Power equipment thermal fault whole cycle management and control method based on temperature rise amount and expert weight fusion

Abstract

The present application relates to the technical field of power equipment intelligent operation and maintenance and fault diagnosis, and discloses a power equipment thermal fault whole cycle management and control method based on temperature rise amount and expert weight fusion, which establishes a temperature rise benchmark model bound with the aging state of the equipment individual, replaces the general fixed threshold with the tolerance limit attenuated with the running life index, dynamically calculates the remaining available decision time in combination with the real-time temperature rise change characteristics and the load working condition, fills the personalized time information gap between defect alarm and emergency disposal, constructs a fusion game model of temperature rise data and expert weight, takes the conflict degree of the risk assessment results of both sides as the confidence quantitative representation, triggers arbitration and adjusts the collection strategy or corrects the weight composition when the conflict exceeds the threshold, makes the uncertainty information run through the transmission to the execution layer, generates the differentiated management and control strategy, eliminates the decision blind area under the certainty illusion caused by the confidence cross-layer discard, and realizes the whole cycle closed loop self-evolution management and control.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance and fault diagnosis technology for power equipment, and in particular to a method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights. Background Technology

[0002] As power systems evolve towards intelligence and digitalization, early warning and scientific management of thermal faults in power equipment have become core issues in ensuring the safe operation of the power grid. Thermal faults are one of the most common types of faults in power equipment, encompassing various types such as current-induced thermal faults (e.g., poor conductor connection, contact oxidation), voltage-induced thermal faults (e.g., insulation degradation, increased dielectric loss), and electromagnetic-induced thermal faults. How to accurately perceive, scientifically assess, and proactively manage the thermal state of equipment is a key direction for continuous exploration in this field.

[0003] The above-mentioned and existing related technologies often have the following drawbacks: 1. In actual on-site operation and maintenance, when the temperature rise rate of power equipment is relatively fast, the operation and maintenance personnel still cannot know the remaining available decision time under the combined effect of the current actual operating load curve and the aging state of the individual equipment after receiving a serious defect alarm. This results in an insurmountable gap in personalized time information between the defect level information provided by the general temperature rise threshold standard and the quantitative requirements for emergency response time under the actual on-site operating constraints of the specific equipment. As a result, the operation and maintenance decision-making is passively invalidated in the accelerated temperature rise stage due to the lack of time buffer prediction bound to the individual characteristics of the equipment. 2. Because the existing assessment model only transmits the deterministic level conclusion after fusion calculation in the output stage of the operation and maintenance strategy, while systematically discarding the confidence information generated during the assessment process that represents the reliability of the conclusion during cross-layer transmission, high-confidence high-risk alarms and low-confidence high-risk false alarms are treated indiscriminately at the strategy response end. At the same time, low-confidence low-risk conclusions and high-confidence low-risk conclusions are also treated the same, causing on-site operation and maintenance to fall into a decision blind spot under the illusion of certainty caused by the break in the uncertainty information transmission chain when executing alarm-handling mapping, and unable to implement differentiated response strategies according to the credibility of the assessment conclusion. Summary of the Invention

[0004] The technical problem to be solved by this invention is that existing technologies have the drawbacks of lacking remaining time for temperature rise alarms and having blind spots in decision-making due to loss of confidence in assessment. To address this, we propose a full-cycle management method for thermal faults in power equipment based on the fusion of temperature rise and expert weights.

[0005] To achieve the above objectives, this application adopts the following technical solution: a full-cycle management method for thermal faults of power equipment based on the fusion of temperature rise and expert weights, comprising the following steps: Step 1: Collect real-time temperature rise data and related parameter data of the power equipment to be managed, and preprocess them to form a temperature rise time series dataset; Step 2: Establish a temperature rise baseline model based on the equipment's historical health operation data. The temperature rise baseline model includes temperature rise tolerance limits that decrease with the number of years of operation. Compare the real-time temperature rise data with the temperature rise baseline model and calculate the multidimensional temperature rise feature vector. Step 3: Construct a multi-level expert weight system to form an expert weight vector; Step 4: Establish a game model that integrates temperature rise and expert weights. Place the multidimensional temperature rise feature vector and the expert weight vector in the same game framework. Calculate the degree of conflict between the temperature rise-side risk assessment result obtained by mapping the multidimensional temperature rise feature vector and the weight-side risk assessment result obtained by mapping the expert weight vector. When the degree of conflict exceeds the preset conflict threshold, trigger arbitration and adjust the data acquisition strategy or modify the composition of the expert weight vector according to the arbitration result. Obtain the optimal fusion coefficient by solving the Nash equilibrium and calculate the comprehensive thermal failure risk index based on the optimal fusion coefficient. Step 5: Based on the current value of the comprehensive thermal failure risk index and the temperature rise change characteristics in the multidimensional temperature rise feature vector, combined with the temperature rise tolerance limit in the temperature rise benchmark model, calculate the remaining available decision time from the current temperature rise state to the thermal breakdown critical state under the combined effect of the current actual operating load curve and the individual aging state of the equipment. Step 6: Generate differentiated full-cycle management and control strategies based on the comprehensive risk index of thermal failure and the remaining available decision time. Transform the management and control strategies into execution instructions and push them to the operation and maintenance system to track the execution effect. Iteratively correct the parameters of the temperature rise benchmark model, multi-level expert weight system and fusion game model based on the execution effect feedback data, and store the management and control cases in the knowledge base to form a closed loop.

[0006] Preferably, the associated parameter data in step 1 includes load current, ambient temperature, ambient humidity, and basic equipment information; In step 2, the multidimensional temperature rise feature vector includes absolute temperature rise, relative temperature difference, temperature rise rate of change, three-phase temperature rise imbalance, and hot spot area characteristics. Step 3's multi-level expert weighting system includes weights for equipment importance, basic equipment health, severity of failure consequences, and credibility of monitoring data.

[0007] Preferably, the method for establishing the temperature rise baseline model in step 2 is as follows: Temperature rise samples under different load rate and ambient temperature combinations were extracted from the historical healthy operation data of the power equipment to be managed, and the temperature rise surface under normal operating conditions was fitted by multivariate nonlinear regression. Historical temperature rise data of similar equipment of the same model and batch were retrieved, and the mean and standard deviation of temperature rise under each operating condition were calculated as a benchmark for horizontal comparison. Based on the equipment's years of operation and the thermal aging characteristics of the insulation material, a relationship is established between the temperature rise tolerance limit and the decrease in temperature over the years of operation.

[0008] Preferably, the calculation method for each dimension of the multidimensional temperature rise feature vector in step 2 is as follows: Absolute temperature rise is the difference between the measured temperature of the equipment and the ambient temperature. The relative temperature difference is the ratio of the current absolute temperature rise to the average temperature rise of similar equipment under the same operating conditions. The rate of temperature change is determined by the slope of the linear fit of the temperature rise sequence within the preset time window before the current moment; The three-phase temperature rise imbalance is taken as the difference between the maximum and minimum temperature rise of the three phases. The hot spot area feature is obtained by extracting the hot spot area through gradient analysis and boundary tracking of infrared thermal imaging images.

[0009] Preferably, the dimensions of the multi-level expert weighting system in step 3 are defined as follows: The importance weight of equipment is assessed based on the equipment voltage level, power supply range, and the number of users affected by a fault; The basic weight of equipment health is based on a comprehensive assessment of the equipment's years of operation, historical failure frequency, last maintenance time, and quality score. The severity of the fault consequences is weighted based on the potential personal safety risks, the scope of the power outage, the restoration time, and the economic losses that the fault may cause. The reliability weight of monitoring data is assessed based on the accuracy level of the temperature measurement method, the data integrity rate, and the degree of environmental electromagnetic interference.

[0010] Preferably, in step 4, adjusting the data collection strategy or modifying the composition of the expert weight vector based on the arbitration result specifically involves: If the risk assessment result on the temperature rise side is higher than the risk assessment result on the weight side, the collection frequency of real-time temperature rise data will be automatically increased and a review request will be pushed to the operation and maintenance expert terminal. If the risk assessment result on the temperature rise side is lower than the risk assessment result on the weight side, then maintain the current data collection frequency and reduce the growth rate of the adjustment factor corresponding to the basic weight of equipment health.

[0011] Preferably, in step 3, the multi-level expert weighting system also includes a dynamic weighting adjustment mechanism: Set a weight adjustment factor based on operating conditions. When the equipment load rate exceeds the preset high load threshold, this factor is greater than one to improve the sensitivity of the severity weight of the failure consequences. Set a trend-based weight adjustment factor. When the rate of change of temperature rise exceeds the preset warning threshold, the factor is greater than one to strengthen the basic weight of equipment health. Set a time-based weight decay factor, which decreases as the time interval since the last weight verification update increases, and applies to the credibility weight of the monitoring data. The dynamically corrected weight vector replaces the original static expert weight vector in the fusion game model.

[0012] Preferably, the remaining available decision time in step 5 characterizes the emergency response buffering capacity of the equipment during the development of thermal faults. This remaining available decision time shortens as the temperature rise change characteristics in the multidimensional temperature rise feature vector increase, and decreases as the temperature rise tolerance limit decreases.

[0013] Preferably, the iterative correction of model parameters in step 6 includes: comparing the deviation between the execution effect feedback data and the comprehensive thermal failure risk index and the remaining available decision time, identifying the adaptive adjustment direction of parameters in the temperature rise benchmark model, the multi-level expert weight system and the fusion game model, so that the generation of thermal failure risk assessment and control strategy in the next cycle is closer to the actual operation and maintenance effect.

[0014] Preferably, the differentiated full-cycle management and control strategy in step 6 includes monitoring frequency adjustment plan, maintenance window arrangement suggestion, load control suggestion, and emergency response triggering conditions; The comprehensive risk index for thermal failures corresponds to a preset four-level risk classification: When the comprehensive risk index of thermal failure is less than the first threshold, it is considered to be in a normal state, and the monitoring frequency is maintained at the regular cycle. When the comprehensive risk index of thermal failure is between the first threshold and the second threshold, it is determined to be in a state of concern, the monitoring frequency is increased and trend tracking is triggered. When the comprehensive risk index of thermal failure is between the second and third thresholds, it is determined to be in a warning state, a planned maintenance scheme is formulated and a power outage window is applied for. When the comprehensive risk index of thermal failure is greater than or equal to the third threshold, it is determined to be an alarm state, and load reduction operation or emergency power outage shall be initiated immediately.

[0015] The technical effects and advantages of this invention are as follows: In this invention, by establishing a temperature rise benchmark model deeply bound to individual equipment, the traditional fixed threshold criterion is transformed into a tolerance limit function that decays exponentially with the number of years of operation. This ensures that the temperature rise assessment standard for each piece of equipment is precisely matched with its actual aging state. During real-time operation, the system continuously collects temperature rise change characteristics and load condition data, dynamically compares the current temperature rise level with the tolerance limit after decay correction, calculates the remaining temperature rise margin, and comprehensively determines the temperature rise growth rate based on the temperature rise change rate and the risk trend factor output by the fusion game model. The remaining available decision time is then calculated in real time by using the ratio of the remaining temperature rise margin to the temperature rise growth rate. This principle enables maintenance personnel to go from merely knowing the defect level to accurately grasping the emergency response window duration, completely bridging the time information gap between general threshold standards and the emergency response needs of individual equipment.

[0016] In this invention, objective temperature rise data and expert subjective experience are placed in the same game framework. The risk assessment results on the temperature rise side and the risk assessment results on the weight side are calculated separately, and the degree of conflict between the two is used as a quantitative representation of the confidence level of the assessment conclusion. When the degree of conflict exceeds a preset threshold, an arbitration mechanism is triggered. The data collection strategy is automatically adjusted or the expert weight composition is modified according to the direction of conflict. This ensures that the uncertainty information in the assessment process is retained and transmitted to the control and execution stage in the form of conflict arbitration decision. Based on the combined judgment of the arbitration result and the risk level, the system generates differentiated control strategies that include monitoring frequency adjustment, load control suggestions, and emergency response triggering conditions. High-risk, low-confidence and low-risk, low-confidence scenarios receive different response methods. The principle completely eliminates the decision-making blind spot caused by the illusion of certainty due to the cross-layer discarding of confidence information. Attached Figure Description

[0017] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts: Figure 1 This is a schematic diagram of the full-cycle closed-loop control logic of the present invention; Figure 2 This is a three-dimensional surface diagram of the temperature rise reference model of the present invention; Figure 3 This is the decision matrix for the differentiated management and control strategy of the present invention; Figure 4 This is an infrared thermal imaging field detection image of the present invention; Figure caption: The color scale on the right is a temperature scale, with a value range of 0.9-26.5. The brighter the color, the higher the temperature, and the darker the color, the lower the temperature. The black box in the figure marks the key monitoring area of ​​the equipment. The label "M2H24.9L7.1" is the core temperature measurement data of this monitoring area, with the highest temperature being 24.9℃ and the lowest temperature being 7.1℃. The red cross marks the key temperature measurement points of the equipment, which can accurately locate local overheating areas. Detailed Implementation

[0018] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, some of the terms involved in this application will be explained first.

[0020] The multi-level expert weighting system refers to a multi-dimensional weighting evaluation framework constructed by the operation and maintenance expert team based on equipment operation and management experience. This framework comprehensively evaluates the risk of thermal failures from four dimensions: equipment importance, equipment health foundation, severity of failure consequences, and reliability of monitoring data. The expert weight vector is a mathematical vector formed by arranging the weight values ​​of each dimension in the multi-level expert weighting system in a preset order. This vector is used as one of the input parameters of the fusion game model to participate in the calculation of the comprehensive risk index of thermal failures. The equipment importance weight is a quantitative indicator of the importance of the equipment based on its voltage level, power supply range, and the number of users affected by the failure in the power system. The equipment health foundation weight is a quantitative indicator of the health status based on the equipment's years of operation, frequency of historical failure records, last maintenance time, and maintenance quality score. The severity of failure consequences weight is a quantitative indicator of the severity of consequences based on the level of personal safety risks that equipment failure may cause, the scope of power outage impact, the time of power restoration, and direct economic losses. The reliability of monitoring data weight is a quantitative indicator of the reliability of the data based on the accuracy level of temperature measurement methods, data integrity rate, and the degree of electromagnetic interference in the operating environment.

[0021] The temperature rise benchmark model is a mathematical model established based on historical health operation data of equipment to describe the temperature rise characteristics of equipment under normal operating conditions. This model includes three components: the normal operating condition temperature rise surface, the statistical distribution characteristics of temperature rise of similar equipment, and the temperature rise tolerance limit that decreases with the number of years of operation. The normal operating condition temperature rise surface is a three-dimensional surface function obtained by fitting temperature rise samples under different load rate and ambient temperature combinations using a multivariate nonlinear regression method. The statistical distribution characteristics of temperature rise of similar equipment refer to the mean and standard deviation statistics of temperature rise data of similar equipment of the same model and batch under various operating conditions, which serve as a horizontal comparison benchmark for detecting temperature rise anomalies. The temperature rise tolerance limit is the maximum temperature rise value that power equipment can withstand under normal operating conditions. This value gradually decreases with the increase of the number of years of equipment operation and the deepening of the thermal aging of insulation materials. The temperature rise tolerance limit relationship that decreases with the number of years of operation refers to the functional relationship between the tolerance limit constructed based on the number of years of equipment operation and the thermal aging characteristics of insulation materials and the number of years of operation.

[0022] The multidimensional temperature rise feature vector is a mathematical vector composed of multiple temperature rise feature indicators. Each dimension describes the current thermal state of the equipment from different perspectives. Absolute temperature rise refers to the difference between the measured temperature of the equipment and the ambient temperature at the same time. Relative temperature difference refers to the ratio of the current absolute temperature rise of the equipment to the statistical average temperature rise of the same type of equipment under the same operating conditions. Temperature rise change rate refers to the change in temperature rise value per unit time, which is determined by the slope obtained by linear fitting of temperature rise sequence data within a preset time window. Three-phase temperature rise imbalance refers to the difference between the maximum and minimum temperature rise values ​​of each phase in three-phase power equipment, which is used to reflect the degree of three-phase load balance and contact resistance differences. Hot spot area feature refers to the area value of the local overheated area extracted after gradient analysis and boundary tracking of infrared thermal imaging images, which is used to characterize the spatial distribution range of thermal faults. Temperature rise change feature refers to the feature dimension in the multidimensional temperature rise feature vector that reflects the trend of temperature rise over time, which corresponds to the temperature rise change rate in specific implementation.

[0023] Related parameter data refers to the set of auxiliary parameters collected synchronously with the temperature rise data, including load current, ambient temperature, ambient humidity, and basic equipment information. Load current refers to the actual current value carried by the equipment during operation, ambient temperature refers to the air temperature in the equipment installation area, and ambient humidity refers to the relative humidity of the air in the equipment installation area. Basic equipment information includes equipment model, rated parameters, commissioning date, and historical maintenance records. Real-time temperature rise data refers to the temperature measurement values ​​collected in real time by temperature sensors deployed in key parts of power equipment. Typical acquisition methods include distributed fiber optic temperature measurement, surface acoustic wave wireless temperature measurement, and infrared thermal imaging temperature measurement. Temperature rise time series dataset refers to the set of temperature rise and related parameter data arranged in chronological order after timestamp alignment, missing value completion, outlier removal, and standardization.

[0024] The fusion game model refers to a mathematical model that integrates multidimensional temperature rise feature vectors and expert weight vectors within the same game framework for adversarial coordination. The temperature rise-side risk assessment result refers to the quantified risk value obtained by mapping the multidimensional temperature rise feature vectors through a pre-defined temperature rise risk assessment function. The weight-side risk assessment result refers to the quantified prior risk value obtained by mapping the expert weight vectors through a pre-defined weight activation function. The degree of conflict refers to the quantified index of the difference between the temperature rise-side risk assessment result and the weight-side risk assessment result, usually expressed as the absolute value of their difference. The pre-defined conflict threshold refers to the critical value for the degree of conflict set to trigger the arbitration mechanism. Arbitration refers to the process where, when the conflict reaches a certain threshold... The mediation mechanism is activated when the temperature rise exceeds the preset conflict threshold. It adjusts the data acquisition strategy or modifies the composition of the expert weight vector according to the direction of the conflict. The data acquisition strategy refers to the configuration scheme of the frequency and method of temperature rise data acquisition. Modifying the composition of the expert weight vector refers to the behavior of adjusting the weight values ​​of each dimension or their adjustment factors in the multi-level expert weight system. Nash equilibrium refers to the strategy combination state reached by the game participants in a non-cooperative game in game theory. In this state, any unilateral change of strategy cannot obtain higher returns. The optimal fusion coefficient refers to the weighting coefficient of each dimension of temperature rise characteristics obtained by solving the Nash equilibrium in the calculation of the comprehensive risk index of thermal failure.

[0025] The comprehensive thermal failure risk index is a comprehensive indicator output by the fusion game model to quantify the probability of thermal failures in equipment. The index ranges from zero to one, with a higher value indicating a higher risk of thermal failure. The actual operating load curve refers to the curve of load current changing with time during actual operation. The individual aging state of the equipment refers to the degree of insulation performance degradation determined by the operating years, historical operating conditions, and maintenance of a specific piece of equipment. The thermal breakdown critical state refers to the state corresponding to the temperature critical point at which the insulation material of the equipment is irreversibly damaged due to continuous temperature rise. The remaining available decision time is the estimated value of the time required for the current temperature rise state to develop into the thermal breakdown critical state. This time value is determined by comprehensively considering the current temperature rise level, temperature rise trend, individual equipment aging degree, and actual operating load conditions.

[0026] The differentiated full-cycle management and control strategy refers to a comprehensive management and control plan generated based on the combination of the comprehensive risk index of thermal failure and the remaining available decision time. This plan includes monitoring frequency adjustment, maintenance window arrangement, load control, and emergency response triggering conditions. Execution instructions refer to the transformation of the management and control strategy into standardized task scheduling commands that can be recognized and executed by the operation and maintenance system and the scheduling automation system. Iterative correction refers to the process of continuously optimizing model parameters based on the deviation between the execution effect feedback data and the expected effect. The knowledge base refers to a database system used to store structured information of historical thermal failure management and control cases, which can be retrieved and called. The closed loop refers to the complete management and control cycle from state perception, risk assessment, strategy generation, instruction execution, effect feedback to model optimization. The results of each cycle back-optimize the initial parameters of the next cycle.

[0027] The Delphi method refers to an expert weighting method that solicits expert opinions through multiple rounds of anonymous questionnaires and gradually converges to reach a consensus. The Analytic Hierarchy Process (AHP) is a multi-criteria decision-making method that decomposes a decision problem into target, criterion, and alternative layers and determines the weights of each factor through pairwise comparison judgment matrices. Multiple nonlinear regression is a statistical analysis method that establishes nonlinear functional relationships between the dependent variable and multiple independent variables. Gradient analysis is a method that calculates temperature change gradients in infrared thermal imaging images to identify areas of rapid temperature change. Boundary tracking is an image processing technique that extracts and depicts the contours of hot spot regions in infrared thermal imaging images. Digital twin simulation technology is a technique that establishes a virtual mapping model based on the equipment's three-dimensional geometric model, material physical parameters, and operating boundary conditions, and predicts the equipment's operating status through simulation calculations. In survival analysis, the proportional hazards model is a semi-parametric regression model used to analyze the relationship between survival time data and multiple covariates.

[0028] Example 1: Refer to Figure 1-4 As shown, the present invention provides a technical solution: a method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights. This embodiment is the simplest and most feasible implementation method, and discloses all the necessary technical features of the independent claims.

[0029] Step S1: Construct a temperature rise sensing system and collect data. The existing power equipment thermal fault management field has formed a technical system based on infrared thermometry and supplemented by various sensing methods. Distributed fiber optic temperature sensors are installed at the external flange of the GIS isolating switch gas chamber in this substation to achieve continuous online monitoring of the surface temperature of the contact area. The temperature measurement accuracy is ±1 degree Celsius, and the sampling period is set to one minute. Simultaneously, load current data for this GIS bay is obtained through the substation's SCADA system, ambient temperature and humidity data are obtained through the environmental monitoring system, and basic equipment information such as the GIS equipment's commissioning year, rated parameters, and historical maintenance records are obtained from the equipment ledger system. The collected data undergoes timestamp alignment processing at edge computing nodes, and missing data is supplemented using cubic spline interpolation. Abnormal jump data is detected and removed according to the three sigma principle. After ambient temperature compensation correction of the infrared thermometry data, a standardized temperature rise time series dataset is formed. Through the above multi-source data collection and standardization processing, a complete and reliable temperature rise time series dataset is obtained, laying the data foundation for subsequent personalized benchmark modeling and feature extraction.

[0030] Step S2: Establish a personalized temperature rise benchmark model and calculate a multidimensional temperature rise feature vector. Existing technologies typically use a uniform temperature rise threshold standard for defect level determination, failing to fully consider the varying impacts of different equipment types, operating years, and load levels on temperature rise tolerance. Temperature rise samples under different load rate and ambient temperature combinations were extracted from historical data of the GIS isolating switch during its healthy operation over the past three years. Load rates were set at 30%, 50%, 70%, and 90%, and ambient temperatures at -10℃, 0℃, 10℃, 20℃, 30℃, and 40℃. A multivariate nonlinear regression method was used to fit the normal operating condition temperature rise surface. Simultaneously, historical temperature rise data from six GIS isolating switches of the same model and batch in the substation were retrieved, and the mean and standard deviation of temperature rise under each operating condition were calculated as a horizontal comparison benchmark. Based on the equipment's fifteen years of operation and the thermal aging characteristics of the insulation material, a temperature rise tolerance limit relationship with decreasing operating years was constructed. The factory-rated temperature rise limit is 65 Kelvin. After aging and attenuation, the current effective temperature rise tolerance limit is approximately 45 Kelvin. The real-time collected temperature rise data is compared with the established temperature rise benchmark model, and the values ​​of each dimension in the multidimensional temperature rise feature vector are calculated, including absolute temperature rise of 22 Kelvin, relative temperature difference of 1.47, temperature rise change rate of 0.18 Kelvin per minute, three-phase temperature rise imbalance of 4 Kelvin, and hot spot area characteristic of 12 square centimeters. Through the above steps, a temperature rise benchmark model bound to the specific characteristics of the equipment is established, realizing the transformation from a general standard threshold to a personalized benchmark. The multidimensional temperature rise feature vector comprehensively reflects all aspects of the current thermal state of the equipment.

[0031] like Figure 2As shown, this figure is a three-dimensional surface plot of the temperature rise benchmark model constructed based on multivariate nonlinear regression fitting. The X-axis represents the equipment load rate (range 30%-90%), the Y-axis represents the ambient temperature (range -10℃ to 40℃), and the Z-axis represents the equipment temperature rise (unit: K). The gradient colored surface at the bottom of the figure is the temperature rise benchmark surface under normal operating conditions, which intuitively presents the normal temperature rise distribution law of the equipment under different load rate and ambient temperature combinations. The more yellow the color, the higher the temperature rise, and the more black the color, the lower the temperature rise. The red plane is the current effective temperature rise tolerance limit (45K) of the equipment after 15 years of operation and insulation aging and attenuation. The blue dashed plane is the factory rated temperature rise tolerance limit (65K) of the new equipment, which clearly shows the characteristic that the equipment temperature rise tolerance limit decreases with the increase of operating years. The green pentagram marks the actual operating point corresponding to the example, which can intuitively compare the relative position of the actual temperature rise of the equipment under the current operating conditions with the benchmark surface and tolerance limit, providing a visual benchmark for judging temperature rise anomalies and calculating the remaining temperature rise margin.

[0032] Step S3: Construct a multi-level expert weighting system. Existing technologies often present expert experience as qualitative references or static weights, lacking an effective framework for systematically and quantitatively expressing the long-term accumulated experience and knowledge of operation and maintenance experts. To address the management needs of this GIS isolating switch, a multi-level expert weighting system is constructed, comprising four dimensions: equipment importance weight, equipment health baseline weight, fault consequence severity weight, and monitoring data reliability weight. This equipment connects to a 110kV busbar; a fault will cause a power outage on the entire busbar and affect the power supply to three downstream substations. The equipment importance weight is rated at 0.95. This equipment has been in operation for fifteen years and has experienced two contact overheating defects since its commissioning. The last overhaul was four years ago. The equipment health baseline... The basic weight is set at 0.65. Considering the potential power outage range, recovery time, and impact on downstream industrial users, the severity of the fault consequences is set at 0.85. The current distributed fiber optic grating temperature measurement accuracy is ±1 degree Celsius, the data integrity rate is 98.5%, and the electromagnetic interference in the operating environment is relatively small. The reliability of the monitoring data is set at 0.90. The weight values ​​of the above four dimensions were determined by five experts with more than ten years of experience in GIS equipment operation and maintenance through the Delphi method after two rounds of questionnaires and consensus convergence, forming an expert weight vector. Through the above steps, the long-term accumulated experience and knowledge of operation and maintenance experts are transformed into quantitative weight parameters that can participate in mathematical calculations, realizing a structured mapping from subjective experience to objective weights.

[0033] Step S4: Establish a fusion game model and calculate the comprehensive risk index of thermal failure. Existing assessment methods either overly rely on data-driven single-dimensional analysis or still depend mainly on manual experience judgment. They lack an effective framework for systematically and quantitatively integrating massive temperature rise monitoring data with the experience and knowledge of operation and maintenance experts. The multi-dimensional temperature rise feature vector calculated in step S2 and the expert weight vector formed in step S3 are placed in the same game framework. The temperature rise-side strategy is to assess the risk of thermal failure based on the temperature rise features of each dimension. For absolute temperature rise, a trapezoidal membership function mapping is used. In the range of 0 to 1, the mapping value is zero when the absolute temperature rise is less than 10 Kelvin, linearly increasing to 0.3 when it is between 10 and 15 Kelvin, linearly increasing to 0.7 when it is between 15 and 25 Kelvin, and reaching 1 when it reaches 25 Kelvin or above. An S-shaped function is used for mapping relative temperature difference, and a linearly truncated mapping is used for the rate of temperature change. The weighted strategy involves adjusting the risk assessment results through expert weights, calculating the degree of conflict between the temperature rise-side risk assessment results and the weighted-side risk assessment results. This degree of conflict is determined by the interaction between the two... The absolute value of the difference was quantified. The calculated risk assessment result on the temperature rise side was approximately 0.68, while the risk assessment result on the weight side was approximately 0.41, resulting in a conflict level of 0.27. This exceeded the preset conflict threshold of 0.25, triggering the arbitration mechanism. Because the risk assessment result on the temperature rise side was higher than that on the weight side, the monitoring data collection frequency was automatically increased from once per minute to once every 30 seconds, and a review request was pushed to the operation and maintenance expert terminal. After arbitration adjustment, the optimal fusion coefficients for each dimension of temperature rise characteristics were obtained by iteratively solving the Nash equilibrium point of the game model. The characteristic coefficients for temperature rise, relative temperature difference, temperature rise change, three-phase imbalance, and hot spot area are 0.25, respectively. Based on the optimal fusion coefficients, the comprehensive risk index for thermal failure is calculated to be 0.62. The above-mentioned fusion game process avoids the neglect of individual differences that may be caused by relying solely on temperature rise data, and also prevents the subjective bias that may be caused by relying solely on expert experience. At the same time, the uncertainty in the assessment process is transformed into differentiated control responses through the conflict arbitration mechanism.

[0034] Step S5: Calculate the remaining available decision-making time. The existing temperature rise monitoring threshold standards cannot answer the crucial question of how much available decision-making time remains between the current temperature rise state and thermal breakdown failure. Based on the current value of the comprehensive thermal failure risk index (0.62) calculated in Step S4 and the temperature rise change characteristic of 0.18 Kelvin per minute in the multi-dimensional temperature rise feature vector, combined with the temperature rise tolerance limit of 45 Kelvin after age-related degradation correction in the temperature rise benchmark model established in Step S2, the remaining available decision-making time is calculated. The difference between the age-related degradation-corrected temperature rise tolerance limit and the current absolute temperature rise is used as the remaining temperature rise margin. The current absolute temperature rise is 22 Kelvin. The remaining temperature rise margin is 23 Kelvin. The temperature rise growth rate is calculated by multiplying the current temperature rise rate by a trend acceleration factor that considers the temperature rise acceleration effect. The trend acceleration factor is set to 1.15, and the temperature rise growth rate is 0.207 Kelvin per minute. The remaining available decision time is approximately 111 minutes, determined by the ratio of the remaining temperature rise margin to the temperature rise growth rate. This remaining available decision time characterizes the equipment's emergency response buffer capacity during the development of a thermal fault. This time decreases as the temperature rise change characteristics increase and decreases as the temperature rise tolerance limit decreases. It reflects the comprehensive impact of the equipment's individual aging state and real-time operating conditions on the development process of the thermal fault, providing an accurate time-quantitative basis for operation and maintenance decisions.

[0035] Step S6: Generate a differentiated control strategy and form a closed loop. Existing control methods separate monitoring and alarm, fault diagnosis, maintenance decision-making, maintenance execution, and effect verification into independent functional modules, failing to achieve a seamless process from abnormal temperature rise detection to maintenance strategy execution and effect feedback. Based on the thermal fault comprehensive risk index of 0.62 calculated in step S4 and the remaining available decision time of 111 minutes calculated in step S5, a differentiated full-cycle control strategy is generated. The thermal fault comprehensive risk index of 0.62 is between the second threshold of 0.5 and the third threshold of 0.7, indicating a warning state. The plan includes: a monitoring frequency adjustment scheme to increase the sampling frequency to continuous monitoring every 30 seconds and trigger an infrared temperature measurement robot to perform temporary inspections; a maintenance window arrangement suggestion, as the remaining available decision-making time is less than two hours, and the conditions for applying for a regular power outage are not met, it is recommended to activate the emergency response process; a load control suggestion, reducing the load current from 680 amps to below 500 amps to slow down the rate of temperature rise; and emergency response triggering conditions, if the rate of temperature rise does not decrease significantly within 30 minutes after load reduction or the risk index rises above 0.70, an emergency power outage should be immediately implemented, transforming the above control strategies into structured tasks. The dispatch execution command generates a load control application form and pushes it to the dispatch automation system; it generates an encrypted monitoring command and pushes it to the online monitoring system; and it generates an emergency response order and pushes it to the mobile terminal of the maintenance team. Fifteen minutes after the load reduction command is executed, the load current drops to 495 amperes, the temperature rise rate drops from 0.18 Kelvin per minute to 0.10 Kelvin per minute, and the risk index drops from 0.62 micrograms to 0.58, verifying the effectiveness of the load control measures. The execution effect feedback data is compared with the comprehensive thermal fault risk index and the remaining available decision time to identify the negative factors in the temperature rise benchmark model. The direction of adjusting the load-temperature rise coupling coefficient and the rule for determining the value of the trend acceleration factor are modified. The structured information of this control case includes the initial temperature rise feature vector, expert weight vector, parameters of the fusion game process, risk index evolution sequence, control strategy content and evaluation of the handling effect. This information is stored in the knowledge base and marked with case identifier, equipment type, and the root cause of the fault is insufficient contact pressure due to aging of the contact spring. This provides knowledge reference for the control of similar equipment in the future. Through the above closed-loop mechanism, the full-cycle self-evolution of control is realized from state perception to risk assessment to strategy execution to effect feedback and model optimization.

[0036] like Figure 3As shown, the control strategy decision matrix is ​​constructed based on the comprehensive risk index of thermal failure and the remaining available decision time. The X-axis represents the comprehensive risk index of thermal failure (range 0-1.0), with a higher value indicating a higher risk of thermal failure; the Y-axis represents the remaining available decision time (unit: minutes), with a lower value indicating a shorter emergency response buffer time. The figure is divided into 5 control zones according to the combination of risk index and remaining time: ① Green normal state zone (risk index < 0.3), implementing routine monitoring strategies; ② Yellow attention state zone (0.3 ≤ risk index < 0.5), implementing intensified monitoring and trend tracking strategies; ③ Orange warning state zone (0.3 ≤ risk index < 0.7), formulating planned maintenance schemes and applying for power outage windows; ④ Red alarm state zone (0.5 ≤ risk index < 1.0), immediately initiating load reduction or emergency power outage measures; ⑤ Dark red emergency state zone (risk index ≥ 0.7 and remaining time < 25 minutes), implementing mandatory emergency power outage measures. The blue pentagram in the figure marks the operating point of the example, corresponding to a comprehensive thermal failure risk index of 0.62 and a remaining available decision time of 111 minutes. It is in the warning state zone, which is completely consistent with the generation logic of the control strategy in the example, providing operation and maintenance personnel with intuitive decision-making basis and strategy mapping rules.

[0037] Example 2: Based on Example 1, this example further discloses the additional technical features of the dynamic weight adjustment mechanism in the dependent claims, and takes the control process of the same GIS isolating switch under another operating condition as an example for illustration.

[0038] The difference between this embodiment and Embodiment 1 lies in that the multi-level expert weight system also includes a dynamic weight adjustment mechanism. Existing variable weight methods mostly adjust weights based on the macroscopic division of equipment operation stages, failing to achieve refined dynamic adjustment linked to real-time operating conditions and temperature rise trends. After forming the static expert weight vector in step S3, a three-factor dynamic adjustment mechanism is introduced. A weight adjustment factor based on operating conditions is set. When the equipment load rate exceeds the preset high load threshold of 80%, this factor takes a value greater than one, specifically one plus 0.2 multiplied by the difference between the load rate and 0.8, used to improve the sensitivity of the severity weight of the fault consequences. The current load rate is 85%, and this factor takes a value of 1.01, dynamically adjusting the severity weight of the fault consequences from 0.85 to 0.86. A weight adjustment factor based on trends is set. When the rate of temperature change exceeds the preset warning threshold of 0.15 Kelvin per minute, this factor takes a value greater than one, specifically one plus 0.3 multiplied by the rate of temperature change and 0. The difference of 1.5 divided by the upper limit cutoff value of 0.35 is used to strengthen the response intensity of the basic weight of equipment health. The current temperature rise rate is 0.18 Kelvin per minute, and the factor is set to 1.026. The basic weight of equipment health is dynamically adjusted from 0.65 to 0.67. A time-based weight decay factor is set, which decreases as the time interval between the last weight verification update increases. This factor is applied to the credibility weight of monitoring data. Six months have passed since the last weight assessment. The decay time constant is set to twelve months, and the decay factor is 0.61. The credibility weight of monitoring data is dynamically adjusted from 0.90 to 0.55. The dynamically corrected weight vector replaces the original static expert weight vector and is input into the fusion game model in step S4. The above dynamic weight adjustment mechanism enables the expert weight to evolve in real time with the equipment operating conditions, temperature rise trend, and time. It realizes the transformation of expert experience from static and fixed values ​​to dynamic and adaptive living knowledge. The remaining steps are the same as in Example 1.

[0039] Example 3: This example discloses alternative implementations of the technical solution, covering alternative construction methods for the temperature rise baseline model and alternative calculation methods for the remaining available decision time.

[0040] The first difference between this embodiment and Embodiment 1 is that the temperature rise benchmark model is constructed using digital twin simulation technology instead of historical data regression fitting. When the equipment to be managed is newly commissioned or lacks sufficient historical healthy operation data, a thermoelectric coupling finite element simulation model is established based on the three-dimensional geometric model of the power equipment to be managed, the material thermophysical parameters, and the operating boundary conditions. The material thermophysical parameters include the conductivity of the contact material, the design value of the contact resistance, the thermal conductivity coefficient and the convective heat transfer coefficient of the insulating gas sulfur hexafluoride, and the operating boundary conditions include the load current, the ambient temperature, and the heat dissipation conditions. The theoretical normal temperature rise value under different load rate and ambient temperature combination conditions is obtained through simulation calculation. This simulated temperature rise value is used to replace the expected normal temperature rise value obtained by historical data regression fitting. In the subsequent calculation of the multidimensional temperature rise feature vector, the absolute temperature rise is taken as the difference between the actual measured value and the simulation benchmark value.

[0041] The second difference between this embodiment and Embodiment 1 is that the remaining available decision time uses a proportional hazards model from survival analysis instead of linear extrapolation. Based on a historical thermal failure case library of similar equipment, a risk function with a multidimensional temperature rise feature vector as covariates is constructed. The regression coefficients of each covariate are determined by maximum likelihood estimation. The failure probability density function of the equipment under the current temperature rise state is calculated. The time when the failure probability density reaches a preset critical value is taken as the predicted failure time, thereby obtaining the remaining available decision time. This alternative method makes full use of the statistical regularity of historical failure cases and can provide a confidence interval estimate of the remaining time.

[0042] Compared with the prior art, the present invention has the following advantages: This addresses the issue of temperature rise alarms lacking remaining available decision-making time tied to individual equipment characteristics. By establishing a personalized temperature rise benchmark model that includes temperature rise tolerance limits that decay with operating years, and combining the temperature rise change characteristics in the multi-dimensional temperature rise feature vector with the current actual operating load curve, the remaining available decision-making time is calculated to be deeply tied to the aging state of the individual equipment. Maintenance personnel are upgraded from only knowing the defect level to accurately grasping the remaining emergency response window time, and decisions are no longer passively invalidated due to a lack of time buffer for prediction.

[0043] This eliminates the decision-making blind spot caused by the cross-layer discarding of assessment confidence. By establishing a fusion game model of temperature rise and expert weights, the degree of conflict between the risk assessment results on the temperature rise side and the risk assessment results on the weight side is calculated. When the degree of conflict exceeds a preset threshold, arbitration is triggered and the data collection strategy is adjusted or the weight composition is modified. This mechanism retains the confidence information generated during the assessment process in the form of the degree of conflict and transmits it to the decision-making stage. High-risk, low-confidence false alarms and low-risk, low-confidence missed alarms are treated differently. On-site operation and maintenance personnel no longer fall into the decision-making blind spot under the illusion of certainty when performing alarm handling mapping.

[0044] It achieves a closed-loop self-evolution throughout the entire lifecycle, from state perception to knowledge feedback. By transforming control strategies into execution instructions and tracking execution effects, it iteratively corrects the parameters of the temperature rise benchmark model, multi-level expert weight system, and fusion game model based on the execution effect feedback data. Control cases are stored in the knowledge base to form a closed loop. The system's control capabilities are continuously optimized as the running time accumulates and the cases become richer, achieving a self-evolution effect that becomes more accurate the more it is used.

[0045] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights, characterized in that, Includes the following steps: S1: Collect real-time temperature rise data and related parameter data of the power equipment to be managed, and preprocess them to form a temperature rise time series dataset; S2: Establish a temperature rise benchmark model based on the historical health operation data of the equipment. The temperature rise benchmark model includes a temperature rise tolerance limit that decreases with the number of years of operation. Compare the real-time temperature rise data with the temperature rise benchmark model and calculate a multi-dimensional temperature rise feature vector. S3: Construct a multi-level expert weight system to form an expert weight vector; S4: Establish a game model that integrates temperature rise and expert weights. Place the multidimensional temperature rise feature vector and the expert weight vector in the same game framework. Calculate the degree of conflict between the temperature rise-side risk assessment result obtained by mapping the multidimensional temperature rise feature vector and the weight-side risk assessment result obtained by mapping the expert weight vector. When the degree of conflict exceeds a preset conflict threshold, trigger arbitration and adjust the data acquisition strategy or modify the composition of the expert weight vector according to the arbitration result. Obtain the optimal fusion coefficient by solving the Nash equilibrium and calculate the comprehensive thermal failure risk index based on the optimal fusion coefficient. S5: Based on the current value of the comprehensive thermal failure risk index and the temperature rise change characteristics in the multidimensional temperature rise feature vector, combined with the temperature rise tolerance limit in the temperature rise benchmark model, calculate the remaining available decision time from the current temperature rise state to the thermal breakdown critical state under the combined effect of the current actual operating load curve and the individual aging state of the equipment. S6: Based on the comprehensive risk index of thermal failure and the remaining available decision time, generate differentiated full-cycle management and control strategies, convert management and control strategies into execution instructions and push them to the operation and maintenance system and track the execution effect. Based on the execution effect feedback data, iteratively correct the parameters of the temperature rise benchmark model, multi-level expert weight system and fusion game model, and store management and control cases in the knowledge base to form a closed loop.

2. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 1, characterized in that: The associated parameter data mentioned in S1 includes load current, ambient temperature, ambient humidity, and basic equipment information; The multidimensional temperature rise feature vector described in S2 includes absolute temperature rise, relative temperature difference, temperature rise rate of change, three-phase temperature rise imbalance, and hot spot area characteristics. The multi-level expert weighting system described in S3 includes weights for equipment importance, basic equipment health, severity of failure consequences, and credibility of monitoring data.

3. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 2, characterized in that: The method for establishing the temperature rise reference model described in S2 is as follows: Temperature rise samples under different load rate and ambient temperature combinations were extracted from the historical healthy operation data of the power equipment to be managed, and the temperature rise surface under normal operating conditions was fitted by multivariate nonlinear regression. Historical temperature rise data of similar equipment of the same model and batch were retrieved, and the mean and standard deviation of temperature rise under each operating condition were calculated as a benchmark for horizontal comparison. Based on the equipment's years of operation and the thermal aging characteristics of the insulation material, a relationship is established between the temperature rise tolerance limit and the decrease in temperature over the years of operation.

4. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 2, characterized in that: The calculation method for each dimension of the multidimensional temperature rise feature vector described in S2 is as follows: The absolute temperature rise is the difference between the measured temperature by the device and the ambient temperature. The relative temperature difference is the ratio of the current absolute temperature rise to the average temperature rise of similar equipment under the same operating conditions; The rate of temperature change is determined by the slope of the linear fitting of the temperature rise sequence within a preset time window before the current moment. The three-phase temperature rise imbalance is taken as the difference between the maximum and minimum temperature rise of the three phases. The hot spot area feature is obtained by extracting the hot spot region area through gradient analysis and boundary tracking of infrared thermal imaging images.

5. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 2, characterized in that: The dimensions of the multi-level expert weighting system described in S3 are defined as follows: The importance weight of the equipment is assessed based on the equipment voltage level, power supply range, and the number of users affected by the fault. The basic weight of equipment health is based on a comprehensive evaluation of the equipment's years of operation, historical failure frequency, last maintenance time, and quality score. The severity weighting of the fault consequences is based on the assessment of the personal safety risks that the fault may cause, the scope of the power outage, the restoration time, and the economic losses. The reliability weight of the monitoring data is assessed based on the accuracy level of the temperature measurement method, the data integrity rate, and the degree of environmental electromagnetic interference.

6. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 1, characterized in that: The specific details of adjusting the data collection strategy or modifying the expert weight vector based on the arbitration result as described in S4 are as follows: If the risk assessment result on the temperature rise side is higher than the risk assessment result on the weight side, the collection frequency of the real-time temperature rise data will be automatically increased and a review request will be pushed to the operation and maintenance expert terminal. If the risk assessment result on the temperature rise side is lower than the risk assessment result on the weight side, then the current data collection frequency is maintained and the growth rate of the adjustment factor corresponding to the basic weight of the equipment health is reduced.

7. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 2, characterized in that: In S3, the multi-level expert weighting system also includes a dynamic weighting adjustment mechanism: Set a weight adjustment factor based on operating conditions. When the equipment load rate exceeds the preset high load threshold, the factor is greater than one to improve the sensitivity of the severity weight of the failure consequences. A trend-based weight adjustment factor is set, which is greater than one when the rate of temperature change exceeds a preset warning threshold, in order to strengthen the basic weight of the equipment health. A time-based weight decay factor is set, which decreases as the time interval between the last weight verification update increases, and applies to the credibility weight of the monitoring data. The dynamically corrected weight vector replaces the original static expert weight vector in the fusion game model.

8. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 1, characterized in that: The remaining available decision time mentioned in S5 characterizes the equipment's emergency response buffering capacity during the development of thermal faults. This remaining available decision time shortens as the temperature rise change characteristics in the multidimensional temperature rise feature vector increase, and decreases as the temperature rise tolerance limit decreases.

9. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 1, characterized in that: The iterative correction of model parameters in S6 includes: comparing the deviation between the execution effect feedback data and the comprehensive thermal failure risk index and the remaining available decision time, identifying the adaptive adjustment direction of parameters in the temperature rise benchmark model, the multi-level expert weight system and the fusion game model, so that the generation of thermal failure risk assessment and control strategy in the next cycle is closer to the actual operation and maintenance effect.

10. The method for full-cycle management and control of thermal faults in power equipment based on the fusion of temperature rise and expert weights as described in claim 1, characterized in that: The differentiated full-cycle management and control strategy described in S6 includes monitoring frequency adjustment schemes, maintenance window arrangement suggestions, load control suggestions, and emergency response triggering conditions; The comprehensive risk index for thermal failures corresponds to a preset four-level risk classification: When the comprehensive risk index of thermal failure is less than the first threshold, it is determined to be in a normal state, and the monitoring frequency is maintained at the regular cycle. When the comprehensive risk index of thermal failure is between the first threshold and the second threshold, it is determined to be in a state of concern, the monitoring frequency is increased and trend tracking is triggered. When the comprehensive risk index of thermal failure is between the second and third thresholds, it is determined to be in a warning state, a planned maintenance scheme is formulated and a power outage window is applied for; When the comprehensive risk index of thermal failure is greater than or equal to the third threshold, it is determined to be an alarm state, and load reduction operation or emergency power outage shall be initiated immediately.

Images