Oil-immersed bushing monitoring method and system based on random forest model and medium

Abstract

A method, system, and medium for monitoring oil-immersed bushings based on a random forest model are disclosed, relating to the field of computer systems based on specific computational models. The method includes: acquiring historical monitoring data of a target bushing system within a preset time period; constructing a random forest model based on the historical monitoring data; acquiring real-time monitoring data of the target bushing system and uploading it to a power system; receiving operational monitoring data of multiple interconnected bushing systems sent by the power system; constructing a test sample set based on the real-time monitoring data and the operational monitoring data; inputting the test sample set into the random forest model to obtain state assessment results for the multiple interconnected bushing systems and uploading them to the power system; acquiring the target assessment result of the target bushing system output by the power system after receiving all state assessment results; and generating a state prompt message for the target bushing system based on the target assessment result. Implementing this application can improve the accuracy of oil-immersed bushing state monitoring.

Description

Technical Field

[0001] This application relates to the field of computer systems based on specific computational models, and in particular to a method, system and medium for monitoring oil-immersed casing based on a random forest model. Background Technology

[0002] Oil-immersed bushings are key components of power equipment such as transformers and reactors, and their operating status directly affects equipment safety. A bushing consists of the bushing body (including paper insulation and insulating oil), the insulating oil chamber, and the high-voltage terminal conductor, and undergoes changes in various physical quantities such as electricity, heat, and gas during operation. With the expansion of power grids and the increase in load levels, bushings face more complex operating environments, placing higher demands on their condition monitoring.

[0003] In related technologies, monitoring systems typically install various sensors on the bushing to collect parameters such as dielectric loss, capacity, and partial discharge. The collected data is then input into pre-trained fault diagnosis models for analysis. These models are usually trained based on historical operating data and fault cases, and determine whether the equipment is in an abnormal state by setting fixed thresholds. In practical applications, the monitoring system periodically collects data and compares it with preset thresholds. When the monitored parameters exceed the thresholds, the system issues an alarm signal.

[0004] However, with the improvement of equipment manufacturing processes and the advancement of equipment maintenance, the probability of failure of the new generation of oil-immersed bushings has decreased, and the number of failure samples is seriously insufficient. Therefore, the failure samples used for model training in related technologies are mostly derived from experimental simulations. However, there are significant differences between the equipment operating environment and the experimental simulation environment, and the condition monitoring of oil-immersed bushings in related technologies is not accurate enough. Summary of the Invention

[0005] This application provides a method, system, and medium for monitoring oil-immersed casing based on a random forest model, which can improve the accuracy of oil-immersed casing condition monitoring.

[0006] Firstly, this application provides a method for monitoring oil-immersed bushings based on a random forest model, applied to a monitoring system. The method includes: acquiring historical monitoring data of a target bushing system within a preset time period, and constructing a random forest model based on the historical monitoring data; acquiring real-time monitoring data of the target bushing system and uploading it to the power system; receiving operational monitoring data of multiple interconnected bushing systems sent by the power system, and constructing a test sample set based on the real-time monitoring data and the operational monitoring data; inputting the test sample set into the random forest model to obtain the state assessment results of the target bushing system on multiple interconnected bushing systems, and uploading the state assessment results to the power system; acquiring the target assessment result of the target bushing system output by the power system after receiving all the state assessment results; and generating a status prompt message for the target bushing system based on the target assessment result.

[0007] In the above embodiments, the monitoring system establishes a random forest model based on historical data of the target casing system and performs cross-evaluation by combining real-time operational data from multiple interconnected casing systems, thus achieving multi-dimensional status monitoring. It not only utilizes the historical performance of individual casings but also integrates the collective behavioral characteristics of casings in the same batch, making it not limited to its own model. This improves the accuracy and reliability of status assessment and effectively solves the problem of insufficient accuracy in traditional single-threshold judgment methods.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of acquiring real-time monitoring data of the target bushing system and uploading it to the power system specifically includes: acquiring the operating environment parameters of the target bushing system and determining the data correction coefficient of the operating environment parameters relative to the standard environment parameters; correcting the monitoring data of the target bushing system based on the data correction coefficient to obtain real-time monitoring data, and uploading the real-time monitoring data to the power system.

[0009] In the above embodiments, the monitoring system introduces an environmental parameter correction mechanism, which enables the system to dynamically adjust the correction coefficient of the monitoring data according to the difference between the actual operating environment and the standard environment. This ensures the comparability and accuracy of the monitoring data under different environmental conditions, avoids the interference of environmental factors on the evaluation results, and improves the adaptability of condition monitoring.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, before the step of acquiring historical monitoring data of the target bushing system within a preset time period and constructing a random forest model based on the historical monitoring data, the method further includes: when the cumulative running time of the target bushing system reaches a preset running time, acquiring the bushing body parameters, operating environment parameters, and operating status parameters of the target bushing in the target bushing system; calculating the deviation direction and deviation index of the target bushing relative to a standard bushing model based on the bushing body parameters, operating environment parameters, and operating status parameters; the standard bushing model is a reference model under ideal operating conditions constructed based on historical operating data of bushings of the same insulation level of the target bushing and expert evaluation rules; calculating the operating correlation degree between the target bushing and other bushings in the power system based on the deviation direction and deviation index, determining the bushing system where the operating correlation degree is within a preset correlation threshold, and identifying the bushing system that has reached the preset running time as an interconnected bushing system.

[0011] In the above embodiments, the monitoring system constructs a standard casing model as a reference benchmark for ideal operating conditions based on casing body parameters, operating environment parameters, and operating status parameters. By calculating the deviation of the target casing from the standard model and its operational correlation with other casings, a correlation network between the casing systems is established. This correlation analysis method based on multi-dimensional parameters ensures the accuracy of the interconnected casing systems.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, before calculating the deviation direction and deviation index of the target casing relative to the standard casing model based on casing body parameters, operating environment parameters, and operating state parameters, the method further includes: when the cumulative running time of the target casing system has not reached the preset running time, acquiring the factory test data and early operating data of the target casing system; constructing a mapping relationship between the factory test data and the early operating data to obtain a systematic deviation; constructing a compensation factor based on the systematic deviation and operating environment parameters to obtain a backup evaluation model; and performing a state evaluation of the target casing system based on the backup evaluation model until the cumulative running time reaches the preset running time.

[0013] In the above embodiments, the monitoring system establishes a backup evaluation model based on factory test data and early operating data at the initial stage of equipment operation. Through systematic deviation analysis and compensation factor construction, it achieves effective evaluation of bushings with insufficient operating time, solves the problem of lack of historical data for newly commissioned equipment, and ensures the continuity and integrity of condition monitoring.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining the target assessment result of the target bushing system output by the power system after receiving all status assessment results, the method further includes: statistically analyzing the percentage of bushing systems judged to be in a warning state among multiple interconnected bushing systems and the target bushing system; when the percentage exceeds the collective fault judgment threshold, correcting the status prompt information of the target bushing system to a collective fault warning.

[0015] In the above embodiments, by introducing a collective failure determination mechanism, the monitoring system can identify and warn of potential batch problems, effectively improving the early warning capability for systemic hidden dangers and providing an important basis for equipment management decisions.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining historical monitoring data of the target casing system within a preset time period and constructing a random forest model based on the historical monitoring data, the method further includes: adjusting the preset time period to the latest time period after the target casing system's running time enters the next data management cycle; obtaining the latest monitoring data of the target casing system within the latest time period, and retraining the random forest model based on the latest monitoring data.

[0017] In the above embodiments, the monitoring system establishes a data management cycle mechanism to realize the dynamic updating of the random forest model, and retrains the model regularly with the latest data, ensuring the model's adaptability to the current state of the equipment and improving the timeliness and accuracy of the evaluation results.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining the latest monitoring data of the target casing system within the latest time period and retraining the random forest model based on the latest monitoring data, the method further includes: dividing an independent validation dataset from the latest monitoring data; the validation dataset does not participate in the retraining of the random forest model; evaluating the validation dataset using the unupdated random forest model and the updated random forest model respectively to obtain a pre-performance index and a post-performance index; and deploying the updated random forest model to the monitoring system when the post-performance index is better than the pre-performance index.

[0019] In the above embodiments, the monitoring system ensures the effectiveness of model updates by introducing an independent validation dataset and a performance indicator comparison mechanism, avoiding performance degradation that may result from blind updates, and guaranteeing the stability and reliability of the monitoring system.

[0020] In a second aspect, embodiments of this application provide a monitoring system comprising: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, the computer program code including computer instructions, wherein the one or more processors invoke the computer instructions to cause the monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a monitoring system, cause the monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a monitoring system, cause the monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the monitoring system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. By adopting a multi-dimensional evaluation scheme based on a random forest model, which integrates the dual mechanisms of historical data training and cross-validation of multiple bushing systems, it can fully utilize the historical performance of individual bushings while effectively integrating the collective behavioral characteristics of multiple interconnected bushings to form a more comprehensive evaluation system. This effectively solves the problems of insufficient evaluation accuracy and high false alarm rate caused by relying solely on single device data and fixed thresholds in existing technologies. As a result, it achieves more accurate and reliable bushing status monitoring, improves the accuracy of early warnings, reduces the false alarm rate, and provides a more reliable basis for equipment operation and maintenance decisions.

[0026] 2. By employing a multi-dimensional parameter-based casing correlation analysis method, a standard casing model is constructed by combining casing body parameters, operating environment parameters, and operating status parameters. The operational correlation between different casings is calculated through deviation direction and deviation index. Therefore, it is possible to scientifically identify and screen casing systems with similar operating characteristics, establish a more reasonable interconnection evaluation network, and effectively solve the problems of insufficient scientific basis for interconnection casing selection and inadequate correlation analysis in existing technologies. This leads to more accurate group status assessment, improves the early warning accuracy of batch failures, and enhances the reliability of the monitoring system.

[0027] 3. Due to the adoption of a dynamic model update mechanism based on data management cycles, the evaluation model is continuously optimized by periodically acquiring the latest monitoring data and retraining the random forest model. Therefore, it can adapt to the dynamic changes in equipment status in a timely manner, maintain the timeliness of the model, and effectively solve the problems of poor adaptability and the decay of evaluation accuracy over time caused by model solidification in the existing technology. This enables the continuous and effective operation of the monitoring system and ensures the long-term reliability of the status evaluation results. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an oil-immersed casing monitoring method based on a random forest model in an embodiment of this application.

[0029] Figure 2 This is another flowchart illustrating the oil-immersed casing monitoring method based on the random forest model in this application embodiment;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a monitoring system in an embodiment of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] A batch of new oil-immersed bushings was put into operation at a 500kV substation. These bushings employed an improved paper insulation structure and a new type of insulating oil, theoretically offering higher reliability. However, after six months of operation, the monitoring system detected slight anomalies in the dielectric loss values ​​of several bushings. Due to the new type of bushings and the lack of corresponding fault samples, traditional monitoring models based on historical data struggled to accurately determine the equipment's condition. Ultimately, this led to two bushings experiencing breakdown failures due to accelerated insulation aging, causing the transformer to shut down. Further investigation revealed that the same batch of insulating paper was used in the manufacturing process, indicating a potential quality problem that the existing monitoring system was unable to detect in a timely manner.

[0035] In related technologies, the operating status of casing is determined by using a single device for independent monitoring, training a model based on laboratory simulation data, and setting fixed thresholds. The following describes a scenario using a random forest-based monitoring method for oil-immersed casing.

[0036] A power company uses a traditional bushing monitoring system, which primarily relies on laboratory-simulated fault data to train diagnostic models. After a batch of imported oil-immersed bushings was put into service, these bushings were equipped with online monitoring devices for dielectric loss, capacity, and partial discharge. However, due to significant differences between the actual operating environment (such as temperature variations, load fluctuations, and pollution levels) and the laboratory environment, the monitoring system frequently generated false alarms. To avoid these false alarms, management adjusted the corresponding alarm thresholds, resulting in multiple bushings from the same batch experiencing insulation degradation within a short period, which was not detected and addressed in a timely manner.

[0037] The oil-immersed casing monitoring method based on the random forest model in this application embodiment achieves group analysis and accurate evaluation of equipment in the same batch by establishing an association model of the interconnected casing system.

[0038] After adopting the monitoring method described in this application, a power grid company detected a slight fluctuation in the dielectric loss value of one of a batch of newly commissioned oil-immersed bushings. The system first conducted a preliminary assessment of the bushing, finding that although the parameter changes did not exceed traditional thresholds, they exhibited an abnormal trend. The system then initiated a batch correlation analysis process: first, it acquired real-time monitoring data for all bushings in the same manufacturing batch, including parameters such as dielectric loss, capacity, and partial discharge; then, it constructed an evaluation model based on the random forest algorithm, using data from bushings with stable operating conditions in the batch as a control sample, and inputting data from suspicious bushings into the model for predictive analysis, thus promptly identifying bushings that had deviated from normal operating ranges.

[0039] As can be seen, the oil-immersed casing monitoring method based on the random forest model in this application embodiment can not only accurately assess the casing status, but also effectively solve problems such as the difference between laboratory data and actual operation and batch correlation analysis, thereby realizing early warning and prevention of batch failures.

[0040] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating an oil-immersed casing monitoring method based on a random forest model in an embodiment of this application.

[0041] S101. Obtain historical monitoring data of the target casing system within a preset time period, and construct a random forest model based on the historical monitoring data.

[0042] The target casing system refers to the overall system consisting of a specific oil-immersed casing and its related equipment that requires condition monitoring; the preset time period refers to the historical data collection cycle determined in advance based on the equipment operating characteristics and data analysis needs, usually 3-6 months; historical monitoring data is used to represent various monitoring parameters collected within the preset time period, including but not limited to dielectric loss values, capacity values, partial discharge, oil temperature, ambient temperature, etc.; the random forest model refers to an ensemble learning model composed of multiple decision trees, used to achieve a comprehensive assessment of the casing condition.

[0043] After initial configuration, the monitoring system needs to establish a basic evaluation model. Specifically, the system first determines the start and end times of a preset time period, then extracts all monitoring data for the target casing system within that time period from the database, including both routine and special operating condition data. Next, the system preprocesses this data, including outlier handling, missing value imputation, and data standardization. Finally, the system uses the processed data to train a random forest model, including feature selection, parameter optimization, and model validation.

[0044] In some embodiments, the acquisition of historical monitoring data and the model building process can be implemented in several ways: Optionally, the monitoring system can collect data periodically, setting a fixed data sampling interval, acquiring data from sensors at regular intervals, storing the data in a local database after data cleaning and feature extraction, and then dynamically adjusting the number and depth parameters of decision trees based on data distribution characteristics to ultimately build a highly adaptive random forest model; Optionally, the monitoring system can adopt an event-triggered approach, increasing the sampling frequency and recording the complete parameter change process when changes in key monitoring parameters are detected, then setting feature weights based on expert experience, and building a targeted evaluation model through a weighted random forest algorithm. It is understood that other methods can also be used to achieve historical data acquisition and model building, which are not limited here.

[0045] During implementation, the issue of uneven historical data quality may arise, meaning that data quality is poor or there are significant gaps in certain time periods. To address this, the monitoring system can employ a data quality assessment and compensation mechanism: First, the data is segmented for quality assessment, calculating the data completeness and reliability indices for each time window. For data segments with poor quality, data reconstruction is performed by combining the physical model and data characteristics from adjacent time periods. Simultaneously, data quality weights are introduced during the training of the random forest model, making the model more reliant on the characteristics of high-quality data segments. For example, if 30% of the data for a certain month is missing, targeted supplementation can be performed using historical data from similar operating conditions based on the month's load characteristics and environmental conditions.

[0046] It should be noted that although the random forest model used in this application has some data processing complexity, it achieves a stable improvement in the accuracy of data monitoring. Specifically, due to the complexity of the failure modes of the bushing system, involving the coupling of multiple physical processes such as electrical, thermal, and chemical processes, it is difficult to accurately capture various potential fault characteristics using a simple single threshold judgment method. The random forest model, however, has excellent parallel computing characteristics; multiple decision trees can be trained and predicted simultaneously, resulting in high efficiency on modern computing hardware. This model has an inherent feature importance evaluation mechanism that can identify and filter key monitoring parameters, reducing the workload of manual feature engineering. The prediction process of the model is essentially a voting statistics of multiple simple decision trees, which is parallel computing with very low computational overhead, fully meeting the periodic inspection requirements of oil-immersed bushings. Therefore, the use of the random forest model in this application results in less waste of computing resources, and through its efficient parallel computing characteristics and feature processing capabilities, it can improve the efficiency and accuracy of condition monitoring.

[0047] Furthermore, the interconnected bushing evaluation mechanism established in this application is optimized. The system only performs phased group analysis when necessary, while daily real-time monitoring still mainly relies on the local model. Only when an abnormal trend is detected will the collaborative evaluation of interconnected bushings be triggered. This on-demand analysis strategy ensures comprehensive monitoring while avoiding unnecessary computational overhead. Therefore, the technical solution provided in this application achieves accurate monitoring while maintaining high computational efficiency.

[0048] S102. Obtain real-time monitoring data of the target bushing system and upload it to the power system.

[0049] Among them, real-time monitoring data represents various parameter indicators of the target bushing system under its current operating status; power system refers to the central control platform used to manage and coordinate multiple bushing systems; data upload refers to the process of transmitting locally collected monitoring data to the power system for centralized processing; operating environment parameters refer to external factors that affect the operating status of the bushing, including ambient temperature, humidity, load rate, etc.

[0050] After constructing the random forest model, the monitoring system needs to continuously acquire real-time data for state assessment. Specifically, the monitoring system first collects operating parameters of the bushing system in real time using various sensors, including electrical parameters (such as dielectric loss and capacity), physical parameters (such as temperature and pressure), and chemical parameters (such as gas content in oil). Then, the monitoring system performs real-time verification and preprocessing of the collected data to ensure its validity. Next, the monitoring system encapsulates the processed data according to a predefined data format and transmits the data to the power system via a secure communication protocol. During transmission, the monitoring system records the data's timestamp and source information to ensure data traceability.

[0051] In some embodiments, the real-time data acquisition and uploading process can be implemented in multiple ways: Optionally, the monitoring system can adopt a hierarchical acquisition strategy, setting different sampling frequencies according to the importance and variation characteristics of different parameters, sampling key parameters at high frequencies and stable parameters at low frequencies, then integrating data from different frequencies through a data fusion algorithm, and finally uploading them to the power system in batches; Optionally, the monitoring system can implement an adaptive sampling scheme, increasing the sampling frequency when abnormal trends are detected by analyzing parameter change trends in real time, and simultaneously triggering a real-time data upload mechanism to ensure data real-time performance under abnormal conditions. It is understood that other methods can also be used to achieve real-time data acquisition and uploading, which are not limited here.

[0052] During implementation, data transmission interruptions or delays may occur, potentially hindering the power system from obtaining complete monitoring data in a timely manner. To address this, the monitoring system can employ a data caching and retransmission mechanism: a local data buffer is set up, and when a network anomaly is detected, the collected data is temporarily stored in the buffer; simultaneously, a backup communication channel is activated, using backup links such as 4G / 5G mobile networks to ensure data transmission; once the main communication link is restored, the monitoring system checks data integrity and retransmits the cached data to the power system in chronological order. For example, when fiber optic communication is interrupted, the system can switch to wireless communication mode to ensure continuous data transmission.

[0053] S103. Receive operation monitoring data of multiple interconnected bushing systems sent by the power system, and construct a test sample set based on real-time monitoring data and operation monitoring data.

[0054] Among them, the interconnected bushing system refers to the set of other bushing systems that have the same production batch or operating characteristics as the target bushing system; the operation monitoring data refers to various operating parameters collected in real time from the interconnected bushing system; the test sample set is used to represent the data set formed by integrating the monitoring data of the target bushing system and the interconnected bushing system according to specific rules.

[0055] After acquiring local real-time data, the monitoring system needs to perform collaborative analysis with data from other bushing systems. Specifically, the monitoring system first receives interconnected bushing operation data pushed by the power system, including real-time parameters and historical trends of each bushing system. Then, the monitoring system performs format standardization and time synchronization processing on the received data to ensure the comparability of data from different sources. Next, the monitoring system performs feature-level correlation analysis between the local real-time monitoring data and the received operational monitoring data, extracting common and dissimilar features. Finally, according to preset data organization rules, the monitoring system integrates the processed feature data into a standardized set of test samples.

[0056] In some embodiments, the reception of operational monitoring data and the construction of sample sets can be achieved in multiple ways: Optionally, the monitoring system can adopt a hierarchical feature extraction strategy, firstly performing independent feature engineering on the raw data of each casing system to extract time-domain features, frequency-domain features, and statistical features, then constructing a feature correlation matrix, selecting the most representative feature combination based on the correlation strength, and finally generating a structured sample set to be tested; Optionally, the monitoring system can implement a dynamic weight allocation mechanism, assigning different weight coefficients to data from different sources based on factors such as the runtime and historical reliability of each casing system, and constructing a sample set to be tested with credibility indicators through weighted combination. It is understood that other methods can also be used to achieve data reception and sample set construction, which are not limited here.

[0057] During implementation, inconsistencies in the data feature distributions of different bushing systems can arise, affecting the quality of the sample set and the accuracy of subsequent evaluations. To address this, the monitoring system can employ data standardization and distribution alignment mechanisms: First, multi-dimensional statistical analysis is performed on the data from each bushing system to identify differences in data distribution; then, based on physical models and expert knowledge, a suitable standardization transformation function is designed to map data with different distributions to a unified feature space; simultaneously, a distribution alignment algorithm is introduced to achieve a smooth transition of features by minimizing distribution differences between data from different sources. For example, when the temperature data of a certain interconnected bushing exhibits significantly different distribution characteristics compared to other bushings, the system will make targeted distribution adjustments based on environmental factors and load characteristics.

[0058] S104. Input the sample set to be tested into the random forest model to obtain the state assessment results of the target bushing system for multiple interconnected bushing systems, and upload the state assessment results to the power system.

[0059] Among them, the state assessment result represents the quantitative analysis output of the bushing system's operating state through the random forest model; the model input refers to the process of sending the standardized test sample set into the random forest model for calculation according to a predetermined format; and the data upload is used to represent the transmission of the assessment results to the power system for centralized processing.

[0060] After constructing the test sample set, the monitoring system needs to perform state assessment using a trained random forest model. Specifically, the monitoring system first preprocesses the test sample set to ensure the data format matches the model input requirements. Then, the system sequentially inputs the processed sample data into each decision tree of the random forest model to obtain the evaluation results of each decision tree. Next, based on a pre-defined ensemble strategy, the monitoring system integrates the output results of all decision trees to calculate the final state assessment result. Finally, the monitoring system packages and uploads the assessment results and their reliability indicators to the power system, while simultaneously saving local assessment records.

[0061] In some embodiments, the model evaluation and result uploading process can be implemented in multiple ways: Optionally, the monitoring system can adopt a multi-threshold hierarchical evaluation strategy, setting multiple warning level thresholds for different types of abnormal states, calculating the probability distribution of samples at each level using a random forest model, comprehensively obtaining evaluation results with confidence intervals, and selecting different uploading priorities based on the severity of the evaluation results; Optionally, the monitoring system can implement a dynamic correction mechanism for the evaluation results, dynamically adjusting the model's decision weights by comparing the deviation between historical evaluation results and actual operating conditions, improving the accuracy of the evaluation, and uploading the corrected results to the power system in real time. It is understood that other methods can also be used to implement state evaluation and result uploading, which are not limited here.

[0062] During implementation, the instability of model evaluation results may be encountered, meaning that similar input samples can produce highly fluctuating evaluation results. To address this, the monitoring system can employ a stability enhancement mechanism for the evaluation results: first, a time window smoothing process is introduced, weighted averaging the evaluation results across multiple consecutive time points; then, a rate-of-change limit is set for the evaluation results, triggering secondary validation when the change in evaluation results at adjacent time points exceeds the threshold; simultaneously, by analyzing the key features causing the fluctuations, the feature weights are dynamically adjusted to suppress interference from non-critical factors. For example, when frequent fluctuations in evaluation results are detected within a short period, the system will extend the evaluation period, using data analysis over a longer time window to improve the stability of the results.

[0063] S105. Obtain the target assessment result of the target bushing system output by the power system after receiving all the status assessment results.

[0064] Among them, the target assessment result represents the final state determination of the target bushing system based on the cross-assessment results of all interconnected bushing systems; all state assessment results refer to the complete assessment dataset generated by the mutual assessment of the target bushing system and all interconnected bushing systems; cross-validation is used to represent the assessment process of mutual verification between different bushing systems.

[0065] After uploading its local assessment results, the monitoring system waits for the power system to complete its global assessment. Specifically, the monitoring system first confirms that the power system has received assessment results from all interconnected bushing systems, including local assessment results and feedback from other bushing systems. Then, the monitoring system receives the target assessment results generated by the power system based on a cross-validation mechanism, which comprehensively considers the assessment opinions of all interconnected bushings. Next, the monitoring system compares and analyzes the received target assessment results with the local assessment results to verify their consistency. Finally, the monitoring system stores the verified target assessment results in its local database as the basis for subsequent status analysis.

[0066] In some embodiments, the acquisition and processing of target assessment results can be achieved in multiple ways: Optionally, the monitoring system can adopt an active polling strategy, periodically sending query requests to the power system to check whether the global assessment has been completed. Once the assessment is confirmed to be complete, the target assessment results are immediately acquired, and the reasonableness of the results is verified through a local assessment model to ensure the reliability of the assessment results. Optionally, the monitoring system can implement a hierarchical processing mechanism for assessment results, setting different processing priorities according to the urgency of the target assessment results, and initiating a rapid response process for high-risk assessment results to ensure that necessary protective measures are taken in a timely manner. It is understood that other methods can also be used to acquire and process assessment results, which are not limited here.

[0067] During implementation, discrepancies may arise between the target evaluation results and the local evaluation results, affecting the accuracy of subsequent decisions. The monitoring system can employ an evaluation result consistency processing mechanism: first, it calculates the degree of deviation between the local evaluation results and the target evaluation results; when the deviation exceeds a preset threshold, a deep analysis process is triggered. Then, by analyzing the key factors causing the deviation, including differences in data characteristics and changes in model parameters, the root causes of the evaluation discrepancies are identified. Simultaneously, based on the analysis results, the local evaluation model is specifically optimized to improve consistency with the global evaluation. For example, if the local evaluation results are found to be overly conservative while the target evaluation results are more reasonable, the system will appropriately adjust the evaluation criteria of the local model.

[0068] S106. Based on the target assessment results, generate status prompt information for the target casing system.

[0069] Among them, status prompt information refers to warning, suggestion or explanatory information about the operating status of the bushing system generated based on the target assessment results; information generation rules refer to the set of preset rules that transform the assessment results into specific prompt information; information display method is used to represent the presentation form and transmission path of different types of prompt information.

[0070] After acquiring the target assessment results, the monitoring system needs to generate easily understandable and actionable status alerts. Specifically, the monitoring system first determines the current operating status category of the bushing system based on the numerical range and trend of the target assessment results. Then, based on preset information generation rules, the monitoring system transforms the assessment results into specific alert content, including status description, risk level, and handling suggestions. Next, the monitoring system selects the appropriate information display method and push channel according to the urgency of the alert information. Finally, the monitoring system sends the generated alert information to the relevant responsible persons and saves the alert records locally.

[0071] In some embodiments, status alert information can be generated and published in multiple ways: Optionally, the monitoring system can employ a multi-dimensional information fusion strategy, combining assessment results with historical operational data, maintenance records, and expert experience to generate comprehensive alert information that includes fault cause analysis, development trend prediction, and handling solution suggestions, visually displaying the characteristics and impact of abnormal states through a combination of text and graphics; Optionally, the monitoring system can implement an intelligent alert level adjustment mechanism, dynamically optimizing the alert level classification criteria and triggering conditions by analyzing operators' responses to historical alert information, ensuring the practicality and operability of the alert information. It is understood that other methods can also be used to generate and publish status alert information, and this is not limited here.

[0072] During implementation, an excessive number of notifications can distract operators and hinder the timely processing of critical alerts. To address this, the monitoring system can employ an intelligent notification filtering mechanism: First, a priority evaluation model is established, comprehensively considering factors such as the severity of the anomaly, its development speed, and potential impact, assigning dynamic weights to each notification. Then, based on system operating status and operator capabilities, reasonable notification frequency and quantity limits are set, merging or delaying the delivery of low-priority notifications. Simultaneously, by analyzing operators' work patterns and processing habits, the timing and display of notifications are optimized. For example, when the system detects multiple related minor anomalies, these anomalies are consolidated into a single comprehensive notification, avoiding the generation of too many fragmented notifications.

[0073] The above embodiments (steps S101-S106) provide a core monitoring framework based on a random forest model and cross-evaluation of interconnected bushings. However, in complex real-world application scenarios, the implementation of this framework may face further challenges. For example, it is necessary to scientifically define and select interconnected bushing systems to ensure the effectiveness of group evaluation; for newly commissioned bushings lacking sufficient historical monitoring data, accurate early status assessment is required; as equipment ages and the environment changes, it is necessary to ensure that the model can continuously adapt to the latest status of the bushings; and, the operating environments of bushings in different regions vary greatly, requiring the elimination of the impact of these environmental differences on data consistency.

[0074] To address the aforementioned potential problems, this application provides a more comprehensive, rigorous, and adaptive technical solution. This embodiment not only elucidates the scientific construction method of the interconnected bushing system but also introduces key technical aspects such as a backup evaluation model for new equipment, a dynamic updating and verification mechanism for the model, and data correction considering environmental factors. These significantly enhance the accuracy, robustness, and lifecycle applicability of the monitoring method. The following provides a more detailed description of the process provided in this embodiment. Please refer to... Figure 2 This is another flowchart illustrating the oil-immersed casing monitoring method based on the random forest model in this application embodiment.

[0075] S201. When the cumulative running time of the target casing system reaches the preset running time, acquire the casing body parameters, operating environment parameters, and operating status parameters of the target casing in the target casing system.

[0076] Among them, the preset running time refers to the minimum operating time required for the bushing system, which is usually 3-6 months; the bushing body parameters represent a set of indicators reflecting the basic characteristics of the bushing, including insulation class, manufacturing year, production batch number and production process parameters; the operating environment parameters refer to external condition indicators describing the operating scenario of the bushing, including the altitude of the installation site, atmospheric pollution level, annual average temperature and annual average humidity; the operating status parameters are dynamic indicators used to represent the actual operating status of the bushing, including cumulative running time, average load rate and number of start-ups.

[0077] Before commencing condition assessment, the monitoring system must ensure that the target casing system has sufficient operational experience data. Specifically, the monitoring system first checks whether the cumulative runtime of the target casing system meets the preset requirements by comparing runtime timestamps. Once the runtime meets the requirements, the monitoring system extracts manufacturing information and basic parameters from the casing body database, including design specifications, production date, and process parameters. Simultaneously, the monitoring system obtains environmental data from the operating site from the environmental monitoring system, including temperature, humidity, altitude, and pollution levels. Furthermore, the monitoring system extracts historical operating data from the operation management system, including load records, start / stop statistics, and other operational status information. Finally, the monitoring system standardizes and converts these parameters to form a complete parameter dataset.

[0078] In some embodiments, parameter acquisition and data preprocessing can be achieved in multiple ways: Optionally, the monitoring system can implement a hierarchical data acquisition strategy. First, a parameter classification system is constructed, classifying all parameters according to their intrinsic characteristics, environmental factors, and operating status. Then, corresponding data sources and acquisition cycles are set for different categories. Finally, a data fusion algorithm is used to integrate all types of parameters into a unified format. Optionally, the monitoring system can employ an intelligent parameter filtering mechanism. By establishing a parameter importance assessment model, the contribution of each parameter is analyzed, and the acquisition strategy is dynamically adjusted according to the degree of influence of the parameters on the equipment status, with a focus on the acquisition quality of key parameters. It is understood that other methods can also be used to achieve parameter acquisition and data preprocessing, which are not limited here.

[0079] During implementation, inconsistencies in data quality from different parameter sources can arise, affecting the accuracy of subsequent analyses. To address this, the monitoring system can employ a data quality assessment and compensation mechanism: First, establish a parameter quality evaluation system, including assessment indicators for data completeness, accuracy, and timeliness; then, score each parameter to identify data items with poor quality; next, use data repair algorithms to correct or reconstruct low-quality data based on the physical relationships between parameters and historical data characteristics; finally, cross-validation ensures the reasonableness of the repaired data. For example, when data collection for a certain environmental parameter is interrupted, the system will combine data characteristics from adjacent time periods and meteorological forecast data for supplementary calculations.

[0080] Since newly commissioned equipment often lacks sufficient historical operating data, it is difficult to directly build a reliable random forest model.

[0081] To address this cold start problem and ensure continuous monitoring, in some embodiments, the monitoring system performs an initial assessment of oil-immersed casing systems with insufficient runtime. Specifically, when the cumulative runtime of the target casing system has not reached the preset runtime, the monitoring system acquires the factory test data and early operating data of the target casing system; constructs a mapping relationship between the factory test data and the early operating data to obtain systematic deviations; constructs compensation factors based on systematic deviations and operating environment parameters to obtain a backup assessment model; and performs a status assessment of the target casing system based on the backup assessment model until the cumulative runtime reaches the preset runtime.

[0082] Among them, factory test data refers to the results of various performance tests conducted on the bushing after manufacturing, including data on medium loss testing, capacity testing, and pressure resistance testing; early operation data represents the actual operating parameters collected in the early stages of bushing commissioning, usually including monitoring data within the first month after commissioning; systematic deviation is used to represent the systematic differences between the factory test environment and the actual operating environment; compensation factor refers to the set of correction coefficients used to correct systematic deviation; and standby evaluation model represents an alternative model used to temporarily evaluate the operating status of the bushing when historical data is insufficient.

[0083] When handling newly commissioned casing systems, the monitoring system needs to address the issue of insufficient historical data. Specifically, the system first obtains the complete factory test report for the casing from the manufacturer, including test data and conditions for various performance indicators. Then, the system collects early operational data after the casing is commissioned, focusing on monitoring parameters corresponding to the factory test items. Next, the system uses data analysis algorithms to establish the correspondence between the factory data and early operational data, calculating the systematic deviations of various parameters. Afterward, the system combines current operating environment parameters to construct adaptive compensation factors for dynamically adjusting evaluation criteria. Finally, based on the compensated parameters, the system constructs a backup evaluation model, which is used for temporary state assessments until the cumulative operating time reaches a preset requirement.

[0084] In some embodiments, the data mapping and model building process can be implemented in multiple ways: Optionally, the monitoring system can adopt a multi-level mapping analysis strategy, first classifying and preprocessing the parameters of the factory test data and early operation data, then using machine learning algorithms to establish mapping models for different categories of parameters, and finally optimizing the mapping relationship through parameter correlation analysis to ensure the accuracy of systematic deviation calculation; Optionally, the monitoring system can implement a dynamic compensation mechanism, by establishing an environmental factor sensitivity model, calculating the degree of influence of environmental changes on various parameters in real time, dynamically adjusting the calculation method of compensation factors, and continuously optimizing the compensation effect through a feedback mechanism. It is understood that other methods can also be used to implement data mapping and model building, which are not limited here.

[0085] During implementation, a significant discrepancy may arise between factory test data and actual operating data, leading to inaccurate calculations of systematic deviations. To address this, the monitoring system employs a difference analysis and correction mechanism: First, the factory test data is standardized to eliminate the influence of different test environments; then, time-series analysis is used to study the changing trends of early operating data, identifying characteristics of the stable operating phase; next, a physical model is established to analyze the causes of parameter differences, distinguishing between systematic deviations caused by environmental factors and changes in the equipment's inherent characteristics; finally, the correction effect is verified using a validation dataset to ensure the reliability of the backup evaluation model. For example, when a significant fluctuation in the medium loss value of a bushing is detected in the initial stage of operation, the system will analyze this in conjunction with changes in ambient temperature to accurately identify the true systematic deviation.

[0086] It should be noted that the backup assessment model and the random forest model in this application can be used in conjunction. Specifically, from the perspective of model positioning, the backup assessment model mainly addresses the problem of insufficient early-stage data, while the random forest model is responsible for long-term condition monitoring. The backup assessment model is built based on factory test data and early-stage operational data, providing reliable condition assessments in the initial stages of equipment commissioning; the random forest model, on the other hand, requires sufficient historical data support for continuous monitoring after the equipment reaches its preset operating period. Each model has its advantages: the backup assessment model employs a physics-based compensation mechanism, accurately reflecting the early performance characteristics and environmental impacts of the equipment; the random forest model, through ensemble learning of multiple decision trees, can capture complex failure modes and performance degradation characteristics. This complementarity ensures the accuracy of the monitoring system's assessments throughout the entire equipment lifecycle.

[0087] The monitoring system employs a smooth transition mechanism for processing the two models: as the system approaches the preset runtime, it runs both models simultaneously, weighting and fusing the evaluation results, with the weights dynamically adjusted over time. Specifically, when the cumulative runtime reaches 80% of the preset duration, training of the random forest model begins; within the 80%-100% runtime range, the evaluation weight of the random forest model is gradually increased while the weight of the backup model is decreased; and when the preset runtime is reached, the system fully switches to the random forest model.

[0088] Furthermore, the early data and evaluation experience accumulated by the backup evaluation model are not discarded, but rather used as important training inputs for the random forest model. The monitoring system transforms the systematic biases and compensation factor information in the backup model into feature engineering rules for the random forest model, improving the model's adaptability to environmental factors. Simultaneously, the evaluation results of the backup model are also used as labeled data to enhance the early prediction accuracy of the random forest model. In addition, the monitoring system implements a model backup mechanism: even after a complete switch to the random forest model, the backup evaluation model remains online. When the random forest model exhibits evaluation bias due to data anomalies or drastic environmental changes, the system can refer to the evaluation results of the backup model for correction, further improving the reliability of the monitoring system.

[0089] By employing a dual-model mechanism, this application not only solves the challenges of early equipment monitoring but also achieves a smooth transition and mutual verification between monitoring models, thereby improving the full-cycle reliability of oil-immersed casing condition monitoring. Compared to a single-model approach, it exhibits greater adaptability and stability, providing a more reliable basis for equipment management decisions.

[0090] S202. Based on the casing body parameters, operating environment parameters, and operating status parameters, calculate the deviation direction and deviation index of the target casing relative to the standard casing model.

[0091] Among them, the standard bushing model refers to the reference model under ideal operating conditions constructed based on the historical operating data of bushings with the same insulation level as the target bushing and combined with expert evaluation rules; the deviation direction indicates the changing trend of the operating characteristics of the target bushing relative to the standard model, including the judgment of the trend of parameter increase or decrease; the deviation index is used to indicate the degree of difference between the target bushing and the standard model, and usually adopts a normalized quantitative index.

[0092] After acquiring a complete parameter dataset, the monitoring system needs to assess the degree of deviation of the target bushing's operating status. Specifically, the monitoring system first constructs a standard bushing model based on the operating records of bushings with the same insulation class in the historical database, combined with industry standards and expert experience. Then, the monitoring system compares the actual parameters of the target bushing with the standard model from multiple dimensions, analyzing the deviation of each parameter. Next, based on the physical characteristics and correlations of parameter changes, the monitoring system determines the direction of deviation for each parameter and calculates the corresponding degree of deviation. Afterward, the monitoring system uses a weighted fusion algorithm to comprehensively consider the importance of different parameters and calculates the overall deviation index. Finally, the monitoring system standardizes the deviation direction and deviation index to ensure the comparability of different types of parameters.

[0093] In some embodiments, deviation analysis and index calculation can be implemented in multiple ways: Optionally, the monitoring system can adopt a multi-level deviation analysis strategy, first establishing a three-level evaluation system of parameter layer, feature layer, and state layer; performing basic deviation calculation at the parameter layer; performing parameter combination analysis at the feature layer; and performing comprehensive evaluation at the state layer; finally, obtaining the final deviation index through hierarchical weighting. Optionally, the monitoring system can implement a dynamic weight optimization mechanism, establishing a parameter importance model by analyzing the contribution of various parameters in historical failure cases, and dynamically adjusting the weight coefficients of each parameter in the deviation calculation according to the operating conditions. It is understood that other methods can also be used to implement deviation analysis and index calculation, which are not limited here.

[0094] During implementation, significant differences in the deviation characteristics of various parameters may arise, leading to distortion in the deviation index calculation. To address this, the monitoring system can employ a parameter characteristic normalization mechanism: first, analyze the physical characteristics and variation patterns of various parameters to establish a parameter characteristic model; then, design corresponding normalization functions based on the physical meaning of the parameters to transform parameters with different dimensions and variation ranges into a unified evaluation space; simultaneously, introduce a nonlinear mapping method to handle the nonlinear relationship between parameter deviation and equipment status. For example, when the variation characteristics of temperature parameters and media loss parameters differ significantly, the system will use different normalization functions for transformation to ensure the rationality of the calculation results.

[0095] S203. Calculate the operational correlation between the target bushing and other bushings in the power system based on the deviation direction and deviation index, determine the bushing system where the bushing with the operational correlation within the preset correlation threshold is located, and determine the bushing system that has reached the preset operating time as the interconnected bushing system.

[0096] Among them, the operational correlation degree represents the similarity of the operational characteristics between different casing systems, which is calculated based on the deviation direction and deviation index; the preset correlation threshold is the critical value for determining whether the casing systems have sufficient similarity, which is usually determined based on historical operational experience; the interconnected casing system is used to represent the set of casing systems that have high operational similarity with the target casing system and meet the requirements for operating time.

[0097] After obtaining the deviation characteristics of the target bushing, the monitoring system needs to search for similar bushing systems within the power system. Specifically, the monitoring system first retrieves deviation characteristic data for all operating bushings from the power system database. Then, based on the consistency of the deviation direction and the proximity of the deviation index, the monitoring system calculates the operational correlation between the target bushing and other bushings. Next, the monitoring system compares the calculated correlation with a preset threshold, selecting bushing systems with high correlation. Afterward, the monitoring system checks the operating time of these bushing systems to ensure they have sufficient operational experience. Finally, the monitoring system identifies bushing systems that simultaneously meet the correlation and operating time requirements as interconnected bushing systems, establishing an interconnection evaluation network.

[0098] In some embodiments, correlation calculation and system screening can be implemented in multiple ways: Optionally, the monitoring system can implement a multi-dimensional correlation analysis strategy, first constructing a multi-dimensional correlation index system including temporal correlation, feature similarity, and state consistency, then training a correlation calculation model through machine learning algorithms, and finally dynamically adjusting the correlation weights based on the historical performance of the casing system; Optionally, the monitoring system can adopt a hierarchical screening mechanism, gradually selecting the most suitable interconnected casing systems by setting multi-level screening criteria such as basic correlation, core feature matching degree, and operational state similarity. It is understood that other methods can also be used to implement correlation calculation and system screening, which are not limited here.

[0099] During implementation, the large number of bushing systems can lead to excessive computational resource consumption, affecting the real-time performance of correlation analysis. To address this, the monitoring system can employ an efficient correlation calculation mechanism: first, feature pre-screening reduces the number of bushings to be calculated, allowing detailed analysis only of bushings with similar basic characteristics; then, a parallel computing framework is used to distribute the correlation calculation task across multiple processing units; simultaneously, an incremental update strategy is implemented, recalculating the correlation only for bushings whose status has significantly changed. For example, when a new bushing is put into operation or the status of an existing bushing changes significantly, only the correlation relationships of the affected parts are updated, avoiding a global recalculation.

[0100] S204. Obtain historical monitoring data of the target casing system within a preset time period, and construct a random forest model based on the historical monitoring data.

[0101] Referring to step S101, the monitoring system will train a random forest model based on historical data for casing status assessment.

[0102] S205. After the target casing system enters the next data management cycle, adjust the preset time period to the latest time period.

[0103] Among them, the data management cycle refers to the fixed time interval for the system to regularly update and maintain data, which is usually 1-3 months; the latest time period refers to the new data collection interval calculated backward from the current time point; the time period adjustment is used to indicate the process of updating the time range of data collection and analysis.

[0104] The monitoring system needs to periodically update the time window for data analysis. Specifically, the monitoring system first checks whether the operating time of the target casing system has crossed the current data management cycle node. When a new data management cycle begins, the monitoring system calculates the start and end times of the new time period, including determining the start and end times of the latest time period. Then, the monitoring system updates the data acquisition configuration, adjusting the data collection scope to the latest time period. Finally, the monitoring system archives the data from historical time periods to ensure data continuity and traceability.

[0105] In some embodiments, the dynamic adjustment of time periods can be achieved in several ways: Optionally, the monitoring system can implement an adaptive cycle adjustment mechanism to dynamically adjust the length of the data management cycle based on the stability of the bushing's operating status and the characteristics of data changes, using a shorter update cycle for periods with large state fluctuations to ensure that the model can respond promptly to changes in system status; Optionally, the monitoring system can adopt a smooth transition strategy for time periods, retaining a portion of the overlap period when entering a new time period, and avoiding data abrupt changes caused by time period switching through a gradual update method. It is understood that other methods can also be used to adjust the preset time period, which are not limited here.

[0106] During implementation, the discontinuity of model evaluation results caused by time period switching can affect the accuracy of state assessment. To address this, the monitoring system can employ a time period transition smoothing mechanism: First, before the time period switch, the data characteristics of the old and new time periods are compared and analyzed to identify key factors that may cause a jump in evaluation results; then, a feature smoothing function is designed to progressively adjust key features during the time period transition; simultaneously, a dual-time period parallel evaluation mechanism is introduced, using data from both the old and new time periods for evaluation during the transition period, and obtaining the final result through weighted fusion. For example, when a difference is detected between the data distribution of the new time period and historical data, the system will retain the evaluation results of both time periods for a certain period, ensuring the continuity of evaluation through a smooth transition.

[0107] S206. Obtain the latest monitoring data of the target casing system within the latest time period, and retrain the random forest model based on the latest monitoring data.

[0108] Among them, the latest monitoring data represents the various operating parameter data collected by the target casing system within the adjusted time period; retraining is used to represent the process of optimizing and updating the parameters of the random forest model using the updated dataset; the model training process refers to the complete calculation process of optimizing the parameters of the random forest model using the monitoring data.

[0109] After adjusting the time period, the monitoring system needs to update the evaluation model. Specifically, the monitoring system first collects all monitoring data of the target casing system within the latest time period, including routine monitoring data and data under special operating conditions. Then, the monitoring system preprocesses and extracts features from the newly collected data to ensure data quality meets the model training requirements. Next, the monitoring system retrains the random forest model using the processed, latest dataset, including updating feature weights, optimizing the decision tree structure, and adjusting model parameters. Finally, the monitoring system performs performance validation on the updated model to ensure it accurately reflects the latest operating status of the casing system.

[0110] In some embodiments, model update training can be implemented in several ways: Optionally, the monitoring system can employ an incremental learning strategy, focusing on changes in state characteristics reflected by the latest data based on the original model. By adjusting the growth rules and pruning strategies of the decision tree, the model can adapt to dynamic changes in system performance while maintaining its ability to remember historical features. Optionally, the monitoring system can train multiple sub-models simultaneously, each using datasets at different time scales. By using ensemble learning methods to synthesize the prediction results of each sub-model, the model's generalization ability can be improved. It is understood that other methods can also be used to implement model update training, which are not limited here.

[0111] During implementation, the discrepancy in feature distribution between old and new data can lead to unstable performance of the updated model. To address this, the monitoring system can employ a feature transfer learning mechanism: first, a comparative analysis of the feature distributions of the old and new datasets is performed to identify the changed feature dimensions; then, a feature transformation function is designed to gradually map the feature distribution of the new data to a feature space similar to historical data; simultaneously, an adversarial training strategy is introduced to improve the model's adaptability by minimizing the difference between the old and new feature distributions. For example, when certain monitoring parameters exhibit significant seasonal changes in a new time period, the system uses methods such as feature standardization and distribution alignment to ensure the model accurately identifies the true state changes.

[0112] To ensure the effectiveness of model updates and avoid the new model's performance being inferior to the old model due to data fluctuations or overfitting, thus guaranteeing the stability and reliability of the monitoring system, in some embodiments, the monitoring system performs periodic performance evaluations and updates to the model. Specifically, the monitoring system divides out an independent validation dataset from the latest monitoring data; this validation dataset is not used for retraining the random forest model; the validation dataset is evaluated using both the original and updated random forest models to obtain pre-performance metrics and post-performance metrics; when the post-performance metrics are better than the pre-performance metrics, the updated random forest model is deployed to the monitoring system.

[0113] Among them, the validation dataset represents independent test data used to evaluate model performance, which usually accounts for 20%-30% of the latest monitoring data; the pre-performance metrics refer to the performance parameters such as the accuracy and recall of the model on the validation dataset before the update; the post-performance metrics are used to represent the performance evaluation results of the updated model; and the model deployment refers to the process of putting the validated new model into actual operation.

[0114] After updating the random forest model, the monitoring system undergoes rigorous performance validation. Specifically, the system first randomly selects a certain proportion of samples from the latest monitoring data as a validation dataset, ensuring that the validation data is independent and representative of the training data. Then, the system evaluates the validation dataset using both the pre-update and updated random forest models, obtaining the evaluation results for each model. Next, the system calculates various performance metrics for both models on the validation dataset, including key indicators such as evaluation accuracy, false positive rate, and false negative rate. Finally, the system compares the performance metrics of the two models, and only when the new model outperforms the old model in all metrics will it be officially deployed to the monitoring system.

[0115] In some embodiments, model performance validation can be achieved in multiple ways: Optionally, the monitoring system can adopt a hierarchical validation strategy, firstly by stratifying the validation dataset according to state categories and feature distributions to ensure that the validation data covers various operating states, then by performing performance evaluations on different types of data separately, establishing a weighted comprehensive evaluation system, and finally making deployment decisions based on multi-dimensional performance comparison results; Optionally, the monitoring system can implement a cross-validation mechanism, dividing the validation data into multiple subsets, evaluating the stability of model performance through multiple rounds of cross-validation, and introducing confidence interval analysis to evaluate the statistical nature of performance improvement. It is understood that other methods can also be used to achieve model performance validation, which are not limited here.

[0116] During implementation, the performance of new models may actually decline under certain special operating conditions, affecting the accuracy of model upgrade decisions. To address this, the monitoring system can employ a scenario-based verification mechanism: first, the verification data is categorized by operating conditions to identify different scenarios; then, performance evaluation is performed for each scenario, establishing a scenario weighting system; simultaneously, a performance degradation detection mechanism is introduced to comprehensively analyze the model's performance under various scenarios. For example, if a slight performance decrease is detected in the new model under high load conditions, the system will assess the importance and frequency of this condition, comprehensively deciding whether to deploy the new model or perform targeted optimization.

[0117] S207. Obtain real-time monitoring data of the target bushing system and upload it to the power system.

[0118] Referring to step S102, the monitoring system will acquire and upload real-time monitoring data.

[0119] Considering the differences in operating environments (such as temperature and humidity) among different bushings, these differences directly affect the values ​​of monitoring data. To eliminate the interference of such environmental factors and ensure the comparability of data and the accuracy of evaluation, in some embodiments, the monitoring system will perform environmental factor correction on the monitoring data. That is, the monitoring system will acquire the operating environment parameters of the target bushing system and determine the data correction coefficient of the operating environment parameters relative to the standard environment parameters; based on the data correction coefficient, the monitoring data of the target bushing system will be corrected to obtain real-time monitoring data, and the real-time monitoring data will be uploaded to the power system.

[0120] Among them, the operating environment parameters represent the external environmental factors that affect the operating status of the bushing system, including ambient temperature, humidity, altitude, etc.; the standard environmental parameters refer to the standard working environment conditions when the equipment is designed, usually referring to a temperature of 20℃, humidity of 65%, and an altitude not exceeding 1000m; the data correction coefficient is used to represent the compensation value for the difference between the actual operating environment and the standard environment; and the real-time monitoring data represents the operating parameters of the bushing system after correction for environmental factors.

[0121] While collecting raw monitoring data, the monitoring system needs to consider the impact of environmental factors. Specifically, the system first collects environmental parameters such as temperature, humidity, and air pressure at the operating site using environmental sensors. Then, the system compares the collected environmental parameters with preset standard environmental parameters, calculating the deviation values ​​for each environmental factor. Next, based on the correlation model between environmental factors and monitoring parameters, the system determines the corresponding data correction coefficients. Finally, the system uses these correction coefficients to correct the raw monitoring data, obtaining more accurate real-time monitoring data, and uploads the corrected data to the power system.

[0122] In some embodiments, environmental parameter correction and data calibration can be achieved in multiple ways: Optionally, the monitoring system can adopt a hierarchical correction strategy. First, a sensitivity model between environmental parameters and monitoring indicators is established, different environmental impact weights are set for different monitoring parameters, then correction coefficients are dynamically calculated based on real-time environmental conditions, and finally, the original data is precisely corrected through a weighted combination to ensure that the corrected data accurately reflects the true state of the equipment. Optionally, the monitoring system can implement an adaptive correction mechanism. By continuously analyzing environmental change trends and equipment response characteristics, a dynamic compensation model for environmental impact is established, the calculation method of correction coefficients is adjusted in real time, and the correction effect is verified by combining historical data to continuously optimize the accuracy of the correction strategy. It is understood that other methods can also be used to achieve environmental parameter correction and data calibration, which are not limited here.

[0123] During implementation, abrupt changes in environmental parameters may cause drastic fluctuations in the correction coefficient, leading to unreasonable jumps in monitoring data. To address this, the monitoring system can employ an environmental parameter smoothing mechanism: first, perform time-series analysis on the environmental parameters to identify abnormal environmental changes; then, use methods such as moving averages to smooth the environmental parameters, reducing the impact of sudden changes; simultaneously, introduce a limit on the rate of change of the correction coefficient. When the change in the correction coefficient exceeds a threshold, a gradual correction method is adopted to ensure the stability of the data correction. For example, when a rapid change in ambient temperature is detected within a short period, the system will dynamically adjust the correction coefficient based on the physical characteristics of the temperature change, avoiding unreasonable fluctuations in the corrected data.

[0124] S208 Receive operation monitoring data of multiple interconnected bushing systems sent by the power system, and construct a test sample set based on real-time monitoring data and operation monitoring data.

[0125] Referring to step S103, the monitoring system will integrate the monitoring data of multiple interconnected bushings to form a sample set to be tested.

[0126] S209. Input the sample set to be tested into the random forest model to obtain the state assessment results of the target bushing system on multiple interconnected bushing systems, and upload the state assessment results to the power system.

[0127] Referring to step S104, the monitoring system uses a random forest model to evaluate the status of multiple interconnected bushings.

[0128] S210. Obtain the target assessment result of the target bushing system output by the power system after receiving all the state assessment results.

[0129] Referring to step S105, the monitoring system will obtain the target casing status result after comprehensive evaluation.

[0130] After obtaining the evaluation results of a single bushing, in order to explore the group correlation between devices and promptly identify potential batch defects or systemic risks, in some embodiments, the monitoring system will perform collective failure analysis and early warning. That is, the monitoring system will count the percentage of bushing systems judged to be in a warning state among multiple interconnected bushing systems and target bushing systems. When the percentage exceeds the collective failure judgment threshold, the status prompt information of the target bushing system will be corrected to a collective failure early warning.

[0131] Among them, the warning status indicates an intermediate state in which the operating parameters of the bushing system exceed the normal range but have not yet reached the level of failure; the number percentage refers to the ratio of the number of bushing systems in the warning status to the total number of bushing systems; the collective failure judgment threshold is used to indicate the critical percentage value for judging whether a batch problem has occurred, which is usually set to 30%-50%; the collective failure warning indicates a system-level warning message issued due to the potential problems that are common in the same batch of products.

[0132] When the monitoring system detects an early warning for a single bushing system, it needs to perform a group analysis. Specifically, the monitoring system first counts the total number of all interconnected bushing systems and the target bushing system. Then, the monitoring system checks the status assessment results of each bushing system one by one, calculating the number of systems in an early warning state. Next, the monitoring system calculates the percentage of early warning systems out of the total and compares it with a preset threshold for determining a collective failure. When the percentage exceeds the threshold, the monitoring system upgrades the original single-system early warning to a collective failure early warning, while updating the relevant early warning information. Finally, the monitoring system activates a special monitoring mechanism for this batch of products, increasing the monitoring frequency and analysis depth of all relevant systems.

[0133] In some embodiments, the determination and early warning of collective failures can be achieved in multiple ways: Optionally, the monitoring system can employ a multi-dimensional statistical analysis strategy. First, it can perform feature clustering on the casing systems in the early warning state to identify the similarity of early warning patterns. Then, it can analyze the distribution characteristics of different early warning types, establish an early warning propagation model, and finally dynamically adjust the criteria for determining collective failures based on the spatiotemporal distribution characteristics and diffusion trends of the early warnings. Optionally, the monitoring system can implement a tiered early warning mechanism, setting multiple early warning levels based on the proportion of early warning systems, adopting different response strategies for different levels, and establishing a dynamic adjustment mechanism for early warning escalation and de-escalation to ensure the timeliness and accuracy of early warnings. It is understood that other methods can also be used to determine and warn of collective failures, which are not limited here.

[0134] During implementation, differences in early warning characteristics of casing systems under different operating environments can affect the accuracy of collective fault determination. To address this, the monitoring system can employ an environmental grouping assessment mechanism: first, all casing systems are grouped according to their operating environment characteristics, establishing an environmental similarity evaluation system; then, the early warning rate is calculated within each environmental group, and a judgment threshold is set within each group; simultaneously, an inter-group correlation analysis model is established to comprehensively evaluate the early warning status of different environmental groups. For example, when the early warning rate of casing systems in high-altitude areas is found to be significantly higher than that in plain areas, the system will set differentiated judgment criteria for different altitude conditions to improve the accuracy of collective fault determination.

[0135] S211. Based on the target assessment results, generate status prompt information for the target casing system.

[0136] Referring to step S106, the monitoring system will generate corresponding status warnings or normal operation prompts.

[0137] In this embodiment, by employing techniques such as environmental parameter correction, backup evaluation using factory test data, and model performance verification, collaborative evaluation and early warning of bushings in the same batch can be achieved. By establishing a correlation model for the interconnected bushing system, the technical deficiency of traditional single-device monitoring in detecting batch-related problems is effectively solved; the environmental parameter correction mechanism overcomes the interference of different operating environments on the evaluation results; the backup evaluation model constructed using factory test data solves the problem of insufficient data in the early stages of equipment operation; and rigorous model performance verification ensures the reliability of the evaluation results. This enables accurate prediction and timely early warning of batch-related faults in power equipment, improves the operational reliability of the power system and the level of equipment management, and has significant engineering application value.

[0138] The monitoring system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of the physical device structure of the monitoring system in this application embodiment.

[0139] It should be noted that, Figure 3 The structure of the monitoring system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0140] like Figure 3As shown, the monitoring system includes a CPU 301, which can perform various appropriate actions and processes according to a program stored in ROM 302 or a program loaded into RAM 303 from storage section 308, such as executing the methods described in the above embodiments. RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via bus 304. I / O interface 305 is also connected to bus 304.

[0141] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including hard disks, etc.; and communication section 309 including network interface cards such as LAN (Local Area Network) cards, modems, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0142] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by CPU 301, it performs the various functions defined in the present invention.

[0143] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0144] Specifically, the monitoring system in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the oil immersion casing monitoring method based on the random forest model provided in the above embodiment.

[0145] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the monitoring system described in the above embodiments; or it may exist independently and not assembled into the monitoring system. The storage medium carries one or more computer programs that, when executed by a processor of the monitoring system, cause the monitoring system to implement the oil-immersed casing monitoring method based on the random forest model provided in the above embodiments.

[0146] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0147] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

Claims

1. A method for monitoring oil-immersed casing based on a random forest model, characterized in that, Applied to a monitoring system, the method includes: Acquire historical monitoring data of the target casing system within a preset time period, and construct a random forest model based on the historical monitoring data; Acquire real-time monitoring data of the target bushing system and upload it to the power system; Receive operation monitoring data of multiple interconnected bushing systems sent by the power system, and construct a test sample set based on the real-time monitoring data and the operation monitoring data; The test sample set is input into the random forest model to obtain the state assessment results of the target bushing system on the multiple interconnected bushing systems, and the state assessment results are uploaded to the power system. After receiving all the state assessment results, the target assessment result of the target bushing system is obtained. Based on the target evaluation results, a status indication message for the target casing system is generated.

2. The method according to claim 1, characterized in that, The step of acquiring real-time monitoring data of the target bushing system and uploading it to the power system specifically includes: Obtain the operating environment parameters of the target casing system, and determine the data correction coefficient of the operating environment parameters relative to the standard environment parameters; The monitoring data of the target bushing system is corrected based on the data correction coefficient to obtain real-time monitoring data, and the real-time monitoring data is uploaded to the power system.

3. The method according to claim 1, characterized in that, Before the step of acquiring historical monitoring data of the target casing system within a preset time period and constructing a random forest model based on the historical monitoring data, the method further includes: When the cumulative running time of the target casing system reaches the preset running time, the casing body parameters, operating environment parameters, and operating status parameters of the target casing in the target casing system are obtained; Based on the bushing body parameters, the operating environment parameters, and the operating state parameters, the deviation direction and deviation index of the target bushing relative to the standard bushing model are calculated; the standard bushing model is a reference model under ideal operating conditions constructed based on historical operating data of bushings with the same insulation level as the target bushing and combined with expert evaluation rules. Based on the deviation direction and the deviation index, the operational correlation degree between the target bushing and other bushings in the power system is calculated. The bushing system in which the bushing with the operational correlation degree is within the preset correlation threshold is determined, and the bushing system that reaches the preset operating time is determined as the interconnected bushing system.

4. The method according to claim 3, characterized in that, Before the step of calculating the deviation direction and deviation index of the target casing relative to the standard casing model based on the casing body parameters, the operating environment parameters, and the operating state parameters, the method further includes: When the cumulative running time of the target casing system has not reached the preset running time, the factory test data and early operation data of the target casing system are acquired. A mapping relationship is constructed between the factory test data and the early operation data to obtain the systematic deviation; Based on the aforementioned systematic deviations and operating environment parameters, compensation factors are constructed to obtain a backup evaluation model; The target casing system is assessed based on the backup assessment model until the cumulative runtime reaches the preset runtime.

5. The method according to claim 1, characterized in that, After the step of obtaining the target assessment result of the target bushing system output by the power system after receiving all the state assessment results, the method further includes: The percentage of all casing systems that are identified as being in a warning state is calculated among the multiple interconnected casing systems and the target casing system. When the proportion of such cases exceeds the threshold for determining a collective fault, the status message of the target bushing system is corrected to a collective fault warning.

6. The method according to claim 1, characterized in that, After the steps of acquiring historical monitoring data of the target casing system within a preset time period and constructing a random forest model based on the historical monitoring data, the method further includes: After the target casing system's operating time enters the next data management cycle, the preset time period is adjusted to the latest time period; Obtain the latest monitoring data of the target casing system within the latest time period, and retrain the random forest model based on the latest monitoring data.

7. The method according to claim 6, characterized in that, After the steps of obtaining the latest monitoring data of the target casing system within the latest time period and retraining the random forest model based on the latest monitoring data, the method further includes: A separate validation dataset is extracted from the latest monitoring data; the validation dataset is not used for retraining the random forest model. The validation dataset was evaluated using the random forest model before and after the update, respectively, and the pre-performance metrics and post-performance metrics were obtained. When the subsequent performance metric is better than the preceding performance metric, the updated random forest model is deployed in the monitoring system.

8. A monitoring system, characterized in that, The monitoring system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the monitoring system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the monitoring system, the monitoring system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the monitoring system, the monitoring system performs the method as described in any one of claims 1-7.

Images