Insulation performance prediction method and system for live-line work fusing multiple information

Abstract

The application provides a live-line work insulation performance prediction method and system fusing multiple information, and belongs to the technical field of live-line work safety, and the prediction method is applied to a prediction system comprising a data sensing and collecting module, a data analysis module, a decision prediction module, a monitoring feedback module and an alarm module, and specifically comprises the following steps: collecting environment data, stress data and leakage current of an insulation pull rod device, identifying an operation condition and determining a prediction requirement, selecting a short-term recursive prediction strategy or a long-term multi-dimensional input and output prediction strategy in correspondence, predicting a leakage current change trend, identifying a prediction adjustment requirement in real time during the prediction process, performing a prediction strategy transition switching in correspondence, updating a prediction result, and performing integrity scoring on live-line work insulation performance, so as to realize insulation pull rod device safety early warning. Through the setting of the double-time-scale prediction strategy and the transition switching link, the application improves the insulation performance prediction result accuracy and guarantees the operation safety of the insulation pull rod device.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of live working safety, in particular to a live working insulation performance prediction method and system fusing multiple information. BACKGROUND

[0002] As an important support and insulation component of high-voltage power equipment, the insulating pull rod device is mainly used for supporting the conductor and maintaining electrical isolation, and its electrical performance and mechanical strength are directly related to the safety and stability of the operation of the power system. In the actual service process of the insulating pull rod device, it is long-term in a complex environment of high-voltage electric field, mechanical load coupling, etc. Among them, the fluctuation of the excessively high environmental temperature will cause the structural deformation of the insulating pull rod device and the reduction of the end sealing performance due to the thermal expansion of the material, resulting in the invasion of moisture, and then causing the rapid increase of leakage current. In addition, when the environmental humidity exceeds the standard, a water film will be formed on the surface of the insulating pull rod device, and the moisture penetration will cause the deterioration of the dielectric strength of the epoxy resin matrix, and aggravate the risk of flashover or internal breakdown caused by leakage current and partial discharge. In terms of mechanical stress, the periodic fluctuation of the load will also reduce the material life through the fatigue accumulation effect, resulting in the interface peeling of the insulating pull rod laminated plate. Therefore, in the operation process of the insulating pull rod device, it may actually appear short-term abnormality of sudden increase of leakage current due to short-time extreme working conditions such as heavy rain, sudden load increase, etc., and it may also produce slow accumulation of long-term trend due to material aging and local deterioration. That is, under the influence of complex environment, the insulation performance detection of the insulating pull rod device actually has two needs of short-term high-frequency early warning and long-term trend identification according to different operating conditions. The current commonly used insulation detection methods such as manual periodic detection and offline testing after power off cannot meet the dynamic and complex insulation performance detection requirements of the insulating pull rod device due to the problems of poor timeliness and inability to continuous monitoring.

[0003] In order to solve the above problems, the related technology introduces a long short-term memory model LSTM to realize the insulation performance detection of the insulating pull rod device. The LSTM is a deep learning model that can effectively process time series data, which can capture the long-term and short-term dependence relationship between environmental parameters, load characteristics and historical operation data, realize the short-term dynamic prediction and long-term trend prediction of leakage current, and effectively meet the insulation performance detection requirements of the insulating pull rod device.

[0004] However, in practical applications, short-term predictions require high-frequency sampling data as input, focusing on capturing instantaneous fluctuations, while long-term predictions rely on low-frequency trend data, emphasizing slow, gradual changes. These two approaches differ significantly in data time granularity and feature weight allocation. When facing changing detection requirements and needing to switch prediction strategies, sudden parameter changes often cause gaps in the prediction results, making it impossible to form a continuous and consistent insulation performance prediction curve. This can lead to missed short-term sudden faults or misjudgments of long-term aging trends, resulting in low accuracy in the insulation performance prediction of insulated tie rod equipment. Consequently, the insulation degradation of the equipment cannot be accurately identified, and the operational safety of the insulated tie rod equipment cannot be guaranteed. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies that use Long Short-Term Memory (LSTM) models for short-term dynamic and long-term trend prediction of leakage current. When switching prediction strategies, sudden changes in parameters often cause gaps in the prediction results, making it impossible to form a continuous and consistent insulation performance prediction curve. This results in low accuracy of insulation performance prediction and cannot guarantee the operational safety of insulated tie rod equipment. This invention provides a method and system for predicting the insulation performance of live-line work by integrating multiple information sources. Based on a strategy selection mechanism that identifies operating conditions, it selectively chooses either a short-term recursive prediction strategy or a long-term multi-dimensional input-output prediction strategy to predict the trend of leakage current changes. This effectively meets the insulation performance testing requirements of insulated tie rod equipment. Furthermore, when switching prediction strategies, a transition switching step is set up to eliminate prediction errors caused by sudden parameter adjustments, improving the accuracy of insulation performance prediction results. This allows for accurate identification of insulation degradation in insulated tie rod equipment, ensuring its operational safety.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] A method for predicting the insulation performance of live-line work by integrating multiple information sources includes:

[0008] Collect environmental data, stress data, and leakage current of the insulated tie rod equipment during live-line operation;

[0009] Based on environmental and stress data, identify the current operating conditions and determine the predicted demand according to the operating conditions.

[0010] Based on leakage current, either a short-term recursive prediction strategy or a long-term multi-dimensional input-output prediction strategy is selected according to the prediction requirements to predict the trend of leakage current change.

[0011] Based on the leakage current change trend prediction results and operating conditions, the prediction adjustment needs are identified in real time, and the prediction strategy is switched gradually according to the prediction adjustment needs, and the leakage current change trend prediction results are updated.

[0012] According to the prediction result of the current leakage current change trend and the corresponding leakage current prediction strategy, the insulation performance of the live working is scored for integrity;

[0013] Based on the integrity score result, the safety of the insulating pull rod equipment is warned.

[0014] The strategy selection mechanism based on working condition recognition can accurately match the detection needs of the insulating pull rod equipment in different operating states, and then select a short-term recursive prediction strategy or a long-term multi-input and output prediction strategy to develop a leakage current change trend prediction, which can not only respond to short-term abnormalities in equipment insulation performance in time, but also effectively identify the slow accumulation of insulation performance degradation trends such as material aging and local deterioration, achieving full coverage of short-term and long-term needs in insulation performance detection of the insulating pull rod equipment. A transition switching link is set in the prediction strategy switching process, so that the switching between the short-term recursive prediction strategy and the long-term multi-input and output prediction strategy presents a smooth transition state, eliminating the prediction error mutation caused by the step change of parameters, ensuring the continuity and consistency of the leakage current change trend prediction result, and significantly improving the accuracy of the insulation performance prediction result. Based on the accurate insulation degradation identification result, the insulation performance integrity of the insulating pull rod equipment can be accurately quantified and evaluated, and safety warnings can be issued in time, maintenance measures can be taken in advance, and faults such as flashover and internal breakdown caused by insulation performance failure can be effectively avoided, ensuring the safe operation of the insulating pull rod equipment.

[0015] Further, the current operating condition is identified based on the environmental data and the stress data, and the prediction demand is determined according to the operating condition, including:

[0016] The environmental data and the stress data are preprocessed, and the preprocessed environmental data and stress data are subjected to working condition feature extraction;

[0017] Based on the working condition features, the current operating condition is identified in combination with the corresponding working condition classification rules;

[0018] The duration and the working condition type of the current operating condition are obtained, and the corresponding prediction demand is matched.

[0019] Further, the leakage current is based on, and the short-term recursive prediction strategy or the long-term multi-input and output prediction strategy is selected according to the prediction demand, and the leakage current change trend is predicted, including:

[0020] When the working condition type is a short-term extreme working condition and the duration is lower than a preset time threshold, the prediction demand is a short-term high-frequency prediction demand;

[0021] The short-term recursive prediction strategy is selected according to the short-term high-frequency prediction demand, and the leakage current prediction model parameters and the recursive parameters are set according to the corresponding operating condition;

[0022] With the collected leakage current as the initial input, the leakage current value at a future time point is predicted through a leakage current prediction model, and the leakage current prediction value is fed back to the leakage current prediction model as a new input;

[0023] The prediction of the leakage current value and the feedback of the new input are repeatedly performed to generate a continuous short-term prediction sequence;

[0024] A fixed window size is used to slide through the short-term prediction demand generated by recursion to calculate the trend feature in the window;

[0025] Based on the trend feature and the preset rule, the leakage current change trend prediction result is obtained.

[0026] Further, the leakage current is based on the prediction demand to select a short-term recursive prediction strategy or a long-term multi-input and output prediction strategy to predict the leakage current change trend, which further comprises:

[0027] When the working condition type is a long-term stable working condition and the duration is not less than a preset time threshold, the prediction demand is a long-term trend prediction demand;

[0028] When the long-term multi-dimensional input and output prediction strategy is selected according to the long-term trend prediction demand, the leakage current prediction model parameters are set according to the corresponding running working condition;

[0029] Based on the leakage current and its influencing factors, input data samples are constructed, and a continuous long-term prediction sequence is output in combination with the leakage current prediction model;

[0030] The long-term prediction sequence and the collected environmental data and stress data are divided according to the time granularity, the trend features of the long-term prediction sequence in each period and the trend features of the environmental data and stress data in each period are calculated, and corresponding trend category labels are added respectively;

[0031] Based on the trend feature and the corresponding trend category label, the change trend cause is identified, and the leakage current change trend prediction result is obtained.

[0032] Further, the leakage current change trend prediction result and the running working condition are used to identify the prediction adjustment demand in real time, which comprises:

[0033] The current running working condition and its working condition duration are identified, and the matching degree of the prediction strategy is obtained in combination with the current prediction demand;

[0034] The working condition feature of the current running working condition is obtained, and the working condition adaptability of the prediction strategy is obtained in combination with the leakage current change trend prediction result;

[0035] According to the measured value of the current collected leakage current, the trend consistency index of the leakage current change trend prediction result is obtained;

[0036] Identify the prediction adjustment demand based on the matching degree, working condition adaptability and trend consistency index.

[0037] Further, the prediction strategy transition switching according to the prediction adjustment demand updates the leakage current change trend prediction result, comprising:

[0038] Based on the environment data and stress data, obtain the working condition fluctuation coefficient, and set the corresponding transition window according to the working condition fluctuation coefficient;

[0039] In the transition window, the prediction values of the current prediction strategy and the switched target prediction strategy are output synchronously, and the corresponding prediction fusion value is obtained by combining the dynamic weight;

[0040] When the prediction fusion value and the dynamic weight meet the transition end condition, terminate the transition window, and perform leakage current change trend prediction according to the target prediction strategy.

[0041] Further, when the transition window is terminated and the leakage current change trend prediction is performed according to the target prediction strategy, the following is also performed:

[0042] According to the demand type of the prediction adjustment demand, obtain the transition switching target, and select the prediction strategy transition switching direction based on the transition switching target;

[0043] Based on the prediction strategy transition switching direction, establish the trend feature correspondence relationship between the current prediction strategy and the target strategy, and obtain the trend feature list of the target strategy;

[0044] Extract the supplementary features of the current prediction strategy, and add the supplementary features to the trend feature list of the target strategy;

[0045] Based on the trend features in the trend feature list, update the leakage current change trend prediction result.

[0046] Further, the integrity score of the live working insulation performance is obtained according to the current leakage current change trend prediction result and the corresponding leakage current prediction strategy, comprising:

[0047] Obtain the leakage current safety upper limit of the insulating pull rod equipment in the current live working scene, and obtain the corresponding prediction time window according to the current leakage current prediction strategy;

[0048] According to the current leakage current change trend prediction result, obtain the leakage current prediction value at each time in the corresponding prediction time window;

[0049] Based on the leakage current safety upper limit and the leakage current prediction value at each time in the prediction time window, obtain the integrity score of the live working insulation performance.

[0050] The live-line work insulation performance prediction system fusing multiple information is used for executing the prediction method in any one of the above, and comprises:

[0051] The data sensing and collecting module is arranged at the insulation pull rod device, and is used for collecting environmental data, stress data and leakage current of the insulation pull rod device in the live-line work process.

[0052] The data analysis module is connected with the data sensing and collecting module, and is used for identifying the operation condition and determining the prediction requirement according to the collected environmental data, stress data and leakage current.

[0053] The decision prediction module is connected with the data sensing and collecting module and the data analysis module respectively, and is used for selecting a prediction strategy according to the prediction requirement, and predicting the leakage current change trend according to the leakage current.

[0054] The monitoring feedback module is connected with the data analysis module and the decision prediction module respectively, and is used for predicting the prediction adjustment requirement according to the leakage current change trend prediction result and the operation condition, and controlling the decision prediction module to adjust the applied prediction strategy according to the prediction adjustment requirement.

[0055] The alarm module is connected with the decision prediction module, and is used for performing integrity scoring on the live-line work insulation performance according to the leakage current change trend and the prediction strategy, and performing safety warning on the insulation pull rod device according to the integrity scoring.

[0056] Further, the data sensing and collecting module comprises:

[0057] The temperature and humidity sensor is arranged at the middle outer wall of the insulation pull rod device, and is used for collecting the environmental data of the insulation pull rod device in the live-line work process.

[0058] The stress sensor is arranged at the middle surface of the axis of the insulation pull rod device and the metal end, and is used for collecting the stress data of the insulation pull rod device in the live-line work process.

[0059] The leakage current sensor is arranged at the pull rod end of the insulation pull rod device, and is used for collecting the leakage current of the insulation pull rod device in the live-line work process.

[0060] The live-line work insulation performance prediction device fusing multiple information comprises a processor and a storage medium.

[0061] The storage medium is used for storing instructions.

[0062] The processor is used for operating according to the instructions to execute the steps of the live-line work insulation performance prediction method in any one of the above.

[0063] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the live working insulation performance prediction method according to any one of the preceding embodiments.

[0064] The present application has the following advantages:

[0065] (1) The strategy selection mechanism based on working condition recognition can accurately match the detection needs of the insulating pull rod equipment in different operating states, and then select a short-term recursive prediction strategy or a long-term multi-dimensional input-output prediction strategy to develop a leakage current trend prediction, which can not only respond to short-term abnormalities in the insulation performance of the equipment, but also effectively identify the slow accumulation of insulation performance degradation trends such as material aging and local deterioration, thereby achieving full coverage of short-term and long-term needs in the insulation performance detection of the insulating pull rod equipment. A transition switching link is provided in the prediction strategy switching process, which makes the switching between the short-term recursive prediction strategy and the long-term multi-dimensional input-output prediction strategy present a smooth transition state, eliminates the prediction error mutation caused by the step change of parameters, ensures the continuity and consistency of the leakage current trend prediction results, and significantly improves the accuracy of the insulation performance prediction results. Based on the accurate insulation degradation identification results, the insulation performance integrity of the insulating pull rod equipment can be accurately quantified and evaluated, and safety warnings can be issued in a timely manner to take maintenance measures in advance, effectively avoiding faults such as flashover and internal breakdown caused by insulation performance failure, and ensuring the safe operation of the insulating pull rod equipment;

[0066] (2) In the short-term prediction, a continuous prediction sequence is generated based on recursive feedback, and trend features are extracted combined with a sliding window, which not only ensures high-frequency capture of instantaneous fluctuations, but also avoids the randomness of a single prediction value through trend feature analysis, making the identification of abnormal trends such as short-term leakage current surge more reliable. In the long-term prediction, input samples are constructed by associating environmental and stress factors, and the driving mechanism of long-term degradation trends such as material aging is reflected through multi-dimensional trend feature analysis and cause analysis, to ensure the reliability of the obtained long-term trend prediction results;

[0067] (3) Through multi-dimensional evaluation of matching degree, working condition adaptability and trend consistency index, the dynamic quantification of the adaptability of the prediction strategy is realized, providing a precise basis for strategy switching. During the transition switching, a transition window is set based on the working condition fluctuation coefficient, the dynamic weight is fused to combine the prediction values of the two strategies, and the feature supplement and trend feature correspondence are established to ensure the continuity and integrity of the change trend during the switching process, effectively eliminating the prediction fault caused by parameter mutation. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 is a flowchart of the present application;

[0069] Figure 2is a structural schematic diagram of an embodiment of the application;

[0070] Figure 3 is a position schematic diagram of a data sensing and collecting module at an insulated pull rod device of an embodiment of the application;

[0071] Wherein: 1, data sensing and collecting module, 11, temperature and humidity sensor, 111, temperature sensor, 112, humidity sensor, 12, stress sensor, 121, tensile load sensor, 122, shear load sensor, 13, leakage current sensor, 2, data analysis module, 3, decision prediction module, 4, monitoring feedback module, 5, alarm module, 6, insulated pull rod device. DETAILED DESCRIPTION

[0072] The application will be further described below in combination with the drawings and embodiments.

[0073] Embodiment one: The live working insulation performance prediction method fusing multiple information, as shown in the figure, comprises: Figure 1

[0074] Collecting environmental data, stress data and leakage current of the insulated pull rod device during the live working process;

[0075] Based on the environmental data and the stress data, identifying the current operating condition, and determining the prediction requirement according to the operating condition;

[0076] Based on the leakage current, selecting a short-term recursive prediction strategy or a long-term multi-dimensional input and output prediction strategy according to the prediction requirement, and predicting the leakage current change trend;

[0077] According to the leakage current change trend prediction result and the operating condition, identifying the prediction adjustment requirement in real time, performing prediction strategy transition switching according to the prediction adjustment requirement, and updating the leakage current change trend prediction result;

[0078] According to the current leakage current change trend prediction result and the corresponding leakage current prediction strategy, performing integrity scoring on the live working insulation performance;

[0079] Based on the integrity scoring result, performing safety warning of the insulated pull rod device.

[0080] In order to realize accurate detection and prediction of the insulation performance of the insulated pull rod device, in the data collection link, considering that the insulation performance degradation of the insulated pull rod device is closely related to the environment, stress and its own operating state, multi-dimensional data needs to be collected synchronously during the live working process of the insulated pull rod device.

[0081] ​Among them, the environmental data covers key parameters such as temperature and humidity, which can capture the influence of temperature and humidity fluctuations on the formation of water film on the surface of the equipment and the penetration of moisture. The stress data focuses on the periodic changes and instantaneous impacts of the load, which can be used to analyze the effects of mechanical stress on material fatigue accumulation and interface peeling. The leakage current data directly reflects the real-time state of the insulation performance of the equipment, and is a core indicator for judging the degree of insulation degradation. Therefore, during the live working process of the insulating pull rod equipment, the above three types of data are continuously collected to provide complete and real-time basic data support for subsequent working condition identification, prediction strategy selection and performance evaluation.

[0082] On the basis of the collected environmental data and stress data, further identification of operating conditions and determination of prediction requirements are carried out.

[0083] Specifically, based on the environmental data and stress data, the current operating condition is identified, and the prediction requirement is determined according to the operating condition, including:

[0084] The environmental data and stress data are preprocessed, and the operating condition features of the preprocessed environmental data and stress data are extracted;

[0085] Based on the operating condition features, the current operating condition is identified by combining the corresponding operating condition classification rules;

[0086] The duration and type of the current operating condition are obtained, and the corresponding prediction requirement is matched.

[0087] Considering the complex operating environment of the insulating pull rod equipment, the environmental data and stress data often have noise, missing or abnormal fluctuations, etc. If the original data is directly used for operating condition identification, the identification result will be inaccurate, which will affect the determination of subsequent prediction requirements. Therefore, the environmental data and stress data are preprocessed, including outlier rejection and data standardization, to eliminate noise interference.

[0088] And because the insulation performance degradation characteristics of the equipment under different operating conditions are quite different, only by accurately extracting operating condition features, combining classification rules to identify operating conditions, and matching prediction requirements according to operating condition duration and type, can the applicability of subsequent prediction strategies be ensured, and different needs of short-term high-frequency early warning and long-term trend identification can be met.

[0089] Therefore, operating condition features are extracted from the preprocessed environmental data and stress data to mine key information that can reflect the essence of the operating condition, such as temperature change rate, humidity peak value, stress fluctuation period, etc.

[0090] After obtaining the working condition characteristics, the current running working condition is accurately identified in combination with the preset working condition classification rules. Specifically, the working condition classification rules are constructed based on the degradation mechanism and historical running data of the insulating pull rod equipment. For example, based on the historical running data, the short-term extreme working condition is defined as humidity > 90% and duration > 1 hour, and load peak value > 1.2 times of the rated value, the long-term stable working condition is defined as temperature fluctuation < 5 ℃ / day, humidity < 60%, and load fluctuation frequency < 0.1 Hz.

[0091] In the identification process, the matching degree of the current working condition characteristics and each category rule is calculated, such as the number of matching parameters, the deviation degree, etc., to determine the working condition type.

[0092] After identifying the current running working condition type, the duration thereof needs to be further counted, and the corresponding prediction requirement is determined in combination with the two.

[0093] For the short-term extreme working condition, if the duration is short, i.e. lower than the preset time threshold, and the characteristic parameter presents a sudden rise and fall trend, the prediction requirement is focused on short-term high-frequency early warning, which is a short-term high-frequency prediction requirement, requiring the leakage current prediction model to capture the future leakage current instantaneous change, so as to timely trigger the emergency response. If the duration of the short-term extreme working condition is long, and the characteristic parameter is maintained at a high level fluctuation, such as continuous heavy rain leading to long-time humidity exceeding the standard, in addition to the short-term early warning, the acceleration effect of this working condition on long-term degradation also needs to be considered, and the preliminary judgment of the subsequent long-term trend is included in the prediction requirement.

[0094] For the long-term stable working condition, if the duration is long, i.e. not lower than the preset time threshold, and the characteristic parameter has no significant fluctuation, the prediction requirement is mainly the identification of long-term trend, which is a long-term trend prediction requirement, requiring the leakage current prediction model to output the leakage current change trend in the future long period, such as week, month, etc., to evaluate the cumulative effect of material aging. If there is a slow gradual change in the characteristic parameter in the long-term stable working condition, the weight distribution of this gradual change factor needs to be strengthened in the long-term trend prediction, to ensure that the potential degradation acceleration signal can be identified in time.

[0095] And the above-mentioned preset time threshold can be set according to the requirement.

[0096] Through the demand matching mechanism combining the working condition type and the duration, the selection of the prediction strategy is more suitable for the actual running state of the equipment, effectively avoiding the problem of insufficient accuracy caused by single prediction strategy.

[0097] After determining the relevant prediction requirement, the corresponding prediction strategy can be further selected, and the corresponding leakage current change trend prediction is carried out through the corresponding leakage current prediction model.

[0098] The embodiment specifically takes long short-term memory network as a modeling tool to establish a corresponding leakage current prediction model to effectively capture long-term dependencies and trend changes in time series. Specifically, the size and application frequency of the tensile load and shear load of the insulating pull rod in actual operation are taken as the main input characteristics, and historical temperature, humidity and other environmental variables are introduced to construct the input data sequence. The leakage current prediction model is trained by supervised learning to learn the evolution trend of leakage current under different loads and environmental conditions, so that it has the ability to predict future leakage current changes. And further introduce the attention mechanism (Attention Module) in the LSTM main structure, enhance the response accuracy of the leakage current prediction model to sudden events and key time periods, improve the weight perception of key moments such as load mutation or rapid rise of humidity, so as to improve the prediction accuracy. And the leakage current prediction model adopts mean square error (MSE) as the main loss function in the training process, and introduces a weighted penalty mechanism for sudden paragraphs to ensure more accurate prediction at key leakage risk points. The trained leakage current prediction model is deployed on an embedded edge computing platform to realize on-site online operation and real-time feedback.

[0099] In terms of prediction strategy, the leakage current prediction model adopts two prediction modes. One is recursive prediction, that is, the leakage current prediction model only predicts the leakage current at the next time point each time, and the output is fed back as the input for the next prediction, which can be used for continuous prediction with short time steps. One is a long-term multi-input and output prediction strategy, which takes a period of historical data as input and outputs the leakage current prediction value of multiple time steps in the future at a time, which is suitable for medium and long-term trend analysis and whole segment performance evaluation.

[0100] For short-term high-frequency demand under short-term extreme working conditions, such as sudden rise of humidity and peak load exceeding the standard, the recursive prediction strategy is preferred. Under such working conditions, the leakage current is easily affected by instantaneous environment and stress fluctuations, and needs to capture the sudden increase trend with high frequency granularity in a short time. The single-step output and feedback iteration characteristics of recursive prediction can respond to subtle changes at each time point in real time. Especially when the duration of short-term extreme working conditions is short and the characteristic parameters rise and fall suddenly, recursive prediction can capture the mutation nodes of leakage current in time through the continuous generation of short-term sequences, providing a quick warning window for emergency response.

[0101] Under such operating conditions, the leakage current is predicted according to the prediction demand to select a short-term recursive prediction strategy to predict the leakage current trend, comprising:

[0102] When the working condition type is short-term extreme working condition and the duration is less than the preset time threshold, the prediction demand is short-term high-frequency prediction demand;

[0103] According to the short-term high-frequency prediction demand, a short-term recursive prediction strategy is selected, and the leakage current prediction model parameters and the recursive parameters are set according to the corresponding operating conditions;

[0104] With the collected leakage current as the initial input, the leakage current value at a future time point is predicted through the leakage current prediction model, and the leakage current prediction value is fed back to the leakage current prediction model as a new input;

[0105] The prediction of the leakage current value and the feedback of the new input are repeatedly performed to generate a continuous short-term prediction sequence;

[0106] A fixed window size is used to slide through the recursively generated short-term prediction demand, and the trend characteristics in the window are calculated;

[0107] Based on the trend characteristics and preset rules, the leakage current change trend prediction result is obtained.

[0108] In the parameter setting stage, the leakage current prediction model parameters and the recursive parameters need to be configured in combination with the short-term extreme condition characteristics corresponding to the prediction demand. For example, when the condition is that heavy rain causes humidity to rise sharply and is accompanied by short-time load impact, the weight coefficient of the humidity factor in the model parameters needs to be increased, and the number of hidden layer units of LSTM in the leakage current prediction model needs to be increased to enhance the fitting ability of high-frequency fluctuation data. The recursive parameters need to be set according to the warning response time requirement, including the recursive step and the maximum number of recursions, and the upper and lower threshold values of the prediction value are set based on the historical maximum leakage current value under this condition to avoid prediction drift caused by extreme data disturbance.

[0109] After the relevant parameters are set, the collected leakage current is used as the initial input, the leakage current value at a future time point is predicted through the leakage current prediction model, and the leakage current prediction value is fed back to the leakage current prediction model as a new input. In this way, the iterative feedback process is completed.

[0110] The trend characteristics are calculated by using a fixed window sliding traversal method, and the window size is set according to the fluctuation period of the short-term extreme condition. During the sliding process, the trend characteristics in each window are calculated, including the average growth rate of the leakage current, the peak value occurrence frequency, the fluctuation variance, etc. These characteristics can effectively reflect the change law of the local period.

[0111] Based on the calculated trend characteristics, the corresponding leakage current change trend prediction result is generated in combination with the preset rules. The preset rules need to be set in combination with the degradation threshold of the insulation pull rod device, such as when the average growth rate in the window is >100% and the peak value is ≥30μA, it is determined as a high-risk rapid increase trend, when the average growth rate is between 50%-100% and the fluctuation variance is stable, it is determined as a controllable growth trend, etc.

[0112] By matching the calculated trend features with rules, the final output prediction results not only contain specific leakage current value sequences, but also include trend type labels and key time nodes.

[0113] The working condition adaptability configuration and iterative feedback mechanism of this parameter make the prediction sequence accurately track the leakage current instantaneous fluctuation under short-term extreme working conditions, effectively reducing the corresponding prediction error. By extracting trend features through a sliding window, the randomness of a single prediction value is avoided, and by aggregating and analyzing local trends, the accuracy of trend judgment is effectively improved, effectively avoiding the flashover risk of insulation rod equipment. The generated continuous prediction sequence and trend table can also provide fine-grained feature basis for subsequent strategy transition switching, ensuring smooth connection from short-term early warning to long-term trend analysis.

[0114] For long-term trend prediction needs under long-term stable working conditions, such as gentle fluctuations in temperature and humidity, and stable loads, a long-term multi-input-output prediction strategy is adapted. This strategy takes historical data segments as input and outputs multiple time step results at once, covering the slow degradation process caused by material aging. When there are slow gradual changes in characteristic parameters in long-term stable working conditions, the multi-input-output mode can also integrate the long-term trend features of parameters such as temperature, humidity, and load, reflecting the cumulative impact of these factors on leakage current in the prediction results, providing quantitative basis for long-term life assessment for the entire period.

[0115] Under such operating conditions, the long-term multi-input-output prediction strategy is selected based on the leakage current according to the prediction requirements to predict the leakage current trend, including:

[0116] When the working condition type is a long-term stable working condition and the duration is not less than a preset time threshold, the prediction requirement is a long-term trend prediction requirement;

[0117] When selecting a long-term multi-dimensional input-output prediction strategy according to the long-term trend prediction requirement, set the leakage current prediction model parameters according to the corresponding operating conditions;

[0118] Based on the leakage current and its influencing factors, input data samples are constructed, and a continuous long-term prediction sequence is output by combining the leakage current prediction model;

[0119] According to the time granularity, the long-term prediction sequence and the collected environmental data and stress data are divided, the trend features of the long-term prediction sequence and the trend features of the environmental data and stress data in each period are calculated, and the corresponding trend category labels are added;

[0120] Based on the trend features and corresponding trend category labels, the causes of the change trend are identified, and the leakage current change trend prediction result is obtained.

[0121] In the corresponding parameter setting stage, the characteristics of long-term stable working conditions need to be matched, and the leakage current prediction model parameters are configured. In the long-term stable working condition, the degradation of the insulating pull rod equipment is mainly caused by material aging, slow accumulation of environmental influence and stable mechanical stress. The model parameters of the leakage current prediction model need to focus on enhancing the fitting ability of the low-frequency trend characteristics. For example, in the leakage current prediction model LSTM, the learning rate can be appropriately reduced to avoid overfitting of high-frequency noise, the time step can be increased to cover longer period changes, and the number of hidden layer neurons can be adjusted to improve the modeling ability of multiple factors. In addition, for different long-term stable working condition subtypes, such as high temperature and dry long-term working condition, normal temperature and humidity long-term working condition, the environmental factor weight needs to be differentiated, such as increasing the weight of temperature parameter in high temperature and dry working condition, and appropriately increasing the influence weight of humidity parameter in normal temperature and humidity working condition, to ensure that the leakage current prediction model can accurately adapt to the degradation driving mechanism of the specific working condition.

[0122] After completing the corresponding parameter setting, the multi-dimensional data of leakage current and its influencing factors need to be integrated to build input data samples. The influencing factors include environmental data such as daily average temperature, average humidity, and stress data such as daily load average, load fluctuation amplitude. Align these data with the corresponding period of leakage current data in time to build input data samples containing multiple characteristics. Then, according to the input data samples, the leakage current prediction model is used for prediction, and the continuous long-term prediction sequence is output.

[0123] This input sample through multi-factor fusion and model parameter setting adapted to long-term working conditions can accurately reflect the slow change trend of leakage current in the long-term prediction sequence, and compared with the single factor driven prediction model, it can effectively reduce the corresponding prediction error.

[0124] After outputting the continuous long-term prediction sequence, the long-term prediction sequence and the collected environmental data and stress data are divided according to the preset time granularity, such as weekly. For each period, the trend characteristics of the long-term prediction sequence are calculated, such as the weekly average growth rate of leakage current, the difference between the maximum and minimum values of the week, the slope of the trend line, etc. For environmental data, the average temperature change rate and average humidity fluctuation range of each period are calculated. For stress data, the load average fluctuation amplitude and load peak value occurrence frequency of each period are calculated. Subsequently, the trend category labels are added to these trend characteristics, such as slow rise, basic stability, and slight decline for leakage current trend characteristics, temperature slowly rising, humidity stable, etc. for environmental data trend characteristics, and load fluctuation gently, load average slightly increasing, etc. for stress data trend characteristics.

[0125] Based on the trend characteristics and corresponding trend category labels of each period, the causes of the leakage current change trend are accurately judged through correlation analysis. For example, when the leakage current trend characteristics of a period are slowly rising, the corresponding environmental data trend characteristics are slowly rising humidity, and the stress data trend characteristics are gently fluctuating load, combined with the preset cause analysis rules, such as long-term slow rise in humidity easily leading to moisture penetration and causing slow rise in leakage current, it can be identified that the main cause of the leakage current rise in this period is the long-term cumulative effect of humidity. Or, when the leakage current slowly rises and the stress data trend characteristics are slightly increased load average, and the environmental data trend is stable, it can be judged that the cause is mainly the material fatigue accumulation under the long-term action of mechanical stress.

[0126] Through detailed trend characteristics and cause analysis, sufficient feature support is provided for the smooth transition of short-term prediction strategy and long-term prediction strategy, ensuring the coherence and consistency between different prediction strategies, and further improving the overall reliability of the insulation performance detection of the insulating pull rod device.

[0127] However, in the actual operation process of the insulating pull rod device, the operating conditions are not constant, but dynamically change between short-term extreme conditions and long-term stable conditions. For example, a sudden rainstorm may cause the device to switch from long-term stable normal temperature and humidity conditions to short-term extreme high humidity impact conditions, and after the rainstorm ends, the device will gradually return to long-term stable state; or in long-term stable operation, due to seasonal alternation, the environmental parameters change slowly, causing the operating conditions to gradually transition from high temperature and dryness to normal temperature and high humidity.

[0128] The above-mentioned short-term recursive prediction strategy and long-term multi-input-output prediction strategy can adapt to the prediction needs in short-term extreme conditions and long-term stable conditions respectively, but in the dynamic transition process of operating conditions, if a single strategy is still used, the prediction accuracy will be reduced due to the decreased adaptability of the strategy to the operating conditions. For example, when the device transitions from short-term extreme conditions to long-term stable conditions, the short-term recursive prediction strategy is difficult to capture the long-term trend after the stable operating conditions due to its excessive focus on high-frequency fluctuations, while the long-term multi-input-output prediction strategy is unable to timely track the leakage current fluctuations in the transition stage due to its slow response to instantaneous changes.

[0129] Therefore, in order to ensure the accuracy and continuity of the prediction when the operating conditions dynamically change, based on the leakage current change trend prediction results and real-time operating conditions, real-time prediction adjustment needs are identified, and then it is determined whether the prediction strategy needs to be switched, in order to compensate for the limitations of a single prediction strategy in operating condition transition, and to ensure that the prediction results always match the actual operating state of the device, providing reliable support for the insulation performance evaluation and safety warning of the insulating pull rod device.

[0130] The prediction adjustment demand is identified in real time according to the leakage current change trend prediction result and the operation condition, and comprises:

[0131] The matching degree of the prediction strategy is obtained by identifying the current operation condition and the operation condition duration and combining the current prediction demand;

[0132] The operation condition characteristics of the current operation condition are obtained, and the operation condition adaptability of the prediction strategy is obtained by combining the leakage current change trend prediction result;

[0133] The trend consistency index of the leakage current change trend prediction result is obtained according to the current collected leakage current measured value;

[0134] The prediction adjustment demand is identified based on the matching degree, the operation condition adaptability and the trend consistency index.

[0135] Based on the demand for adaptability of the prediction strategy during dynamic conversion of the operation condition, the applicability of the current prediction strategy is quantitatively evaluated through multi-dimensional indexes, and then the prediction adjustment demand is accurately identified. Specifically, the matching degree, the operation condition adaptability and the trend consistency index of the prediction strategy are taken as the evaluation indexes of the applicability.

[0136] For the matching degree calculation of the prediction strategy, the current operation condition, the operation condition duration and the current prediction demand need to be comprehensively evaluated. First, the type and duration of the current operation condition are determined, and then the current prediction demand is compared to calculate the adaptation degree through the preset matching degree scoring rule.

[0137] The matching degree scoring rule can be: matching degree score = operation condition type adaptation score x duration correction coefficient, wherein the operation condition type adaptation score is 90 when the short-term extreme operation condition is adapted to the short-term recursive prediction strategy and the long-term stable operation condition is adapted to the long-term multi-input and output prediction strategy, otherwise the basic score is 40, and the duration correction parameter is set according to the degree of deviation of the operation condition duration from the typical adaptation time, the greater the deviation, the smaller the duration correction coefficient.

[0138] For the operation condition adaptability calculation of the prediction strategy, the fitting degree of the operation condition characteristics of the current operation condition and the leakage current change trend prediction result needs to be focused on. First, the key features of the current operation condition are extracted, such as the humidity sudden rise amplitude in the short-term extreme operation condition, the load impact frequency, the temperature fluctuation range in the long-term stable operation condition, and the load mean change rate, etc. Then, it is analyzed whether the leakage current change trend prediction result can accurately reflect the influence of these features, for example, in the short-term extreme operation condition of humidity sudden rise, if the prediction result shows a slow growth trend of leakage current, which is not consistent with the operation condition characteristics that the humidity sudden rise should cause the rapid rise of leakage current, then the operation condition adaptability is low, and if the growth trend of the prediction result is positively correlated with the humidity sudden rise amplitude, then the operation condition adaptability is high.

[0139] For the trend consistency index calculation of the prediction strategy, the current collected leakage current measured value is taken as the benchmark to measure the degree of agreement between the prediction result and the actual change. First, by comparing the predicted leakage current trend curve with the measured curve, the deviation rate of the two at the same time node is calculated. Then, combined with the deviation rate average, maximum deviation value and other indicators of the entire monitoring period, the trend consistency index is obtained according to the consistency scoring formula. The consistency scoring formula can be: consistency score = 1 - deviation rate average.

[0140] Based on the evaluation results of the above three indicators, the prediction adjustment requirement is identified through a preset decision rule. For example, when the matching degree is less than 60 points, the working condition adaptability is low, and the trend consistency index is less than 75%, it is determined that the emergency adjustment requirement needs to be switched immediately; when two of the indicators do not meet the standard, it is determined that the regular adjustment requirement needs to be completed within a certain time; when only one indicator does not meet the standard, it is determined that the observation adjustment requirement needs to be continuously monitored for the subsequent working condition and prediction result changes.

[0141] Through the above multi-dimensional quantitative evaluation, the prediction adjustment requirement is accurately identified, the blindness of strategy switching is avoided, and through the clear index threshold and decision rule, the error of human judgment is reduced, the identification accuracy of the adjustment requirement is improved, a reliable foundation is laid for the smooth transition of the subsequent prediction strategy, and the consistency between the leakage current trend prediction and the equipment running state is ensured.

[0142] After identifying the prediction adjustment requirement through the matching degree, working condition adaptability and trend consistency index, it means that the currently used prediction strategy cannot accurately adapt to the running working condition of the equipment, and if it is not adjusted in time, it may cause deviation in the prediction result of the leakage current trend, affecting the accurate judgment of the insulation performance of the insulated pull rod equipment.

[0143] At this time, in order to ensure the continuity and accuracy of the prediction, the transition switching of the prediction strategy needs to be carried out according to the identified prediction adjustment requirement. This transition switching is not a simple strategy replacement, but a smooth connection mechanism between the two prediction strategies to cope with the dynamic changes of the running working condition. Through reasonable transition switching, the prediction error mutation caused by step change of parameters is eliminated, the continuity and consistency of the leakage current trend prediction result are ensured, and then the leakage current trend prediction result is updated to ensure that the prediction result can always reflect the insulation performance state of the equipment, providing reliable data support for the safety evaluation and early warning of the insulated pull rod equipment.

[0144] Specifically, the transition switching of the prediction strategy according to the prediction adjustment requirement, updating the leakage current trend prediction result, includes:

[0145] The working condition fluctuation coefficient is obtained based on the environmental data and stress data, and a corresponding transition window is set according to the working condition fluctuation coefficient;

[0146] In the transition window, the prediction values of the current prediction strategy and the switched target prediction strategy are output synchronously, and a corresponding prediction fusion value is obtained by combining the dynamic weight;

[0147] When the prediction fusion value and the dynamic weight meet the transition end condition, the transition window is terminated, and the leakage current trend prediction is performed according to the target prediction strategy.

[0148] After identifying the prediction adjustment requirement, in order to realize the smooth transition switching of the prediction strategy and update the leakage current trend prediction result, the transition process needs to be regulated by the working condition fluctuation coefficient.

[0149] The working condition fluctuation coefficient is obtained based on the dynamic changes of environmental data and stress data, and is used to quantitatively evaluate the severity of working condition changes. In the calculation, the instantaneous change rate of temperature and humidity, and the standard deviation of load fluctuation are selected as the core parameters. After normalization, the working condition fluctuation coefficient is calculated by using the weighted summation formula. The higher the coefficient, the more severe the working condition changes. Then, according to the size of the working condition fluctuation coefficient, the length of the transition window is set. The larger the working condition fluctuation coefficient, the more severe the changes of environmental and stress parameters, and the larger the difference in adaptability between the two prediction strategies. Taking the short-term recursive strategy as an example, although it can capture instantaneous fluctuations, it is difficult to adapt to the stable trend that will come soon. The target strategy, i.e. long-term multi-input and output, although it adapts to the new working condition, it needs a longer time to learn the rules after the severe changes. Therefore, a longer transition window must be set to allow the prediction values of the two strategies to fully fuse and gradually offset the prediction deviation caused by severe fluctuations. Therefore, the larger the fluctuation coefficient, the longer the transition window, ensuring that the transition process matches the working condition change speed.

[0150] In the transition window, a double-strategy parallel output and dynamic weight fusion mechanism is adopted. Specifically, the current prediction strategy and the target prediction strategy are run synchronously to generate the leakage current prediction values at the corresponding time points. The dynamic weight is distributed based on the real-time changes of the working condition fluctuation coefficient. At the initial moment, the weight of the current prediction strategy is 1 and the weight of the target strategy is 0. As the window advances, the weight of the current prediction strategy decreases in a linear decay manner, and the weight of the target strategy increases in a linear increasing manner. The decay or increase rate is determined by the window length. The prediction fusion value is the sum of the products of the prediction values of the two strategies and the corresponding weights, which not only retains the adaptability of the current strategy to the previous working condition, but also gradually introduces the fitting ability of the target strategy to the new working condition, avoiding the prediction mutation caused by single strategy switching.

[0151] The judgment of the transition end condition needs to meet the double requirements of the predicted fusion value and the dynamic weight. Specifically, when the fluctuation amplitudes of the fusion values at three consecutive time points are all less than a preset threshold, and the target strategy weight reaches 1 and remains stable, it is judged that the transition end condition is met, the transition window is terminated, and the target prediction strategy is formally switched to for the leakage current change trend prediction. If the condition is not met within the window duration, the window is automatically extended to ensure that the transition process is fully completed.

[0152] This way of dynamically adjusting the transition window through the working condition volatility technology can accurately match the strategy switching rhythm with the working condition change speed. In the case of severe working condition changes, the window is extended to smooth the transition, and in the case of gentle changes, the window is shortened to improve the response efficiency. Furthermore, through the dynamic weight fusion mechanism, the prediction fault during strategy switching is effectively eliminated, realizing the seamless connection of short-term recursive and long-term multi-input-output strategies, and ensuring the continuity and accuracy of the leakage current change trend prediction.

[0153] In addition, although the transition process has been smoothly connected through the fusion mechanism after the termination of the transition window and the switching to the target prediction strategy, the target strategy may ignore the key details captured by the current strategy in the old working condition, such as the potential impact of the instantaneous fluctuation characteristics in the short-term extreme working condition on the long-term trend, due to its focus on the trend characteristics of the new working condition. In addition, the short-term strategy focuses on high-frequency fluctuation characteristics, and the long-term strategy focuses on low-frequency trend characteristics. The direct switching may lead to incomplete feature transmission, affecting the target strategy's comprehensive judgment of the leakage current change trend. Therefore, it is necessary to establish a feature correspondence relationship and supplement key features to ensure that the target strategy can integrate the feature information of the new and old working conditions and improve the accuracy and completeness of the prediction results.

[0154] Specifically, when the transition window is terminated and the leakage current change trend prediction is performed according to the target prediction strategy, the following is also performed:

[0155] According to the demand type of the prediction adjustment requirement, a transition switching target is obtained, and a prediction strategy transition switching direction is selected based on the transition switching target;

[0156] Based on the prediction strategy transition switching direction, a trend feature correspondence relationship between the current prediction strategy and the target strategy is established, and a trend feature list of the target strategy is obtained;

[0157] Supplementary features of the current prediction strategy are extracted, and the supplementary features are added to the trend feature list of the target strategy;

[0158] Based on the trend features in the trend feature list, the leakage current change trend prediction result is updated.

[0159] Firstly, the transition switching target is determined according to the demand type of the predicted adjusted demand, and then the transition switching direction of the prediction strategy is determined. For example, when the demand type is switching from short-term extreme working condition to long-term stable working condition, the transition switching target is long-term multi-input and output prediction strategy, and the switching direction is short-term to long-term. When the demand type is switching from long-term stable working condition to short-term extreme working condition, the transition switching target is short-term recursive prediction strategy, and the switching direction is long-term to short-term. The determination of the switching direction can provide direction guidance for the establishment of the feature correspondence relationship.

[0160] Based on the switching direction, the trend feature dimensions concerned by the current prediction strategy and the target strategy are determined to establish the feature correspondence relationship. Taking the switching from short-term to long-term as an example, the trend features of the short-term prediction strategy include the instantaneous growth rate of leakage current, high-frequency fluctuation peak value, etc., and the trend features of the long-term prediction strategy include the weekly average growth rate, low-frequency trend slope, etc. Through the correspondence relationship, the instantaneous growth rate is associated with the initial growth momentum in the long-term trend, the high-frequency peak value is associated with the fluctuation risk point in the long-term trend, etc. Subsequently, the trend features identified by the target strategy in the new working condition are extracted, such as the temperature cumulative impact feature in the long-term stable working condition, the load mean value change feature, to form the initial trend feature list of the target strategy. The feature parameters recorded in this trend feature list are historical data.

[0161] Then, the supplementary features with supplementary value for the target strategy are extracted from the historical prediction results of the current prediction strategy. If the current strategy is short-term recursive, the supplementary features can include the maximum leakage current value in the short-term extreme working condition, the duration of sudden increase of leakage current, etc. These features can reflect the damage degree of the material under extreme conditions and affect the long-term degradation speed. If the current strategy is long-term multi-input and output strategy, the supplementary features can include the periodic small fluctuation feature hidden in the long-term trend, which provides a judgment basis for the short-term strategy to judge the instantaneous abnormality.

[0162] These supplementary features are mapped to the feature dimensions of the target strategy according to the correspondence relationship, and are added to the trend feature list of the target strategy, so that the trend feature list covers the key features of the new and old working conditions.

[0163] The target strategy takes the trend feature list after supplementation as input, and re-trains or adjusts the leakage current prediction model parameters. For example, after incorporating the long-term supplementary features, the short-term strategy will adjust the judgment threshold of the instantaneous abnormality in combination with the periodic fluctuation benchmark to avoid misjudging the normal fluctuation in the long-term trend as a sudden abnormality. On this basis, the updated prediction results are output through the adjusted leakage current prediction model.

[0164] By the above-mentioned explicit switching direction and feature corresponding relationship mode, the pertinence and accuracy of the trend feature transmission are ensured, feature loss or mismatch is avoided, the addition of supplementary features can also make the trend feature list of the target strategy more comprehensive, the adaptability of the leakage current prediction model to complex working conditions is improved, the continuity and integrity of the change trend in the switching process are ensured, and the prediction fault caused by parameter mutation is effectively eliminated.

[0165] On the basis of obtaining the leakage current change trend prediction result of the insulating pull rod device, in order to quantitatively evaluate the insulation state, an insulation integrity scoring mechanism is further introduced to quantitatively express the leakage current trend and the insulation performance degradation risk in real time.

[0166] Specifically, the integrity score of the live working insulation performance according to the current leakage current change trend prediction result and the corresponding leakage current prediction strategy comprises:

[0167] Obtaining the leakage current safety upper limit of the insulating pull rod device in the current live working scene, and obtaining the corresponding prediction time window according to the current leakage current prediction strategy;

[0168] According to the current leakage current change trend prediction result, the leakage current prediction value at each time in the corresponding prediction time window is obtained;

[0169] Based on the leakage current safety upper limit and the leakage current prediction value at each time in the prediction time window, the integrity score of the live working insulation performance is obtained.

[0170] The scoring formula of the above-mentioned integrity score is:

[0171] ;

[0172] Wherein, is the set leakage current safety upper limit; is the leakage current prediction value at the i-th time; H is the length of the prediction time window; and S is the score value.

[0173] In the formula, is used to measure the deviation degree of the prediction value at the i-th time relative to the upper limit value, and the lower the score value S, the higher the leakage risk and the worse the integrity.

[0174] This integrity score is performed in real time in the prediction process, and when the score value S is lower than the set safety threshold, a corresponding safety warning of the insulating pull rod device is issued.

[0175] Embodiment Two: The embodiment of the present application provides a live working insulation performance prediction system fusing multiple information, as shown in Figure 2 illustrated, which can be used to implement the method of embodiment one in real time, comprising:

[0176] The data sensing and collecting module 1 is arranged at the insulating pull rod device 6 and is used for collecting environmental data, stress data and leakage current of the insulating pull rod device in the live working process;

[0177] The data analysis module 2 is connected with the data sensing and collecting module and is used for identifying the operation condition and determining the prediction requirement according to the collected environmental data, stress data and leakage current;

[0178] The decision prediction module 3 is connected with the data sensing and collecting module and the data analysis module respectively and is used for selecting a prediction strategy according to the prediction requirement and predicting the leakage current change trend according to the leakage current;

[0179] The monitoring feedback module 4 is connected with the data analysis module and the decision prediction module respectively and is used for predicting the adjustment requirement according to the leakage current change trend prediction result and the operation condition and controlling the decision prediction module to adjust the applied prediction strategy according to the adjustment requirement;

[0180] The alarm module 5 is connected with the decision prediction module and is used for performing integrity scoring on the insulating performance of the live working according to the leakage current change trend and the prediction strategy and performing safety warning on the insulating pull rod device according to the integrity scoring.

[0181] The data sensing and collecting module includes:

[0182] The temperature and humidity sensor 11 is arranged at the middle outer wall of the insulating pull rod device and is used for collecting the environmental data of the insulating pull rod device in the live working process;

[0183] The stress sensor 12 is arranged at the middle surface of the axis of the insulating pull rod device and the metal end and is used for collecting the stress data of the insulating pull rod device in the live working process;

[0184] The leakage current sensor 13 is arranged at the pull rod end of the insulating pull rod device and is used for collecting the leakage current of the insulating pull rod device in the live working process.

[0185] As a key mechanical support and insulation element in high-voltage electrical equipment, the surface of the insulating pull rod device is easily affected by dust, moisture, salt mist and other pollutants during long-term operation, forming a surface conduction channel, and then causing leakage current, even causing flashover failure. Before these phenomena occur, the leakage current often flows earliest along the pull rod end region of the insulating pull rod device, which is close to the conductive connection end, the electric field edge effect is significant, and the local electric stress is high. At this time, the pull rod end of the insulating pull rod device becomes an important channel for the initial or convergence of leakage current. From the perspective of structure and maintenance, the pull rod end region is relatively open, which is convenient for the installation of sensors and later maintenance, and will not cause mechanical damage to the insulation structure body, which helps to ensure the long-term safe operation of the equipment. Therefore, the leakage current sensor is specifically arranged at the pull rod end of the insulating pull rod device, which can realize early detection and continuous monitoring of the leakage current.

[0186] The temperature and humidity sensor specifically includes a temperature sensor 111 and a humidity sensor 112. The middle region of the insulating pull rod device body is a sensitive region with relatively stable thermal field and thermal coupling, which can represent the overall temperature change trend of the pull rod. This position avoids the high-voltage conductor terminal and the metal connecting piece, which helps to reduce the electric field interference and improve the temperature measurement accuracy, while ensuring the electrical isolation of the sensor and the equipment. Therefore, the temperature sensor is installed at the middle outer wall position of the insulating pull rod device in this embodiment, and through this arrangement, the thermal aging trend can be effectively monitored, providing a key input for subsequent leakage current prediction.

[0187] For the humidity sensor, the monitoring point is set at the middle outer wall close to the ambient temperature monitoring point, which has good local humidity response capability and can form a thermal-humidity coupling monitoring point with the temperature sensor, facilitating the analysis of the influence of humidity and temperature on the leakage current under the synergistic action, and improving the prediction accuracy of the risk of leakage current mutation.

[0188] The stress sensor mainly includes a tensile load sensor 121 for collecting the tensile stress of the insulating pull rod device and a shear load sensor 122 for collecting the shear stress of the insulating pull rod device. Among them, the axial direction region of the middle surface of the insulating pull rod device is the main stress concentration area under the action of tensile load, which can represent the stress state of the pull rod of the entire insulating pull rod device during operation. Therefore, the measurement point of the stress sensor for monitoring tensile stress is set at the axial middle surface of the insulating pull rod device. In addition, the cylindrical surface region of the insulating pull rod device close to the metal connecting end is a significant region of stress concentration and interface coupling effect, which is a precursor site of initial damage such as debonding and slipping, therefore, the measurement point of the stress sensor for monitoring shear stress is further set at the metal end of the insulating pull rod device, which can be specifically selected in the cylindrical surface region of the metal end.

[0189] The installation positions of the sensors in the data sensing and collecting module on the insulating pull rod device are specifically as shown in Figure 3

[0190] In this embodiment, the current mutual inductance sensor with a ring-shaped closed structure is a leakage current sensor, and is fixed and installed in a ring-shaped manner. When installing, the surface of the pull rod end of the insulating pull rod device is cleaned thoroughly to remove oil stains, moisture and conductive contaminants, so as to ensure that the leakage current sensor is closely attached to the surface of the insulator and new discharge hidden dangers are avoided. Secondly, the coil is wrapped around the position of the pull rod end of the insulating pull rod device closest to the conductive connecting piece to form a complete closed inductive loop, and a heat shrink fixing ring is used to reinforce the leakage current sensor to ensure that it does not loosen or deviate in long-term operation. Then, a double-layer heat shrink sleeve is used to package and protect the coil body to improve its moisture-proof, dust-proof and ultraviolet resistance. The signal lead-out wire should use a shielded twisted pair wire, which is led out from the end to the non-electric field area, and is reasonably routed to avoid crossing with the high-voltage part, and finally connected to the data analysis module to transmit the collected data.

[0191] For the temperature and humidity sensor, a flat area 3-5 cm away from the outer surface of the middle part of the insulating pull rod device is first selected, and the area is cleaned thoroughly with anhydrous ethanol and the oxide layer is removed to ensure that the contact surface is free of impurities to ensure good adhesion of the temperature and humidity sensor. Then, to improve the response speed and accuracy of temperature acquisition, a layer of thermal conductive glue is applied to the bottom of the temperature sensor, while ensuring that it does not conduct electricity and does not damage the insulation. Then, the temperature and humidity sensor is fixed in an attached manner on the middle outer wall of the insulating pull rod device, and is wrapped and installed using high-insulation-grade high-temperature polyimide tape to avoid displacement and falling due to mechanical loosening or thermal expansion and contraction. For humidity measurement, since the surface of the insulating pull rod is not easy to directly absorb moisture, a layer of hydrophilic film is wrapped around the humidity sensor to enhance water vapor sensing, and 3.5 mm ventilation slots are reserved on both sides to avoid water vapor accumulation and strong wind interference. When laying the temperature and humidity sensor lead, the signal wire is selected to be a double-layer shielded wire and is kept away from the high-voltage cable, the terminal is treated with silver plating and tin plating, and is sealed with a three-layer heat shrink tube, and an iron-yttrium-magnet ring is added to suppress high-frequency interference.

[0192] ​For the stress sensor, a stress meter is specifically used as a tensile load sensor, and its installation position is selected on the cylindrical surface within the range of ±5 cm from the midpoint of the connecting piece of the insulating pull rod device at both ends of the pull rod, so as to realize the collection of the corresponding tensile stress data. Before installation, the target surface is finely sanded until the uniform and smooth insulating substrate layer is exposed, and then degreasing cleaning is performed with anhydrous ethanol. The tensile load sensor is arranged in one direction along the axis direction of the insulating pull rod device, and is pasted with modified epoxy resin high-strength insulating glue, which is cured for not less than 24 hours under constant pressure, so as to ensure that the strain gauge is tightly attached to the substrate. When the tensile load sensor lead is arranged, the signal line is selected to be a double-layer shielding line and is far away from the high-voltage cable, the terminal is treated by silver plating and tin lining, and then is sealed with a three-layer heat shrink tube, and an iron-yttrium ring is additionally installed to suppress high-frequency interference. After installation is completed, zero point calibration and static load test should be performed to verify the response linearity and repeatability of the tensile load sensor, so as to ensure that it has stable and reliable axial tensile stress detection capability.

[0193] A shear stress sensor is used as a shear load sensor, and its installation position is selected on the cylindrical surface region about 20 to 30 cm away from the metal connecting end near the pull rod of the insulating pull rod device, so as to realize the collection of the corresponding shear stress data. Before installation, the target surface is finely sanded until the uniform and smooth insulating substrate layer is exposed, and then degreasing cleaning is performed with anhydrous ethanol. Installation is performed in a paired arrangement, and the shear load sensor is pasted on the two symmetric positions on the two sides of the pull rod axis of the insulating pull rod device at a cross angle of ±45°, so as to form a strain rosette structure and realize accurate extraction of the shear stress. Pasting is performed with modified epoxy resin high-strength insulating glue, which is cured for not less than 24 hours under constant pressure, so as to ensure that the strain gauge is tightly attached to the substrate. When the shear load sensor lead is arranged, the signal line is selected to be a double-layer shielding line and is far away from the high-voltage cable, the terminal is treated by silver plating and tin lining, and then is sealed with a three-layer heat shrink tube, and an iron-yttrium ring is additionally installed to suppress high-frequency interference. After installation is completed, zero point calibration and static load test should be performed to verify the response linearity and repeatability of the shear load sensor, so as to ensure that it has stable and reliable shear stress detection capability.

[0194] The data analysis module, the decision prediction module and the monitoring feedback module can select high-performance calculators, servers, microcontrollers and other data analysis processing devices, carry relevant algorithms of running condition recognition, prediction demand matching, prediction strategy selection, leakage current change trend prediction, prediction adjustment demand identification and prediction strategy transition switching, and can realize efficient and accurate prediction of the leakage current change trend through the collected environmental data, stress data and leakage current.

[0195] The alarm module can also select high-performance calculators, servers, microprocessors and other data analysis and processing devices, which are internally provided with corresponding integrity score formulas, and can obtain the integrity score of the insulating pull rod device according to the leakage current trend prediction result obtained by the decision prediction module and the specific prediction strategy selected. The alarm module is provided with a corresponding safety threshold, which can trigger a dynamic response according to the safety threshold. When the safety threshold is exceeded, the relevant integrity score and the data applied can be directly connected to the monitoring background of operation and maintenance, and the integrity score of the insulating pull rod device can be visually displayed through the relevant software interface. When the safety threshold is lower, the corresponding hierarchical alarm can be activated, and the yellow or red warning can be triggered according to the score difference below the safety threshold. If the score difference below the safety threshold is low, the yellow warning is triggered, the maintenance work order containing GIS positioning is generated and pushed to the mobile terminal of the operation and maintenance personnel, and the relevant data is further fed back to the decision prediction module to start the incremental training of the relevant leakage current prediction model to optimize the prediction accuracy. If the score difference below the safety threshold is high, the corresponding hard point interlocking trip of the insulating pull rod device is directly executed to cut off the circuit, the sound and light alarm is triggered synchronously, and the relevant time axis is recorded.

[0196] Embodiment three: the embodiment of the present application provides a live working insulation performance prediction device fusing multiple information, comprising a processor and a storage medium;

[0197] The storage medium is used for storing instructions;

[0198] The processor is used for operating according to the instructions to execute the steps of any method according to embodiment one.

[0199] Embodiment four: the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of any method according to embodiment one.

[0200] The above-mentioned embodiment is only a preferred scheme of the present application, and does not limit the present application in any form. Other variants and modifications can be made without exceeding the technical scheme recorded in the claims.

Claims

1. A method for predicting the insulation performance of live-line work by integrating multiple information sources, characterized in that, include: Collect environmental data, stress data, and leakage current of the insulated tie rod equipment during live-line operation; Based on environmental and stress data, identify the current operating conditions and determine the predicted demand according to the operating conditions. Based on leakage current, either a short-term recursive prediction strategy or a long-term multi-dimensional input-output prediction strategy is selected according to the prediction requirements to predict the trend of leakage current change. Based on the leakage current change trend prediction results and operating conditions, the prediction adjustment needs are identified in real time, and the prediction strategy is switched gradually according to the prediction adjustment needs, and the leakage current change trend prediction results are updated. Based on the current leakage current trend prediction results and the corresponding leakage current prediction strategy, the integrity score of the insulation performance of live-line work is performed. Safety warnings for insulated tie rod equipment are issued based on integrity scoring results. The method of predicting leakage current trends by selecting either a short-term recursive prediction strategy or a long-term multi-dimensional input-output prediction strategy based on the prediction requirements includes: When the operating condition type is a short-term extreme operating condition and the duration is less than the preset time threshold, the predicted demand is a short-term high-frequency predicted demand. Select a short-term recursive prediction strategy based on the short-term high-frequency prediction requirements, and set the leakage current prediction model parameters and recursive parameters according to the corresponding operating conditions. Using the collected leakage current as the initial input, the leakage current value at a future time point is predicted through the leakage current prediction model, and the predicted leakage current value is fed back to the leakage current prediction model as a new input. Repeatedly perform the prediction of leakage current value and new input feedback to generate a continuous short-term prediction sequence; Short-term forecast demand generated recursively by sliding through a fixed window size, and the trend characteristics within the window are calculated. Based on trend characteristics and preset rules, the leakage current change trend prediction results are obtained; The step of transitioning the prediction strategy based on the predicted adjustment needs and updating the leakage current change trend prediction results includes: The operating condition fluctuation coefficient is obtained based on environmental and stress data, and a corresponding transition window is set according to the operating condition fluctuation coefficient. Within the transition window, the predicted values ​​of the current prediction strategy and the target prediction strategy to be switched are output synchronously, and the corresponding prediction fusion value is obtained by combining the dynamic weights. When the predicted fusion value and dynamic weight meet the transition termination condition, the transition window is terminated, and the leakage current change trend is predicted according to the target prediction strategy.

2. The method for predicting the insulation performance of live-line work by integrating multiple information sources as described in claim 1, characterized in that, The process of identifying the current operating condition based on environmental and stress data, and determining the predicted demand based on the operating condition, includes: Environmental and stress data are preprocessed, and working condition features are extracted from the preprocessed environmental and stress data. Based on operating condition characteristics and corresponding operating condition classification rules, the current operating condition is identified. Obtain the duration and type of the current operating condition, and match the corresponding forecast requirements.

3. The method for predicting the insulation performance of live-line work by integrating multiple information sources according to claim 2, characterized in that, The method of predicting leakage current trends by selecting a short-term recursive prediction strategy or a long-term multi-input-output prediction strategy based on the prediction requirements, according to the leakage current, also includes: When the operating condition type is a long-term stable operating condition and the duration is not less than a preset time threshold, the predicted demand is a long-term trend predicted demand. When selecting a long-term multidimensional input-output prediction strategy based on long-term trend prediction requirements, the leakage current prediction model parameters should be set according to the corresponding operating conditions. Input data samples are constructed based on leakage current and its influencing factors, and a continuous long-term prediction sequence is output by combining the leakage current prediction model. The long-term forecast sequence and the collected environmental and stress data are divided according to time granularity. The trend characteristics of the long-term forecast sequence and the trend characteristics of the environmental and stress data in each time period are calculated, and corresponding trend category labels are added respectively. Based on trend characteristics and corresponding trend category labels, the causes of changing trends are identified, and the prediction results of leakage current changing trends are obtained.

4. The method for predicting the insulation performance of live-line work by integrating multiple information sources according to claim 1, characterized in that, The process of identifying and predicting adjustment needs in real time based on the leakage current change trend prediction results and operating conditions includes: Identify the current operating conditions and their duration, and obtain the matching degree of the forecasting strategy in combination with the current forecasting requirements; Obtain the operating characteristics of the current operating conditions, and combine the leakage current change trend prediction results to obtain the operating condition fitness of the prediction strategy; Based on the measured values ​​of the current leakage current, obtain the trend consistency index of the predicted leakage current change trend; Based on matching degree, working condition adaptability and trend consistency index, the need for predicted adjustment is identified.

5. The method for predicting the insulation performance of live-line work by integrating multiple information sources according to claim 1, characterized in that, During the termination transition window, when predicting the leakage current change trend according to the target prediction strategy, the following is also performed: Based on the demand type of the predicted adjustment demand, obtain the transition target, and select the transition direction of the prediction strategy based on the transition target; Based on the direction of the transition of the prediction strategy, establish the correspondence between the trend characteristics of the current prediction strategy and the target strategy, and obtain the list of trend characteristics of the target strategy. Extract supplementary features from the current prediction strategy and add them to the trend feature list of the target strategy; The leakage current change trend prediction results are updated based on the trend features in the trend feature list.

6. The method for predicting the insulation performance of live-line work by integrating multiple information sources according to claim 1, characterized in that, The process of evaluating the integrity of insulation performance for live-line work based on the current leakage current trend prediction results and corresponding leakage current prediction strategies includes: Obtain the safe upper limit of leakage current for the insulated tie rod equipment under the current live working scenario, and obtain the corresponding prediction time window based on the current leakage current prediction strategy; Based on the current leakage current change trend prediction results, obtain the leakage current prediction value at each moment in the corresponding prediction time window; Based on the upper limit of leakage current safety and the predicted leakage current value at each moment in the prediction time window, the integrity score of the insulation performance of live-line work is obtained.

7. A live-line working insulation performance prediction system integrating multiple information sources, used to execute the prediction method according to any one of claims 1 to 6, characterized in that, include: The data sensing and acquisition module is installed at the insulating tie rod equipment to collect environmental data, stress data and leakage current of the insulating tie rod equipment during live-line operation. The data analysis module, connected to the data sensing and acquisition module, is used to identify operating conditions and determine predicted requirements based on the acquired environmental data, stress data, and leakage current. The decision prediction module is connected to the data sensing and acquisition module and the data analysis module, respectively. It is used to select the prediction strategy according to the prediction requirements and predict the trend of leakage current change based on the leakage current. The monitoring and feedback module is connected to the data analysis module and the decision prediction module respectively. It is used to identify the prediction adjustment needs based on the prediction results of the leakage current change trend and the operating conditions, and to control the prediction strategy of the decision prediction module to adjust the application over time according to the prediction adjustment needs. The alarm module, connected to the decision prediction module, is used to score the integrity of the insulation performance of live-line work based on the leakage current change trend and prediction strategy, and to provide safety warnings for the insulated tie rod equipment based on the integrity score.

8. The live-line working insulation performance prediction system integrating multiple information sources according to claim 7, characterized in that, The data sensing and acquisition module includes: A temperature and humidity sensor is installed on the outer wall of the middle part of the insulated tie rod equipment to collect environmental data of the insulated tie rod equipment during live-line operation. Stress sensors are installed on the central surface of the axis of the insulating tie rod and at its metal end to collect stress data of the insulating tie rod during live-line work. A leakage current sensor is installed at the end of the tie rod of the insulated tie rod equipment to collect the leakage current of the insulated tie rod equipment during live-line operation.

9. A live-line working insulation performance prediction device integrating multiple information sources, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the live-line working insulation performance prediction method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for predicting the insulation performance of live-line work as described in any one of claims 1 to 6.

Images