Method and device for on-line evaluation of insulation aging state and life prediction of high-voltage series reactor

Abstract

The embodiment provides a high-voltage series reactor insulation aging state online evaluation and life prediction method and device, a plurality of types of sensors are arranged at insulation positions and electrical parameters and environmental parameters are accessed, temperature field, partial discharge and vibration parameters are collected online, original data are preprocessed to generate multi-source monitoring data, multi-dimensional aging features are extracted from the multi-source monitoring data, and standardized aging feature vectors are generated by standardizing the multi-dimensional aging features relative to individualized benchmarks, a multi-stress comprehensive insulation aging model is constructed according to the relationships of three types of stresses, i.e., thermal aging, electrical aging and mechanical aging, and a sub-item aging degree is generated by periodical integration and accumulation, a comprehensive aging degree is generated after the comprehensive aging degree is integrated and combined with a deviation correction, a worst grade in a monitoring abnormality grade is generated by comparing the comprehensive aging degree with a grading standard to generate an insulation state grade, a remaining service life is calculated by extrapolating a trend of the comprehensive aging degree and a confidence interval is quantified to generate a maintenance decision and a warning, a system deviation is updated according to outage test data and dissection data of retired equipment, and the model is returned to the individualized benchmark.

Description

Technical Field

[0001] This application relates to the field of power equipment condition monitoring and intelligent operation and maintenance technology, specifically to a method and device for online assessment of insulation aging status and life prediction of high-voltage series reactors. Background Technology

[0002] High-voltage series reactors operate for extended periods in environments with high voltage, high current, and strong electromagnetic fields. The insulation system gradually ages under the combined effects of electric field stress, thermal stress, mechanical stress, and environmental stress. Current maintenance methods primarily rely on periodic power outage tests to assess insulation condition, with test cycles typically measured annually. Dielectric loss measurements, insulation resistance measurements, and partial discharge detection are all performed under power outage conditions. This periodic offline testing approach cannot promptly detect insulation degradation between tests, and a significant proportion of insulation faults occur within months of the periodic test, posing a risk of missed fault detection.

[0003] Existing online monitoring systems mostly monitor a single parameter, such as temperature, partial discharge, or vibration. Insulation aging is the result of multiple factors, and a single parameter cannot fully reflect the true aging state of the insulation, resulting in a low detection rate and a high false negative rate. Current assessment methods often use threshold judgment and condition level classification, comparing monitored values ​​with fixed thresholds to determine condition levels such as normal, warning, and abnormal. However, these methods fail to quantify the degree of insulation aging, making it difficult to answer key questions such as how much insulation has aged and how long it can still operate safely, thus hindering the development of refined maintenance decisions.

[0004] Different reactors have varying design parameters, manufacturing processes, operating environments, and load characteristics, resulting in different insulation aging patterns. Existing methods often employ uniform evaluation models and threshold standards, failing to reflect individual differences and leading to significant discrepancies between evaluation results and actual aging conditions. Traditional thermal aging models only consider temperature factors, neglecting electric field aging and mechanical aging mechanisms, resulting in large errors in remaining life prediction and often only providing qualitative replacement recommendations in practical engineering. Furthermore, existing methods lack a mechanism for fusing and calibrating data from power outage tests and decommissioned equipment, making it difficult to continuously correct model accuracy during operation. Therefore, a method for online insulation aging assessment and life prediction that combines multi-source online monitoring, multi-stress aging modeling, individualized benchmark standardization, and power outage test data calibration is needed. Summary of the Invention

[0005] To address the shortcomings of existing maintenance methods that rely on periodic power outage tests for timely detection of insulation degradation during test intervals, online monitoring often focuses on single parameters, assessment methods rely on fixed thresholds to fail to quantify aging levels, and fail to reflect individual differences while lacking lifespan prediction and fusion calibration capabilities, this application provides an online assessment and lifespan prediction method and device for the insulation aging status of high-voltage series reactors. This method can collect insulation aging-related parameters through multiple types of sensors, quantify the comprehensive aging degree through a combination of three stress sub-models (thermal aging, electrical aging, and mechanical aging), adapt to individual equipment differences through individualized benchmark standardization, extrapolate the remaining service life with confidence intervals through aging degree trend, and continuously update the model through fusion calibration of power outage test data and decommissioned equipment data, thereby improving the monitoring coverage, quantification capabilities, and lifespan prediction accuracy of insulation aging.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] Firstly, this application provides a method for online assessment of the insulation aging state and life prediction of a high-voltage series reactor, comprising: deploying multiple types of sensors on the insulation part of the high-voltage series reactor and connecting them to electrical and environmental parameters; collecting temperature field, partial discharge, and vibration parameters according to a preset sampling rule family, and performing preprocessing to generate multi-source monitoring data; extracting multi-dimensional aging features from the multi-source monitoring data and standardizing them relative to individualized benchmarks to generate a standardized aging feature vector; reading the standardized aging feature vector; constructing a multi-stress comprehensive insulation aging model based on the relationship between thermal aging, electrical aging, and mechanical fatigue; substituting the stress history into the model and integrating it periodically to generate sub-item aging degrees; and performing a fusion based on a preset... The rule family integrates the individual aging degrees and corrects for deviations in the multidimensional aging characteristics to generate a comprehensive aging degree. Based on the comprehensive aging degree and the grading standard, the worst level among the monitoring anomaly levels is selected to generate the insulation status level. The evolution trend of the comprehensive aging degree is fitted and extrapolated to the failure threshold condition to calculate the remaining service life and quantify the confidence interval. Based on the insulation status level and the remaining service life matching decision rule family, maintenance decisions and early warnings are generated. Power outage test data and decommissioned equipment data are compared with the insulation status level to identify deviations and generate model calibration values. Based on the model calibration values, the multi-stress comprehensive insulation aging model is updated and fed back to the individualized benchmark.

[0008] Furthermore, it also includes: arranging temperature sensors in the winding encapsulation layer, vibration sensors in the core clamps and encapsulation layer ends, and partial discharge sensors in the body and bushing flanges, based on the insulation structure of the reactor under test; connecting electrical parameters via current transformers and voltage transformers, and connecting environmental parameters including ambient temperature, relative humidity, pollution level, and ultraviolet radiation intensity; setting differentiated sampling periods based on the dynamic characteristics of each parameter to collect and generate raw data online; removing outliers from the raw data according to a physically reasonable range and statistical distribution; interpolating missing segments according to historical patterns; performing noise reduction filtering on partial discharge and vibration signals; and aligning data from different sampling periods to a unified time reference to generate multi-source monitoring data.

[0009] Furthermore, it also includes: extracting the hottest spot temperature and temperature rise value from the temperature dimension of the multi-source monitoring data, extracting the discharge quantity statistics and discharge trend from the partial discharge dimension, extracting the fundamental frequency amplitude and harmonic ratio from the vibration dimension, and extracting the reactance value deviation from the electrical dimension to generate multi-dimensional aging characteristics; selecting the mean and fluctuation range of each of the multi-dimensional aging characteristics during the stable operation period in the initial stage of commissioning to establish an individualized benchmark, subtracting the mean of the individualized benchmark from each of the multi-dimensional aging characteristics and then dividing by the fluctuation range to obtain a standardized aging feature vector, and using the statistical distribution of the same type of equipment as the initial benchmark when the initial commissioning data is missing.

[0010] Furthermore, it also includes: constructing a thermal aging sub-model based on the thermal aging rate relationship of the insulating material; constructing an electrical aging sub-model based on the inverse power law relationship between inter-turn voltage stress and ground voltage stress; and constructing a mechanical aging sub-model based on the fatigue accumulation relationship between electrodynamic cycles and temperature cycles. The thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model. The temperature history corresponding to the standardized aging feature vector is substituted into the thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model, respectively, and integrated periodically. The stress cycle fatigue damage within the cycle is statistically accumulated, and the increments of each cycle are added to generate the thermal aging degree, the electrical aging degree, and the mechanical aging degree, respectively.

[0011] Furthermore, it also includes: integrating the thermal aging degree, electrical aging degree, and mechanical aging degree according to a preset fusion rule family and taking the dominant component among them; combining the deviation of the multi-dimensional aging characteristics from the individualized benchmark and calculating the correction amount according to the accelerating effect of the environmental parameters on the thermal aging rate according to a preset weight to generate a comprehensive aging degree; classifying the aging level according to the comprehensive aging degree against a preset grading standard; independently assessing the monitoring anomaly level based on the deviation of the multi-dimensional aging characteristics; taking the worst level among the aging level and the monitoring anomaly level to generate an insulation state level; and locating key aging parts according to the characteristic deviation distribution of each monitoring point.

[0012] Furthermore, it also includes: fitting the evolution trend of the historical sequence of the comprehensive aging degree in a preset trend model family according to the goodness of fit, estimating the recent average aging rate, and extrapolating the comprehensive aging degree to the failure threshold condition according to the aging rate to calculate the remaining service life; propagating the parameter uncertainty of the multi-stress comprehensive insulation aging model to the remaining service life through random sampling to quantify its confidence interval, and generating maintenance decisions and graded early warnings based on a preset decision rule family that matches the insulation state level with the remaining service life.

[0013] Furthermore, it also includes: acquiring dielectric performance parameters during the power outage test of the tested reactor, acquiring insulation physicochemical test results during the decommissioning and dismantling of reactors of the same type, comparing the power outage test data and the dissection data of the decommissioned equipment with the insulation condition level to identify system deviations and generate model calibration quantities; performing Bayesian updates on the parameters of the multi-stress integrated insulation aging model based on the model calibration quantities, triggering the individualized benchmark reset for reactors that have undergone major repairs, and dynamically re-estimating the individualized benchmark using a sliding benchmark for reactors that have not undergone major repairs for a long time.

[0014] Secondly, this application provides an online assessment and life prediction device for the insulation aging state of a high-voltage series reactor, comprising: a multi-source sensing module, used to deploy multiple types of sensors on the insulation part of the high-voltage series reactor and connect electrical and environmental parameters, and collect temperature field, partial discharge and vibration parameters online according to a preset sampling rule family, and then perform outlier removal, missing value interpolation and time alignment on the original data to generate multi-source monitoring data; a feature extraction module, used to extract multi-dimensional aging features from the multi-source monitoring data according to temperature, partial discharge and vibration dimensions, establish an individualized benchmark based on the stable operation period in the initial stage of commissioning, and standardize the multi-dimensional aging features relative to the individualized benchmark to generate a standardized aging feature vector; and an aging modeling module, used to construct a thermal aging sub-model, an electrical aging sub-model and a mechanical aging sub-model according to the thermal aging rate relationship, the electric field stress inverse power law relationship and the mechanical fatigue accumulation relationship, and jointly constitute a multi-stress comprehensive insulation aging model, and then perform the following on the model: The stress history corresponding to the standardized aging feature vector is substituted into each of the sub-models and accumulated by periodic integration to generate thermal aging degree, electrical aging degree, and mechanical aging degree. The state assessment module is used to integrate the thermal aging degree, electrical aging degree, and mechanical aging degree according to a preset fusion rule family and combine deviation correction to generate a comprehensive aging degree. Based on the comprehensive aging degree, it compares with a preset grading standard and takes the worst grade among the monitored abnormal grades to generate an insulation status grade and locate key aging parts. The life prediction module is used to fit the evolution trend of the historical sequence of the comprehensive aging degree and estimate the current aging rate. It extrapolates the comprehensive aging degree to the failure threshold condition to calculate the remaining service life and quantifies the confidence interval. Based on the insulation status grade and the remaining service life, it generates maintenance decisions and graded early warnings. Based on the power outage test data and decommissioned equipment data, it compares with the insulation status grade to identify deviations, updates the multi-stress comprehensive insulation aging model, and then feeds back to the individualized benchmark.

[0015] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the online assessment and life prediction method for the insulation aging state of the high-voltage series reactor.

[0016] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the online assessment and life prediction method for the insulation aging state of the high-voltage series reactor.

[0017] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the steps of the online assessment and life prediction method for the insulation aging state of the high-voltage series reactor.

[0018] As can be seen from the above technical solution, this application provides a method and device for online assessment of insulation aging status and life prediction of high-voltage series reactors. By deploying multiple types of sensors at the insulation parts and preprocessing the raw data to generate multi-source monitoring data, multi-dimensional aging characteristics are standardized relative to individualized benchmarks to adapt to individual differences of equipment. The comprehensive aging degree is quantified by combining three types of stress sub-models of thermal aging, electrical aging and mechanical aging and combining deviation correction. The insulation status level is generated by comparing the worst level with the grading standard and the monitoring anomaly level. The remaining service life with confidence interval is given by extrapolating the aging degree trend and random sampling, and maintenance decisions and graded early warnings are generated. The system deviation is identified and the multi-stress comprehensive insulation aging model is updated based on power outage test data and decommissioned equipment dissection data, and the dynamic re-estimation is fed back to the individualized benchmark to form a closed loop. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0020] Figure 1 This is a flowchart illustrating the online assessment and lifetime prediction method for the insulation aging state of a high-voltage series reactor in this application embodiment;

[0021] Figure 2 This is a schematic diagram of the structure of the online assessment and life prediction device for insulation aging of high-voltage series reactors in the embodiments of this application;

[0022] Figure 3 This is a schematic diagram of the multi-source sensor arrangement and data preprocessing process for a dry-type air-core reactor in an embodiment of this application.

[0023] Figure 4 This is a schematic diagram illustrating the multidimensional aging feature extraction and multi-stress integrated insulation aging model in the embodiments of this application;

[0024] Figure 5 This is a schematic diagram of the comprehensive aging degree calculation, cumulative update, and insulation condition level assessment process in the embodiments of this application.

[0025] Figure 6 This is a schematic diagram of the closed-loop iterative process for predicting remaining useful life and calibrating power outage test data in the embodiments of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.

[0027] Considering that existing operation and maintenance methods rely on periodic power outage tests, which make it difficult to detect insulation degradation during test intervals in a timely manner, online monitoring often targets a single parameter, and assessment methods often use fixed thresholds to judge without quantifying the degree of aging and lack the ability to predict individual differences and lifespan, this application provides an online assessment and lifespan prediction method for the insulation aging status of high-voltage series reactors. Through multi-source online monitoring, multi-stress aging modeling, individualized benchmark standardization, and power outage test data calibration, the method can perform online quantitative assessment of the insulation aging status and predict the remaining lifespan of high-voltage series reactors.

[0028] This application provides an embodiment of a method for online assessment of insulation aging status and life prediction of high-voltage series reactors, see [link to embodiment]. Figure 1 The method specifically includes the following:

[0029] Step S101: Deploy multiple types of sensors at the insulation part of the high-voltage series reactor and connect them to electrical and environmental parameters. Collect temperature field, partial discharge and vibration parameters according to the preset sampling rule family, and then perform preprocessing to generate multi-source monitoring data. Extract multi-dimensional aging features from the multi-source monitoring data and generate a standardized aging feature vector relative to the individualized benchmark.

[0030] In this embodiment, the method operates within a high-voltage series reactor online monitoring and condition assessment system, such as... Figure 1 As shown, the system adopts a hierarchical structure of multi-source sensing, feature extraction, aging modeling, condition assessment, and lifetime prediction. The multi-source sensing layer deploys temperature, partial discharge, and vibration sensors at the insulation components and connects them to electrical and environmental parameters to achieve online acquisition of insulation aging-related parameters.

[0031] In one feasible implementation, see Figure 3 For dry-type air-core reactors, fiber optic temperature sensors are arranged along the height direction on the surface of the winding encapsulation layer, and 6 to 10 temperature measuring points are evenly arranged along the winding height. For oil-immersed reactors, temperature sensors are arranged at different heights in the oil tank to monitor the oil temperature distribution, while environmental parameters such as ambient temperature, relative humidity, pollution level and ultraviolet radiation intensity are also monitored. The number of each temperature measuring point is determined according to the winding height and structure.

[0032] Partial discharge is monitored using a combination of ultra-high frequency (UHF) and ultrasonic sensors. The UHF sensors are installed at the body and bushing flange, while the ultrasonic sensors are installed on the outer surface of the encapsulation layer and the tank wall. Signals exceeding the trigger threshold are recorded and statistically analyzed. Vibration monitoring uses piezoelectric vibration sensors installed at the support base and the end of the encapsulation layer. Electrical parameters are connected via current and voltage transformers.

[0033] Each parameter has a differentiated sampling period set according to its dynamic characteristics, with electrical parameters having the shortest sampling period, followed by vibration and temperature. Partial discharge is monitored using a triggered continuous monitoring method. The sampling period is calibrated based on the smallest time scale that can capture the corresponding physical process. Sampling loss segments are marked with a sampling missing identifier and transferred to the interpolation process.

[0034] In one specific implementation, outliers are detected and removed from the collected raw data based on their physical reasonable range and statistical distribution. Missing segments are imputed using linear interpolation and historical patterns. Noise reduction filtering is performed on partial discharge and vibration signals to suppress interference. The removed and imputed samples are registered with data quality labels for subsequent feature extraction.

[0035] Data from different sampling periods are aligned to a unified time base to generate multi-source monitoring data. The multi-source monitoring data uses the device identifier as the primary key and organizes temperature, partial discharge, vibration and electrical parameter fields according to timestamps. These fields are then read by timestamp at the feature extraction entry point in step S102. Records that fail to align are registered with alignment anomaly flags and filled with adjacent time values.

[0036] Based on this, the hottest spot temperature and temperature rise value are extracted from the temperature dimension of the multi-source monitoring data; the discharge quantity statistics and discharge trend are extracted from the partial discharge dimension; the fundamental frequency amplitude and harmonic ratio are extracted from the vibration dimension; and the reactance value deviation is extracted from the electrical dimension, thus generating multi-dimensional aging characteristics. During the stable operation period in the initial stage of commissioning, the mean and fluctuation range of each characteristic are statistically analyzed to establish an individualized benchmark. The mean value of each characteristic is subtracted, and then divided by the fluctuation range to obtain a standardized aging feature vector. When initial commissioning data is missing, the statistical distribution of a group of similar equipment is used as the initial benchmark. The standardized aging feature vector is then read in step S102.

[0037] Step S102: Read the standardized aging feature vector, construct a multi-stress comprehensive insulation aging model based on the relationship between thermal aging, electrical aging and mechanical fatigue, and substitute the stress history into the periodic integral to generate the sub-item aging degree. Based on the preset fusion rule family, synthesize the sub-item aging degree and combine it with the deviation of the multi-dimensional aging feature to generate the comprehensive aging degree. According to the comprehensive aging degree, compare it with the grading standard and take the worst level among the monitoring abnormal levels to generate the insulation state level.

[0038] The standardized aging feature vector output from step S101 and the individualized baseline are read into the aging modeling process, such as... Figure 4 As shown. The two are associated with the equipment identifier. When the individualized benchmark lacks the corresponding feature, the benchmark is registered as not having an identifier and the process is transferred to the benchmark supplementation process. The aging model is constructed by building sub-models according to the three types of stress: thermal, electrical and mechanical, and together they form a multi-stress comprehensive insulation aging model.

[0039] The thermal aging sub-model is established based on the thermal aging rate relationship of insulating materials, converting the temperature history within a cycle into the cumulative aging equivalent at a reference temperature. The electrical aging sub-model is established based on the inverse power law relationship between inter-turn voltage stress and ground voltage stress, calculating the equivalent electric field stress based on the actual voltage waveform under harmonic and transient overvoltage conditions. The mechanical aging sub-model is established based on the fatigue accumulation relationship between electrodynamic cycles and temperature cycles.

[0040] In one specific implementation, the thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model. The parameters of each sub-model are selected according to the typical values ​​of the insulation material type. The thermal aging activation energy is 0.8 to 1.3 electron volts, and the electrical aging voltage index is 9 to 12. Alternatively, the parameters can be fitted based on the aging data of similar equipment. If parameter identification fails, the model will revert to the typical values ​​in the material manual and the parameter reversion flag will be registered.

[0041] Stress parameters are extracted in real time based on measured data. Thermal stress is identified from the measured temperature field, with the hottest temperature as the representative temperature for thermal aging. Electrical stress is calculated from the measured voltage and current to determine the inter-turn voltage stress and the voltage stress to ground. Mechanical stress is identified from the measured vibration and current to determine stress cycle events. Each stress parameter is organized with the calculation period as the caliber.

[0042] Within each calculation cycle, the temperature history and electric field stress history are substituted into the thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model, respectively, and integrated periodically. The stress cycle within the cycle is statistically accumulated to obtain the thermal aging degree increment, electrical aging degree increment, and mechanical aging degree increment. The increments of each cycle are summed to generate the thermal aging degree, electrical aging degree, and mechanical aging degree, respectively.

[0043] Based on this, the thermal aging degree, electrical aging degree and mechanical aging degree are combined according to the preset fusion rule family, and the dominant component is selected. The deviation of the multi-dimensional aging characteristics from the individualized benchmark is combined with the preset weight to calculate the correction amount to generate the comprehensive aging degree. When the partial discharge deviation increases, the correction amount is increased. When the relative humidity and dirtiness increase, the thermal aging correction amount is increased according to their accelerating effect on the thermal aging rate. The correction weight is determined based on historical data learning.

[0044] Aging levels are determined based on the comprehensive aging degree compared to a preset grading standard. The monitoring anomaly levels of partial discharge, temperature, and vibration are independently assessed based on the deviation of the multi-dimensional aging characteristics. The worst level between the aging level and the monitoring anomaly level is used to generate the insulation status level. The insulation status level uses the equipment identifier as the primary key, along with the organization level value, dominant aging factors, and key aging location fields. When the difference between the aging level and the monitoring anomaly level exceeds a certain limit, a verification flag is triggered, which is then read at the lifetime prediction entry point in step S103.

[0045] Step S103: Fit the evolution trend of the comprehensive aging degree and extrapolate it to the failure threshold condition to calculate the remaining service life and quantify the confidence interval. Generate maintenance decisions and early warnings based on the decision rule family matching the insulation status level and the remaining service life. Obtain power outage test data and decommissioned equipment data and compare them with the insulation status level to identify deviations and generate model calibration values. Update the multi-stress comprehensive insulation aging model based on the model calibration values ​​and return it to the individualized benchmark.

[0046] The comprehensive aging degree and insulation condition level output from step S102 are read into the life prediction stage, such as... Figure 6 As shown, the historical sequence of the comprehensive aging degree is fitted to the evolution trend by selecting the best fit from a preset family of trend models. The family of trend models includes linear, exponential, and multinomial models, and the model with the best fit is selected as the evolution trend model.

[0047] Based on this, the recent average aging rate is estimated, and it is determined whether the aging rate is accelerating. The overall aging degree is then extrapolated to the failure threshold condition to calculate the baseline remaining service life. When there are future load plans and seasonal operating patterns, the aging rate is adjusted according to the changes in operating conditions, and the adjusted remaining service life is given.

[0048] In one specific implementation, the parameter uncertainty of the multi-stress integrated insulation aging model is propagated to the remaining service life through random sampling. The uncertainty of aging rate fluctuation and future operating conditions is superimposed to quantify the confidence interval. The remaining service life is expressed as a pair of predicted values ​​and confidence intervals, and a corresponding curve of time and failure probability is generated for maintenance planning.

[0049] Maintenance decisions and tiered early warnings are generated based on a preset set of decision rules that match the insulation condition level with the remaining service life. A red warning is triggered when the insulation condition level is severe or the remaining service life is below a preset threshold; an orange warning is triggered when the insulation condition level is poor or the remaining service life is in the middle range; and a trend warning is triggered when the aging rate accelerates abnormally.

[0050] Maintenance decisions are based on the dominant aging factors, providing targeted recommendations. When thermal aging is dominant, the focus is on checking the cooling system and heat dissipation surfaces; when electrical aging is dominant, the focus is on checking the insulation dielectric properties and overvoltage protection; when mechanical aging is dominant, the focus is on checking fasteners and winding deformation; and when partial discharge is abnormal, the focus is on discharge detection and location. Maintenance decisions and tiered early warnings are output to the operation and maintenance management system using equipment identification as the primary key, and the decision records are registered.

[0051] Dielectric performance parameters are obtained during the power outage test of the tested reactor, and insulation physicochemical test results are obtained when the same type of reactor is decommissioned and dismantled. The power outage test data and the dismantled data of the decommissioned equipment are compared with the insulation status level to identify system deviations and generate model calibration values. The comparison results are recorded in the calibration record.

[0052] Accordingly, Bayesian updates are performed on the parameters of the multi-stress integrated insulation aging model based on the model calibration values, and the individualized benchmark is dynamically re-evaluated. Reactors undergoing major repairs trigger a reset of the individualized benchmark, while reactors that have not undergone major repairs for a long period use a sliding benchmark. The updated parameters and individualized benchmarks are fed back to the individualized benchmark in step S101 and the multi-stress integrated insulation aging model in step S102. Equipment of the same model, manufacturer, and operating environment shares calibration knowledge, forming a closed loop of monitoring, evaluation, verification, and updating.

[0053] As can be seen from the above description, the online assessment and life prediction method for the insulation aging status of high-voltage series reactors provided in this application embodiment can cover insulation aging-related parameters through multi-source online monitoring, quantify the comprehensive aging degree through the combination of three types of stress sub-models of thermal aging, electrical aging and mechanical aging, adapt to individual differences of equipment through individualized benchmark standardization, and give the remaining service life with confidence interval through aging degree trend extrapolation.

[0054] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 3 It can also specifically include the following:

[0055] Step S201: Based on the insulation structure of the reactor under test, temperature sensors are arranged in the winding encapsulation layer, vibration sensors are arranged in the core clamp and the end of the encapsulation layer, and partial discharge sensors are arranged in the body and bushing flange. Electrical parameters are connected through current transformers and voltage transformers, and environmental parameters including ambient temperature, relative humidity, pollution level and ultraviolet radiation intensity are also connected. Based on the dynamic characteristics of each parameter, a differentiated sampling period is set to collect and generate raw data online.

[0056] Step S202: Remove outliers from the original data based on the physical reasonable range and statistical distribution, interpolate missing segments according to historical patterns, perform noise reduction filtering on partial discharge and vibration signals, and align data from different sampling periods to a unified time reference to generate multi-source monitoring data.

[0057] In step S101, the multi-source sensing layer arranges sensors according to the insulation structure of the reactor under test, such as... Figure 3As shown, for dry-type air-core reactors, fiber optic temperature sensors are arranged along the height direction on the surface of the winding cladding layer; piezoelectric vibration sensors are arranged at the core clamps and the ends of the cladding layer; and ultra-high frequency and ultrasonic partial discharge sensors are arranged at the body and bushing flange. The operating current and voltage at both ends are connected via current transformers and voltage transformers, and environmental parameters such as ambient temperature, relative humidity, pollution level, and ultraviolet radiation intensity are monitored. Differentiated sampling periods are set for each parameter based on its dynamic characteristics, with electrical parameters having the shortest sampling period, followed by temperature and vibration. Partial discharge is continuously monitored using a trigger-based method, and missing sampling sections are marked with a sampling omission flag.

[0058] The original data is analyzed and outliers are detected and removed based on a physically reasonable range and statistical distribution. Figure 3 As shown, missing segments are interpolated using linear interpolation and historical patterns. Noise reduction filtering is applied to partial discharge and vibration signals to suppress on-site interference, and quality indicators of the interpolated and removed samples are registered. Based on this, data from different sampling periods are aligned to a unified time base to generate multi-source monitoring data. The multi-source monitoring data is organized with the device identifier as the primary key and timestamps for temperature, partial discharge, vibration, and electrical parameter fields. Records with failed time alignment are registered with alignment anomaly indicators and filled with adjacent time values, which are then read by timestamp at the feature extraction entry point in step S301.

[0059] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 4 It can also specifically include the following:

[0060] Step S301: Extract the hottest spot temperature and temperature rise value from the temperature dimension of the multi-source monitoring data, extract the discharge quantity statistics and discharge trend from the partial discharge dimension, extract the fundamental frequency amplitude and harmonic ratio from the vibration dimension, and extract the reactance value deviation from the electrical dimension to generate multi-dimensional aging characteristics.

[0061] Step S302: Select the mean and fluctuation range of each multidimensional aging feature during the stable operation period in the early stage of commissioning to establish an individualized benchmark. Subtract the mean of the individualized benchmark from each multidimensional aging feature and divide it by the fluctuation range to obtain a standardized aging feature vector. If the data in the early stage of commissioning is missing, use the statistical distribution of the same type of equipment group as the initial benchmark.

[0062] Step S201, where multi-source monitoring data is ready, is read into the feature extraction stage, such as... Figure 4As shown, the hottest spot temperature and temperature rise value are extracted from the temperature dimension to characterize the thermal stress intensity; the discharge quantity statistics and discharge trend are extracted from the partial discharge dimension to characterize the development of insulation defects; the fundamental frequency amplitude and harmonic ratio are extracted from the vibration dimension to characterize the mechanical fastening state; and the reactance deviation is extracted from the electrical dimension to characterize the inter-turn insulation state. These are combined to generate multi-dimensional aging characteristics. Samples with data quality labels are weighted down during extraction.

[0063] To establish an individualized benchmark, the mean and fluctuation range of each of the multidimensional aging characteristics were statistically analyzed during the stable operation period in the initial stage of commissioning. Figure 4 As shown. The normalized aging feature vector is obtained by subtracting the mean of the individualized baseline from each of the multidimensional aging features and then dividing by the fluctuation range. The normalized aging feature vector reflects the degree of change relative to the initial state. When initial operation data is missing, the statistical distribution of a group of similar equipment is used as the initial baseline with an added uncertainty explanation. The normalized aging feature vector is organized by timestamp with the equipment identifier as the primary key, and is read at the modeling entry point in step S401.

[0064] In this embodiment, the multidimensional aging features include at least the hottest temperature extracted from the temperature dimension. With temperature rise It is used to characterize the intensity and distribution of thermal stress; the discharge quantity extracted from the partial discharge dimension. With discharge trend Used to characterize weak discharges and defect development within insulation; fundamental frequency amplitude extracted from vibration dimension. Compared with higher harmonics Used to characterize mechanical fastening conditions and electrodynamic fatigue; the reactance value deviation extracted from the electrical dimension. With leakage current It is used to directly characterize the degree of insulation degradation between winding turns.

[0065] The individualized baseline standardization calculation is performed according to the following formula:

[0066]

[0067] in, These are the standardized aging characteristic values ​​(dimensionless). These are the currently measured multidimensional aging characteristic values. This serves as the baseline for the average value during the stable operation period in the initial stage of commissioning. This represents the standard deviation of the fluctuation range during the stable operation period in the initial stage of commissioning. When data from the initial stage of commissioning is missing, the statistical distribution of a group of similar equipment is used as the initial benchmark.

[0068] The thermal aging sub-model is constructed based on the relationship between Montsinger and Arrhenius, and its equivalent thermal aging rate is calculated as follows:

[0069]

[0070] in, Equivalent thermal aging rate (dimensionless). The hottest temperature (unit: °C) is used as the reference temperature to calculate the equivalent thermal aging rate.

[0071] The electro-aging sub-model is constructed based on the inverse power law model, and its electro-aging lifetime calculation formula is as follows:

[0072]

[0073] in, It is the electrical aging life factor (dimensionless). These are constants related to insulating materials. Equivalent electric field strength (unit: kV / mm). It is the voltage aging index (dimensionless, with a value range of 9~12).

[0074] The mechanical aging sub-model is constructed based on the fatigue cumulative damage model, and its mechanical aging degree is calculated as follows:

[0075]

[0076] in, Mechanical aging degree (dimensionless). For the first The actual number of stress cycles at each stress level The total number of cycles required to reach fatigue failure at the current stress level (determined by the material's SN curve) is calculated based on the statistical accumulation of mechanical fatigue through electrodynamic and temperature cycling. The thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model.

[0077] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 4 It can also specifically include the following:

[0078] Step S401: Construct a thermal aging sub-model based on the thermal aging rate relationship of the insulating material, construct an electrical aging sub-model based on the inverse power law relationship between inter-turn voltage stress and ground voltage stress, and construct a mechanical aging sub-model based on the fatigue accumulation relationship between electrodynamic cycle and temperature cycle. The thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model.

[0079] Step S402: Substitute the temperature history corresponding to the standardized aging feature vector into the thermal aging sub-model, the electrical aging sub-model and the mechanical aging sub-model respectively, and integrate them periodically. Statistically accumulate fatigue damage for stress cycles within the cycle, and sum the increments of each cycle to generate thermal aging degree, electrical aging degree and mechanical aging degree respectively.

[0080] After the standardized aging feature vector in step S301 is ready, three sub-models for thermal aging, electrical aging, and mechanical aging are constructed respectively, such as... Figure 4 As shown, a thermal aging sub-model is constructed based on the thermal aging rate relationship of the insulating material, converting the temperature history into the cumulative aging equivalent at the reference temperature, with the thermal aging activation energy ranging from 0.8 to 1.3 eV. An electrical aging sub-model is constructed based on the inverse power law relationship between inter-turn voltage stress and ground voltage stress, calculating the equivalent electric field stress based on the actual voltage waveform under harmonic and transient overvoltage conditions, with the voltage index ranging from 9 to 12. A mechanical aging sub-model is constructed based on the fatigue accumulation relationship between electrodynamic cycles and temperature cycles. The thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model. The model parameters are selected based on typical values ​​according to the material type or fitted based on aging data of similar equipment. When parameter identification fails, it reverts to the typical values ​​in the material handbook and registers the parameter reversion flag.

[0081] The temperature history and electric field stress history corresponding to the standardized aging feature vector are respectively substituted into the thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model, as follows: Figure 4 As shown. Within each calculation cycle, the stress cycle is statistically accumulated to obtain the thermal aging degree increment, electrical aging degree increment, and mechanical aging degree increment. Based on this, the increments of each cycle are accumulated to generate the thermal aging degree, electrical aging degree, and mechanical aging degree respectively. Each sub-item of aging degree is organized according to the calculation cycle with the equipment identifier as the primary key, for step S501 to read at the comprehensive evaluation entry point. The cycle with integration overflow is registered with a calculation anomaly flag and the value of the previous cycle is used for continuation.

[0082] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 5 It can also specifically include the following:

[0083] Step S501: Based on the preset fusion rule family, the thermal aging degree, the electrical aging degree and the mechanical aging degree are integrated and the dominant component is selected. The deviation of the multidimensional aging characteristics from the individualized benchmark is combined with the acceleration effect of the environmental parameters on the thermal aging rate according to the preset weight to calculate the correction amount and generate the comprehensive aging degree.

[0084] Within each calculation cycle, the aging degree of each component is integrated and accumulated over the cycle. The formula for calculating the total aging degree is:

[0085]

[0086] in, For a moment The cumulative total aging degree (dimensionless). The thermal aging rate is dimensionless. The electrical aging rate is dimensionless. The mechanical aging rate (dimensionless) is calculated by accumulating the increments of stress damage according to a set period.

[0087] Based on the preset fusion rule family, the thermal aging degree is comprehensively considered. The electro-aging degree With the mechanical aging degree The overall aging degree is generated by combining the deviation correction, and its calculation formula is as follows:

[0088]

[0089] in, The overall aging degree (dimensionless). , , These are thermal aging degree, electrical aging degree, and mechanical aging degree (dimensionless). , , These are the combined weighting coefficients for thermal, electrical, and mechanical aging (dimensionless, with the sum of the three being 1). This is a dimensionless correction factor (calculated based on the deviation from a personalized baseline according to multidimensional aging characteristics and environmental temperature and pollution levels). The deviation is based on the partial discharge quantity. Temperature rise The deviation from the individualized baseline is calculated, and the environmental correction is determined based on the accelerating effect of ambient temperature, relative humidity, and pollution level on the thermal aging rate.

[0090] Based on the overall aging degree Aging levels are determined according to preset grading standards:

[0091] when The current condition is Level I normal, and the deviation of each characteristic is within the baseline.

[0092] when At Level II, monitoring is required due to partial discharge or slight temperature deviation.

[0093] when The current state is a Level III anomaly, with significant deviations in characteristics, indicating a defect.

[0094] when The vehicle was classified as Level IV severely aged, triggering an emergency warning and maintenance.

[0095] The monitoring anomaly level is independently assessed based on the deviation of the multidimensional aging characteristics, and the worst level among the aging level and the monitoring anomaly level is used to generate the insulation status level.

[0096] Based on the deviations in the distribution characteristics of each monitoring point, key aging areas can be located:

[0097] By identifying abnormal temperature distribution, the winding encapsulation layer with the largest temperature rise and the most severe heat generation can be located;

[0098] The location of the power source is precisely located using the time difference between ultra-high frequency and ultrasonic waves.

[0099] Vibration spectrum analysis can be used to identify mechanical defects such as loose core clamps or winding deformation.

[0100] Step S502: Divide the aging level according to the comprehensive aging degree and the preset grading standard, independently evaluate the monitoring anomaly level based on the deviation of the multi-dimensional aging characteristics, take the worst level among the aging level and the monitoring anomaly level to generate the insulation status level, and locate the key aging parts according to the characteristic deviation distribution of each monitoring point.

[0101] After each sub-item of aging in step S401 is ready, the overall aging value is generated, such as... Figure 5 As shown. Based on a preset fusion rule family, the thermal aging degree, the electrical aging degree, and the mechanical aging degree are integrated, and the dominant component is taken as the initial value. The deviation of the multidimensional aging characteristics from the individualized benchmark is combined with the deviation of the multidimensional aging characteristics to the individualized benchmark and a correction amount is calculated according to a preset weight. When the partial discharge deviation increases, the corresponding correction amount is increased. When the temperature rise deviation increases, the thermal aging correction amount is increased. When the relative humidity and contamination increase, the corresponding correction amount is increased according to their accelerating effect on the surface aging and thermal aging rate. The correction weights are determined based on historical data learning to generate a comprehensive aging degree.

[0102] The aging level is determined according to the comprehensive aging degree and a preset grading standard, such as... Figure 5 As shown, a comprehensive aging degree of 0.8 is considered severe aging. The monitoring anomaly levels of partial discharge, temperature, and vibration are independently assessed based on the deviation of the multi-dimensional aging characteristics. The worst level among the aging level and the monitoring anomaly level is used to generate the insulation condition level. Based on this, key aging areas are located according to the characteristic deviation distribution of each monitoring point; temperature analysis locates the winding layer with the largest temperature rise; partial discharge is located using ultra-high frequency and ultrasonic time difference analysis; and vibration analysis locates structural anomalies. The insulation condition level uses the equipment identifier as the primary key, along with the dominant aging factor and key aging area fields. When the difference between the aging level and the monitoring anomaly level exceeds a certain limit, a verification flag is triggered for step S601 to read.

[0103] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 6 It can also specifically include the following:

[0104] Step S601: Fit the evolution trend of the historical sequence of the comprehensive aging degree in the preset trend model family according to the goodness of fit, estimate the recent average aging rate, and extrapolate the comprehensive aging degree to the failure threshold condition according to the aging rate to calculate the remaining service life.

[0105] Step S602: By randomly sampling, the parameter uncertainty of the multi-stress integrated insulation aging model is propagated to the remaining service life and its confidence interval is quantified. Maintenance decisions and graded early warnings are generated based on a preset decision rule family that matches the insulation state level with the remaining service life.

[0106] After the insulation condition class and overall aging degree are ready in step S501, the remaining service life is predicted, such as... Figure 6 As shown, the historical sequence of the comprehensive aging degree is fitted to the evolution trend in a preset family of trend models based on goodness of fit. The family of trend models includes linear, exponential, and polynomial models. The recent average aging rate is estimated, and it is determined whether the aging rate has an accelerating trend. Based on this, the comprehensive aging degree is extrapolated to the failure threshold condition to calculate the baseline remaining service life. When there are future load plans and seasonal operating conditions, the aging rate is adjusted according to the changes in operating conditions, and the adjusted remaining service life is given.

[0107] The parameter uncertainties of the multi-stress integrated insulation aging model are propagated to the remaining service life through random sampling, such as... Figure 6 As shown, the confidence intervals for aging rate fluctuations and uncertainties in future operating conditions are superimposed. The remaining service life is expressed as a pair of predicted values ​​and confidence intervals, generating a curve corresponding to time and failure probability. Based on this, maintenance decisions and graded warnings are generated according to a preset family of decision rules matching the insulation status level and the remaining service life. A red warning is triggered when the status level is severe or the remaining service life is below a preset threshold. The maintenance decisions and graded warnings are output to the operation and maintenance management system with the equipment identifier as the primary key and the decision record is registered.

[0108] Regarding the overall aging degree An exponential fit is performed on the historical sequence, and its evolutionary trend model is as follows:

[0109]

[0110] in, For a moment The overall aging degree (dimensionless). The initial overall aging degree (dimensionless). This is the aging rate coefficient (unit: 1 / year). Time (unit: years). The remaining service life is calculated by extrapolating the overall aging degree to the preset failure threshold condition.

[0111] By randomly sampling, the parameter uncertainties of the multi-stress integrated insulation aging model are propagated to the remaining service life, and the aging rate fluctuations and future operating condition uncertainties are propagated to the remaining service life, outputting the remaining service life prediction value and failure probability with confidence intervals.

[0112] Maintenance decisions and tiered early warnings are generated based on a preset set of decision rules that match the insulation condition level with the remaining service life.

[0113] Red alert: Triggered when the insulation condition is severely aged or the remaining service life is less than 2 years, the decision is to immediately arrange power outage for maintenance and insulation replacement;

[0114] Orange alert: Triggered when the insulation status is under watch or the remaining service life is between 2 and 5 years. The decision is to shorten the online monitoring cycle and limit overload operation.

[0115] Based on the dominant aging stress, specific maintenance recommendations are provided:

[0116] When thermal aging is the primary process, check the cooling system and heat dissipation surfaces;

[0117] When electrical aging is the primary process, check the insulation dielectric properties and overvoltage protection.

[0118] When mechanical aging is the primary factor, check for deformation of fasteners and windings.

[0119] Dielectric performance parameters, including the dielectric loss tangent, are obtained during the power-off test of the reactor under test. Insulation resistance and polarization index; insulation physicochemical test results were obtained during the decommissioning and dismantling of similar reactors, including measured values ​​of the degree of polymerization (DP) of the insulating paper and microscopic morphology analysis of the insulating material. Power outage test data and decommissioned equipment dissection data were compared with the stated insulation condition level to identify systematic estimation biases and generate model calibration values.

[0120] The parameters of the multi-stress integrated insulation aging model are updated using Bayesian methods based on the model calibration values: the measured conditions are compared with the evaluation level to identify system biases, and the aging model parameters are adjusted according to the calibration results. For reactors that have undergone major repairs, the individualized benchmark is reset; for reactors that have not undergone major repairs for a long time, the individualized benchmark is dynamically re-evaluated using a sliding benchmark. The updated parameters and individualized benchmarks are fed back into the online evaluation model, forming a closed-loop iteration of monitoring, evaluation, verification, and updating.

[0121] In one embodiment of the online assessment and lifetime prediction method for the insulation aging state of high-voltage series reactors in this application, see [reference needed]. Figure 6 It can also specifically include the following:

[0122] Step S701: Obtain dielectric performance parameters during the power outage test of the tested reactor, and obtain insulation physicochemical test results when the same type of reactor is decommissioned and dismantled. Compare the power outage test data and the dismantled data of the decommissioned equipment with the insulation status level to identify system deviations and generate model calibration quantities.

[0123] Step S702: Perform Bayesian update on the parameters of the multi-stress integrated insulation aging model according to the model calibration amount, trigger the individualized benchmark reset for reactors that have undergone major repairs, and dynamically re-evaluate the individualized benchmark using a sliding benchmark for reactors that have not undergone major repairs for a long time.

[0124] After the maintenance decision and early warning output in step S601, perform the power outage test and decommissioning data calibration, such as... Figure 6 As shown. During the power outage test of the tested reactor, dielectric performance parameters such as insulation resistance, polarization index, and dielectric loss angle are obtained. When a similar reactor is decommissioned and dismantled, physicochemical test results such as insulation polymerization degree, moisture content, and actual aging morphology are obtained. The power outage test data and the dismantled data of the decommissioned equipment are compared with the insulation condition level to identify system deviations and generate model calibration values. The comparison results are recorded in the calibration record. When the deviation exceeds the limit, the calibration trigger flag is triggered.

[0125] Based on the model calibration values, Bayesian updates are performed on the parameters of the multi-stress integrated insulation aging model, such as... Figure 6 As shown. The aging model parameters are updated based on actual aging data, the correlation weights between monitoring features and aging degree are adjusted based on calibration results, and the early warning threshold conditions are statistically optimized based on early warning effects. Accordingly, the individualized benchmark is dynamically re-evaluated. Reactors that have undergone major repairs trigger the individualized benchmark reset, while reactors that have not undergone major repairs for a long time use a sliding benchmark. The updated parameters and individualized benchmarks are fed back to the individualized benchmark in step S101 and the multi-stress integrated insulation aging model in step S401. Equipment of the same model, manufacturer, and operating environment shares calibration knowledge, forming a closed loop of monitoring, evaluation, verification, and updating.

[0126] To implement all or part of the aforementioned method, this application provides an embodiment of an apparatus for implementing the online assessment of the insulation aging state and life prediction method for high-voltage series reactors. See [link to relevant documentation]. Figure 2 The device specifically includes the following components:

[0127] The multi-source sensing module is used to deploy various types of sensors on the insulation part of the high-voltage series reactor and connect electrical and environmental parameters. According to the preset sampling rule family, it collects temperature field, partial discharge and vibration parameters online, and then performs outlier removal, missing value interpolation and time alignment on the raw data to generate multi-source monitoring data.

[0128] The feature extraction module is used to extract multi-dimensional aging features from the multi-source monitoring data according to the dimensions of temperature, partial discharge and vibration, establish an individualized benchmark based on the stable operation period in the early stage of commissioning, and standardize the multi-dimensional aging features relative to the individualized benchmark to generate a standardized aging feature vector.

[0129] The aging modeling module is used to construct thermal aging sub-models, electrical aging sub-models and mechanical aging sub-models based on the thermal aging rate relationship, the electric field stress inverse power law relationship and the mechanical fatigue accumulation relationship, and to jointly form a multi-stress comprehensive insulation aging model. The stress history corresponding to the standardized aging feature vector is substituted into each of the sub-models and the thermal aging degree, electrical aging degree and mechanical aging degree are generated by periodic integration.

[0130] The condition assessment module is used to generate a comprehensive aging degree by combining the thermal aging degree, the electrical aging degree and the mechanical aging degree according to a preset fusion rule family and combining the deviation correction. After generating the insulation condition level by comparing the comprehensive aging degree with the preset grading standard and taking the worst level among the monitored abnormal levels, the module locates the key aging parts.

[0131] The life prediction module is used to fit the evolution trend of the historical sequence of the comprehensive aging degree and estimate the current aging rate. It extrapolates the comprehensive aging degree to the failure threshold condition to calculate the remaining service life and quantifies the confidence interval. Then, it generates maintenance decisions and graded early warnings based on the insulation status level and the remaining service life. Finally, it compares the power outage test data and decommissioned equipment data with the insulation status level to identify the deviation, updates the multi-stress comprehensive insulation aging model, and then feeds it back to the individualized benchmark.

[0132] The multi-source sensing module deploys fiber optic temperature sensors, ultra-high frequency and ultrasonic partial discharge sensors, and piezoelectric vibration sensors on the insulation parts of the high-voltage series reactor. Electrical parameters and environmental parameters are connected via current transformers and voltage transformers. Temperature field, partial discharge, and vibration parameters are collected online according to a preset sampling rule family. Outlier removal, missing value imputation, and time alignment are performed on the raw data to generate multi-source monitoring data. The feature extraction module extracts multi-dimensional aging features from the multi-source monitoring data according to temperature, partial discharge, and vibration dimensions. An individualized benchmark is established based on the stable operation period during the initial commissioning phase, and the multi-dimensional aging features are standardized relative to the individualized benchmark to generate a standardized aging feature vector.

[0133] The aging modeling module constructs sub-models for thermal aging, electrical aging, and mechanical aging based on the thermal aging rate relationship of the insulation material, the inverse power law relationship of inter-turn and ground voltage stress, and the cumulative relationship of mechanical fatigue due to electrodynamic and temperature cycles. These sub-models collectively form a multi-stress integrated insulation aging model. The stress history corresponding to the standardized aging feature vector is substituted into each sub-model and accumulated by periodic integration to generate the thermal aging degree, electrical aging degree, and mechanical aging degree. The condition assessment module integrates the aging degrees of each component according to a preset fusion rule family and combines deviation correction to generate a comprehensive aging degree. Based on the comprehensive aging degree, it compares it with a preset grading standard and takes the worst grade among the monitored anomaly grades to generate the insulation condition grade and locate the key aging parts.

[0134] The lifespan prediction module fits the historical sequence of the comprehensive aging degree to predict its evolution trend and estimates the current aging rate. It extrapolates the comprehensive aging degree to the failure threshold to calculate the remaining service life. It quantifies the confidence interval through random sampling and generates a curve corresponding to time and failure probability. Based on the insulation condition level and the remaining service life, it generates maintenance decisions and graded early warnings using a preset set of decision rules. The device acquires power outage test data and decommissioned equipment disassembly data during power outage tests of the tested reactor and the decommissioning disassembly of similar reactors. It compares this data with the insulation condition level to identify system deviations and generate model calibration values. Based on these calibration values, it updates the multi-stress comprehensive insulation aging model and dynamically re-evaluates and returns the data to the individualized benchmark.

[0135] As can be seen from the above description, the online assessment and life prediction device for the insulation aging status of high-voltage series reactors provided in this application embodiment can perform online quantitative assessment and prediction of the remaining life of the insulation aging status of high-voltage series reactors through the coordinated operation of five modules: multi-source sensing, feature extraction, aging modeling, status assessment and life prediction.

[0136] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method.

[0137] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0138] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method.

[0139] In this embodiment, multi-source monitoring data is generated by deploying multiple types of sensors on the insulation parts and preprocessing the raw data. Multi-dimensional aging characteristics are standardized relative to individualized benchmarks to adapt to individual differences in equipment. The comprehensive aging degree is quantified by combining three types of stress sub-models of thermal aging, electrical aging and mechanical aging and combining deviation correction. The insulation status level is generated by comparing the grading standard and the monitoring anomaly level and taking the worst level. The remaining service life with confidence interval is given by extrapolating the aging degree trend and random sampling, and maintenance decisions and graded early warnings are generated. The multi-stress comprehensive insulation aging model is updated by identifying system deviations based on power outage test data and decommissioned equipment dissection data, and dynamically re-evaluated and fed back to the individualized benchmark to form a closed loop.

[0140] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for online assessment of insulation aging status and life prediction of high-voltage series reactors, characterized in that, The method includes: Multiple types of sensors are deployed at the insulation part of the high-voltage series reactor and connected to electrical and environmental parameters. Temperature field, partial discharge and vibration parameters are collected according to a preset sampling rule family and then preprocessed to generate multi-source monitoring data. Multi-dimensional aging features are extracted from the multi-source monitoring data and standardized relative to individual benchmarks to generate standardized aging feature vectors. The standardized aging feature vector is read, and a multi-stress comprehensive insulation aging model is constructed based on the relationship between thermal aging, electrical aging and mechanical fatigue. The stress history is substituted into the model and integrated periodically to generate the sub-item aging degree. The sub-item aging degree is integrated according to the preset fusion rule family and the deviation of the multi-dimensional aging feature is combined to generate the comprehensive aging degree. The insulation state level is generated by taking the worst level among the monitoring anomaly levels according to the comprehensive aging degree and the grading standard. The evolution trend of the comprehensive aging degree is fitted and extrapolated to the failure threshold condition to calculate the remaining service life and quantify the confidence interval. Maintenance decisions and early warnings are generated based on the decision rule family matching the insulation status level and the remaining service life. Power outage test data and decommissioned equipment data are compared with the insulation status level to identify deviations and generate model calibration values. The multi-stress comprehensive insulation aging model is updated based on the model calibration values ​​and fed back to the individualized benchmark.

2. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 1, characterized in that, The process involves deploying multiple types of sensors at the insulation section of the high-voltage series reactor and connecting them to electrical and environmental parameters. After collecting temperature field, partial discharge, and vibration parameters according to a preset sampling rule family, the data is preprocessed to generate multi-source monitoring data, including: Based on the insulation structure of the reactor under test, temperature sensors are arranged in the winding encapsulation layer, vibration sensors are arranged in the core clamps and the end of the encapsulation layer, and partial discharge sensors are arranged in the body and bushing flange. Electrical parameters are connected through current transformers and voltage transformers, and environmental parameters including ambient temperature, relative humidity, pollution level and ultraviolet radiation intensity are also connected. Based on the dynamic characteristics of each parameter, a differentiated sampling period is set to collect and generate raw data online. Outliers are removed from the original data based on the physical reasonable range and statistical distribution. Missing segments are interpolated based on historical patterns. Noise reduction filtering is performed on partial discharge and vibration signals. Data from different sampling periods are aligned to a unified time reference to generate multi-source monitoring data.

3. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 1, characterized in that, The step of extracting multidimensional aging features from the multi-source monitoring data and standardizing them relative to an individualized benchmark to generate a standardized aging feature vector includes: The hottest spot temperature and temperature rise value are extracted from the temperature dimension of the multi-source monitoring data; the discharge quantity statistics and discharge trend are extracted from the partial discharge dimension; the fundamental frequency amplitude and harmonic ratio are extracted from the vibration dimension; and the reactance value deviation is extracted from the electrical dimension to generate multi-dimensional aging characteristics. To establish an individualized benchmark, the mean and fluctuation range of each multidimensional aging feature are statistically analyzed during the stable operation period in the early stage of commissioning. The mean of the individualized benchmark is subtracted from each multidimensional aging feature and then divided by the fluctuation range to obtain a standardized aging feature vector. When the data from the early stage of commissioning is missing, the statistical distribution of a group of similar equipment is used as the initial benchmark.

4. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 1, characterized in that, The process involves reading the standardized aging feature vector, constructing a multi-stress comprehensive insulation aging model based on the relationship between thermal aging, electrical aging, and mechanical fatigue, and then substituting the stress history into the model to generate sub-items of aging degree by periodic integration, including: A thermal aging sub-model is constructed based on the thermal aging rate relationship of the insulating material; an electrical aging sub-model is constructed based on the inverse power law relationship between inter-turn voltage stress and ground voltage stress; and a mechanical aging sub-model is constructed based on the fatigue accumulation relationship between electrodynamic cycle and temperature cycle. The thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model together constitute a multi-stress comprehensive insulation aging model. The temperature history corresponding to the standardized aging feature vector is substituted into the thermal aging sub-model, the electrical aging sub-model, and the mechanical aging sub-model, respectively, and integrated periodically. The stress cycle within the cycle is statistically accumulated to accumulate fatigue damage, and the increment of each cycle is added to generate the thermal aging degree, electrical aging degree, and mechanical aging degree, respectively.

5. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 4, characterized in that, The process involves combining the sub-items of aging degree with the deviation correction of the multi-dimensional aging characteristics according to a preset fusion rule family to generate a comprehensive aging degree. Based on the comprehensive aging degree and the grading standard, the worst grade among the monitored anomaly grades is selected to generate the insulation condition grade, including: Based on a preset set of fusion rules, the thermal aging degree, the electrical aging degree, and the mechanical aging degree are combined and the dominant component is selected. The deviation of the multidimensional aging characteristics from the individualized benchmark is combined with the acceleration effect of the environmental parameters on the thermal aging rate according to a preset weight to generate a comprehensive aging degree. The aging level is divided according to the comprehensive aging degree and the preset grading standard. The monitoring anomaly level is independently evaluated based on the deviation of the multi-dimensional aging characteristics. The worst level among the aging level and the monitoring anomaly level is taken to generate the insulation status level. The key aging parts are located according to the characteristic deviation distribution of each monitoring point.

6. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 1, characterized in that, The process involves fitting the evolution trend of the comprehensive aging degree and extrapolating it to the failure threshold condition to calculate the remaining service life and quantify the confidence interval. Maintenance decisions and early warnings are generated based on the decision rule family matching the insulation condition level and the remaining service life, including: The historical sequence of the comprehensive aging degree is fitted to the evolution trend in a preset trend model family according to the goodness of fit, the recent average aging rate is estimated, and the comprehensive aging degree is extrapolated to the failure threshold condition according to the aging rate to calculate the remaining service life. By randomly sampling, the parameter uncertainty of the multi-stress integrated insulation aging model is propagated to the remaining service life to quantify its confidence interval. Maintenance decisions and graded early warnings are generated based on a preset decision rule family that matches the insulation state level with the remaining service life.

7. The method for online assessment of insulation aging status and life prediction of high-voltage series reactors according to claim 1, characterized in that, The process of comparing the acquired power outage test data and decommissioned equipment data with the insulation condition level to identify deviations and generate model calibration values, updating the multi-stress integrated insulation aging model based on the model calibration values ​​and feeding it back to the individualized benchmark, includes: Dielectric performance parameters are obtained during the power outage test of the tested reactor, and insulation physicochemical test results are obtained when the same type of reactor is decommissioned and disassembled. The power outage test data and the disassembly data of the decommissioned equipment are compared with the insulation status level to identify system deviations and generate model calibration quantities. Based on the model calibration, Bayesian updates are performed on the parameters of the multi-stress integrated insulation aging model. For reactors that have undergone major repairs, the individualized benchmark is reset. For reactors that have not undergone major repairs for a long time, the individualized benchmark is dynamically re-evaluated using a sliding benchmark.

8. A device for online assessment of insulation aging status and life prediction of high-voltage series reactors, characterized in that, The device includes: The multi-source sensing module is used to deploy various types of sensors on the insulation part of the high-voltage series reactor and connect electrical and environmental parameters. According to the preset sampling rule family, it collects temperature field, partial discharge and vibration parameters online, and then performs outlier removal, missing value interpolation and time alignment on the raw data to generate multi-source monitoring data. The feature extraction module is used to extract multi-dimensional aging features from the multi-source monitoring data according to the dimensions of temperature, partial discharge and vibration, establish an individualized benchmark based on the stable operation period in the early stage of commissioning, and standardize the multi-dimensional aging features relative to the individualized benchmark to generate a standardized aging feature vector. The aging modeling module is used to construct thermal aging sub-models, electrical aging sub-models and mechanical aging sub-models based on the thermal aging rate relationship, the electric field stress inverse power law relationship and the mechanical fatigue accumulation relationship, and to jointly form a multi-stress comprehensive insulation aging model. The stress history corresponding to the standardized aging feature vector is substituted into each of the sub-models and the thermal aging degree, electrical aging degree and mechanical aging degree are generated by periodic integration. The condition assessment module is used to generate a comprehensive aging degree by combining the thermal aging degree, the electrical aging degree and the mechanical aging degree according to a preset fusion rule family and combining the deviation correction. After generating the insulation condition level by comparing the comprehensive aging degree with the preset grading standard and taking the worst level among the monitored abnormal levels, the module locates the key aging parts. The life prediction module is used to fit the evolution trend of the historical sequence of the comprehensive aging degree and estimate the current aging rate. It extrapolates the comprehensive aging degree to the failure threshold condition to calculate the remaining service life and quantifies the confidence interval. Then, it generates maintenance decisions and graded early warnings based on the insulation status level and the remaining service life. Finally, it compares the power outage test data and decommissioned equipment data with the insulation status level to identify the deviation, updates the multi-stress comprehensive insulation aging model, and then feeds it back to the individualized benchmark.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the online assessment and life prediction method for the insulation aging status of high-voltage series reactors according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the online assessment and life prediction method for the insulation aging status of high-voltage series reactors as described in any one of claims 1 to 7.

Images