and the positive and negative ideal solutions, and derives a health status score: （4）； In equation (4), For the first i The Euclidean distance between each line segment and the ideal solution. For the first i The Euclidean distance between each line segment and the negative ideal solution.

3. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The ARIMA prediction module uses the ARIMA algorithm for prediction. It eliminates trends through differencing, captures historical dependencies through autoregression, and smooths random fluctuations through moving averages, thereby achieving accurate prediction of future failure probabilities. The ARIMA prediction module uses the following model expression: （5）； In equation (5), For the shift operator; These are the autoregressive coefficients; The autoregressive order indicates how many past time points the model uses for prediction; This is the differential order; the dynamic weight adjustment module adjusts this parameter based on the line health status score. The value observed at time t; q The moving average order; The moving average coefficient; Let be a white noise sequence at time t, which follows a normal distribution. (0, ); Based on the time-series fitting and prediction results of the ARIMA model above, the predicted fault probability P of the transmission line within the next T time period is output, and the expression is: （6）； In equation (6), For the output of the ARIMA model, the future Predicted values ​​of the line health status at any given time; This refers to the critical health state threshold for line faults. For the future The probability that the line health status is lower than the fault threshold during the time period, with a value range of [0,1].

4. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The dynamic weight adjustment module adjusts the weight based on the line health status score. Dynamically adjust the difference order of the ARIMA model d Weights of input features: when When the value is greater than 0.7, setting d=0 makes the model focus more on predicting stationary sequences. When 0.4 < When ≤0.7, set d=1, the model focuses on predicting slight trends; when When ≤0.4, set d=2, the model focuses on predicting drastic fluctuations and abrupt changes.

5. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The closed-loop feedback optimization module dynamically optimizes and adaptively updates the model parameters through a data comparison mechanism and parameter optimization algorithm, so that the system can improve the prediction accuracy as the state of the transmission line changes and the accumulation of actual fault data. The data comparison mechanism is as follows: The closed-loop feedback optimization module first compares the fault probability P1 output by the ARIMA prediction module with the actual fault data P2 collected by the data acquisition module, and calculates the prediction error. ; At the same time, the line health status score output by the entropy weight-Topsis evaluation module will be used. The scoring error is calculated by comparing the results with the actual line condition assessment. Error analysis determines the degree and direction of deviation in model predictions, providing a basis for parameter optimization.

6. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The parameter optimization algorithm in the closed-loop feedback optimization module includes an improved spectral conjugate gradient method, which minimizes the prediction error by iteratively adjusting the model parameters. An improved spectral conjugate gradient optimization algorithm for the ARIMA prediction model, based on prediction error. To optimize the objective, we construct an objective function that minimizes the prediction error: （7）； In equation (7), This is the set of parameters to be optimized in the ARIMA model. This represents the total number of samples used for error statistics.

7. The transmission line fault prediction system based on multi-algorithm fusion according to claim 6, characterized in that: An attenuation factor adapted to the non-stationary time-series characteristics of transmission lines is introduced during the iterative process of the improved spectral conjugate gradient optimization algorithm. The iterative search direction is corrected, and the iterative formula for the search direction is: （8）； In equation (8), For the first The search direction for the next iteration; For the first The gradient of the objective function in the next iteration; These are the spectral conjugate gradient coefficients; It is a symbolic function; It is a dynamic decay factor; Attenuation factor Its core function is to dynamically constrain the search direction, and the line health status score is output by the entropy weight-Topsis evaluation module. The calculation is obtained dynamically, and the calculation formula is as follows: （9）； In equation (9), This is the upper limit of the decay factor. This is the lower limit of the attenuation factor; Rate the health status of the current line segment.

8. The transmission line fault prediction system based on multi-algorithm fusion according to claim 7, characterized in that: For the parameter optimization of the entropy weight-Topsis evaluation model, including outlier preprocessing optimization and dynamic optimization of the evaluation index system: The 3σ criterion is used to remove and correct outliers from the input multi-source index data. The correction formula is as follows: （10）； In equation (10), These are the corrected indicator values; For the first The sample mean of each indicator; For the first The sample standard deviation of the item; Based on indicator information entropy Set the entropy threshold Eliminate information entropy To address the low discrimination index, we updated the evaluation index system and corresponding index weights to improve model stability.

9. The transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The risk level classification module is based on the line health status score. Based on the predicted failure probability P, the risk of transmission lines is divided into four levels, retaining low risk, medium risk, high risk, and emergency risk, providing clear risk guidance for operation and maintenance decisions; In the risk level classification module, the rules for determining the risk level are as follows: Low risk: >0.7 and P<0.3; Medium risk: 0.4 < ≤0.7 and P<0.5; High risk: ≤0.4 and P<0.7; Emergency Risk: ≤0.4 and P≥0.

7.

10. A transmission line fault prediction system based on multi-algorithm fusion according to claim 1, characterized in that: The decision support and visualization platform is the final output of the system. Based on the results of the risk level classification module, the decision support and visualization platform generates inspection suggestions, maintenance priority ranking, and route optimization schemes, and displays them through GIS maps to provide intuitive support for operation and maintenance decisions. The decision support and visualization platform adopts a layered architecture design, including the following functional modules: Data presentation layer: scores the health status of the lines. Predicting the probability of failure Risk levels, GIS maps, statistical charts, and alarm dashboards are displayed in a visual format; Analysis service layer: Based on risk level data, it generates decision support information such as inspection suggestions, maintenance priority ranking, and route optimization schemes; Application Interface Layer: Interfaces with external systems to enable bidirectional data flow and business collaboration.

Images