A real-time monitoring method and system for insulation performance of a disconnector secondary circuit

Abstract

The application relates to the technical field of insulation performance monitoring, in particular to a real-time monitoring method and system for insulation performance of a secondary circuit of an isolating switch, which comprises the following steps: collecting real-time temperature, real-time humidity in an isolating switch mechanism box and a measured current of the secondary circuit; inputting the real-time temperature and the real-time humidity into a pre-constructed reference model to output a theoretical current; determining a confidence weight at a current time according to a change trend of the real-time humidity; calculating a fluctuation coefficient of a prediction residual within a preset time window; iteratively updating a pollution drift based on the confidence weight and the fluctuation coefficient; subtracting the theoretical current and the pollution drift from the measured current to obtain a fault judgment value, and judging an insulation fault according to the fault judgment value. Through the technical scheme, the false positive rate caused by environmental factors can be reduced, and accurate and self-adaptive real-time monitoring of an insulation state is realized.

Description

Technical Field

[0001] This application relates to the field of insulation performance monitoring technology, and in particular to a method and system for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch. Background Technology

[0002] With the rapid development of smart grid technology, the operational reliability of power equipment has become crucial for ensuring grid security. Disconnect switches, as important equipment widely used in substations, bear core functions such as control and signal transmission in their secondary circuits. Because the disconnect switch mechanism box and secondary circuits are exposed to complex electromagnetic and meteorological environments outdoors for extended periods, their insulation performance is highly susceptible to degradation due to environmental factors such as temperature and humidity. Therefore, achieving accurate monitoring of the insulation performance of the disconnect switch secondary circuits in variable outdoor environments is of great significance for preventing short-circuit faults and ensuring power supply safety.

[0003] Currently, insulation performance is typically monitored using a leakage current threshold-based method. This involves installing a leakage current sensor in the secondary circuit to collect the total leakage current to ground in real time, and then directly comparing the measured current value with a preset alarm threshold. Once the measured current exceeds the set threshold, the insulation performance is determined to be abnormal and an alarm signal is issued, thereby enabling the screening of insulation faults.

[0004] However, the measured leakage current is actually a combination of factors, including insulation leakage in the equipment itself and ambient temperature and humidity. This means the measured leakage current contains normal interference from environmental factors such as temperature and humidity, especially common in rainy, humid weather or when there are significant temperature differences between day and night. Current technology cannot distinguish between environmental interference and actual faults in the measured leakage current, leading to frequent false alarms and a high false alarm rate and poor environmental adaptability in insulation performance monitoring. Summary of the Invention

[0005] To address the technical problems of high false alarm rate and poor environmental adaptability in insulation performance monitoring, this application provides a method and system for real-time monitoring of insulation performance of secondary circuits of disconnecting switches, which can reduce the false alarm rate caused by environmental factors and achieve accurate and adaptive real-time monitoring of insulation status.

[0006] In a first aspect, this application provides a method for real-time monitoring of the insulation performance of a secondary circuit of a disconnector switch. The monitoring method includes: collecting real-time temperature, real-time humidity, and measured current of the secondary circuit within the disconnector switch housing; inputting the real-time temperature and real-time humidity into a pre-constructed benchmark model and outputting the theoretical current under the current environment, wherein the benchmark model characterizes the mapping relationship between temperature, humidity, and leakage current of the equipment in a clean state; determining the confidence weight of the current moment based on the changing trend of the real-time humidity, wherein the confidence weight of the humidity rising phase is less than the confidence weight of the humidity falling phase; calculating the difference between the measured current and the theoretical current to obtain a prediction residual, and calculating the fluctuation coefficient of the prediction residual within a preset time window; weighting the prediction residual based on the confidence weight and the fluctuation coefficient, and iteratively updating the contamination drift amount, wherein the contamination drift amount is used to characterize the slow current increment caused by surface contamination; calculating the difference between the measured current and the sum of the theoretical current and the contamination drift amount to obtain a fault judgment value, and determining that an insulation fault exists when the fault judgment value is greater than an alarm threshold.

[0007] By collecting real-time environmental parameters and current data, the theoretical current is calculated using a benchmark model. The amount of pollution drift is then iteratively calculated by combining a confidence weight reflecting the trend of humidity change and a fluctuation coefficient reflecting the degree of signal oscillation. Finally, the fault is determined based on the difference between the measured current and the theoretical current and the amount of pollution drift. This achieves a refined decomposition of leakage current components, which can not only dynamically follow environmental changes to eliminate background interference, but also automatically track the surface contamination trend to eliminate long-term effects. This significantly improves the robustness and accuracy of the insulation fault monitoring algorithm under complex weather conditions.

[0008] Preferably, after collecting the real-time temperature, real-time humidity, and measured current of the secondary circuit inside the disconnector switch box, the monitoring method further includes: performing a moving average filter on the real-time temperature, real-time humidity, and measured current to eliminate random noise; and performing a time-series alignment operation on the real-time temperature, real-time humidity, and measured current after standardization to form time-series data with a unified timestamp.

[0009] Preferably, the benchmark model is a support vector regression model, which uses normalized real-time temperature and humidity as input features and outputs the leakage current corresponding to the input features.

[0010] Preferably, the reference model uses a radial basis function as the kernel function to map the nonlinear relationship between real-time temperature, real-time humidity and leakage current.

[0011] Preferably, the process of constructing the benchmark model includes: selecting historical data of the device in a healthy state as a training set; using the training set to train the benchmark model until the insensitive loss function between the output of the benchmark model and the leakage current in a healthy state is less than a preset difference, thereby obtaining the benchmark model.

[0012] By selecting historical data of equipment in a healthy state as the training set and optimizing model parameters using an insensitive loss function, it can be ensured that the baseline model only learns the physical laws of the equipment during normal operation, providing a reliable data foundation for subsequent performance monitoring through residual analysis.

[0013] Preferably, determining the confidence weight at the current moment based on the real-time humidity change trend includes: taking the difference between the real-time humidity and the humidity at the previous moment as the real-time humidity change trend; the confidence weight is negatively correlated with the product of a preset sensitivity coefficient and the change trend.

[0014] The model automatically reduces the reference weight during the moisture absorption phase when humidity increases, and increases the weight during the dehumidification phase when humidity decreases or stabilizes. This effectively avoids prediction bias caused by the moisture absorption hysteresis effect of insulation materials and prevents misjudgments caused by the moisture absorption hysteresis effect.

[0015] Preferably, the sensitivity coefficient is positively correlated with the moisture hysteresis value of the insulating material.

[0016] The sensitivity coefficient is related to the moisture absorption hysteresis value of the insulating material. It can adaptively adjust the rate of change of confidence weight according to the moisture absorption and desorption characteristics of different insulating materials. It can be adapted to disconnecting switch insulators of different materials, thus improving the monitoring accuracy of performance monitoring.

[0017] Preferably, calculating the fluctuation coefficient of the prediction residual within a preset time window includes: calculating the variance of the prediction residual within the preset time window; and calculating the product of a preset gating sensitivity and the variance, wherein the fluctuation coefficient is negatively correlated with the product.

[0018] A gating mechanism for updating the amount of dirt drift was constructed using the fluctuation coefficient. This mechanism can automatically suppress the updating of the amount of drift when the prediction residual experiences high-frequency and violent oscillations, such as electromagnetic interference. When the signal is stable, it can capture the slowly changing trend of dirt accumulation and effectively filter out noise interference.

[0019] Preferably, the iterative update of the contamination drift includes: calculating a comprehensive update gain, wherein the comprehensive update gain is the product of the confidence weight and the volatility coefficient; calculating the difference between the prediction residual and the contamination drift at the previous time step; and adding the contamination drift at the previous time step to the product of the comprehensive update gain and the difference to obtain the contamination drift at the current time step.

[0020] By using the product of confidence weight and fluctuation coefficient as the comprehensive update gain to iteratively update the amount of contamination drift, dynamic tracking of the slow current increment caused by surface contamination is achieved, ensuring the robustness of the contamination drift update, and thus enabling accurate separation of environmental and contamination components from leakage current.

[0021] In a second aspect, this application also provides a real-time monitoring system for the insulation performance of the secondary circuit of a disconnector switch, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the real-time monitoring method for the insulation performance of the secondary circuit of a disconnector switch according to the first aspect of this application.

[0022] The technical solution of this application has the following beneficial technical effects: First, a baseline model is constructed to predict the theoretical current. Then, combined with the dynamic update of the pollution drift, the measured current is decomposed into environmental components, pollution accumulation components, and fault components. Specifically, the baseline model is used to remove the environmental component caused by temperature and humidity from the measured current. The pollution drift, which is iteratively updated based on confidence weights and fluctuation coefficients, is used to remove the pollution accumulation component caused by surface contamination from the measured current. Finally, only the fault component caused by insulation defects is retained as the criterion. This effectively solves the problem that traditional monitoring methods cannot distinguish between normal environmental leakage and insulation fault leakage in rainy, humid, or drastically temperature-sensitive environments. It significantly reduces the false alarm rate caused by environmental factors and achieves accurate, adaptive, and real-time monitoring of the insulation status of the secondary circuit of the disconnector switch. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for real-time monitoring of the insulation performance of a secondary circuit of a disconnector switch according to an embodiment of this application.

[0024] Figure 2 This is a schematic diagram of the measured current, theoretical current, and pollution drift at each time sampling point according to the embodiments of this application.

[0025] Figure 3 This is a schematic diagram of the fault determination values ​​at each time sampling point according to the embodiments of this application.

[0026] Figure 4 This is a structural block diagram of a real-time monitoring system for the insulation performance of a secondary circuit of a disconnector switch according to an embodiment of this application. Detailed Implementation

[0027] According to the first aspect of this application, this application provides a method for real-time monitoring of the insulation performance of the secondary circuit of a disconnecting switch, applicable to outdoor high-voltage substations in rainy and humid areas. In outdoor high-voltage substations, the disconnecting switch mechanism box and secondary circuit are exposed to environments with large diurnal temperature differences and drastic humidity changes for extended periods. Traditional monitoring methods cannot distinguish between normal leakage caused by environmental humidity and fault leakage caused by equipment insulation aging, and are prone to issuing false alarms due to environmental influences. This application can eliminate environmental interference under complex meteorological conditions by iteratively updating the pollution drift quantity, achieving accurate determination of insulation faults.

[0028] Figure 1 This is a flowchart of a method for real-time monitoring of the insulation performance of a secondary circuit of a disconnector switch according to an embodiment of this application. Figure 1 As shown, the real-time monitoring method for the insulation performance of the secondary circuit of a disconnector switch includes steps S101 to S106, which are described in detail below.

[0029] S101: Collects real-time temperature, real-time humidity, and measured current of the secondary circuit inside the disconnector switch housing.

[0030] In one embodiment, the real-time temperature refers to the air temperature inside the disconnector switch housing, the real-time humidity refers to the relative humidity inside the housing, and the measured current refers to the leakage current to ground of the secondary circuit collected by a current transformer. The above data are synchronously acquired at a fixed sampling frequency using temperature and humidity sensors and leakage current sensors deployed inside the housing. For example, the sampling frequency can be set to 1Hz, meaning data is collected once per second.

[0031] To ensure data quality and eliminate the impact of environmental noise on monitoring accuracy, preprocessing of the raw data is necessary. After collecting real-time temperature, real-time humidity, and measured current of the secondary circuit inside the disconnector switch box, the monitoring method also includes: performing moving average filtering on the real-time temperature, real-time humidity, and measured current to eliminate random noise; and performing time-series alignment operation on the real-time temperature, real-time humidity, and measured current after standardization to form time-series data with unified timestamps.

[0032] Understandably, the raw signals acquired by sensors will contain high-frequency interference noise. Moving average filtering is a time-domain filtering method that can effectively suppress random noise. By setting a sliding window of a preset length, such as a window length of 5, the arithmetic mean of the data within the window is calculated as the filtering result at the current moment, thereby eliminating interference noise in the raw signal.

[0033] Thus, by synchronously acquiring and preprocessing real-time temperature, real-time humidity, and measured current of the secondary circuit, a data foundation is provided for subsequent real-time monitoring of insulation performance.

[0034] S102, input the real-time temperature and the real-time humidity into a pre-built benchmark model, and output the theoretical current under the current environment. The benchmark model characterizes the mapping relationship between temperature, humidity and leakage current of the device in a clean state.

[0035] In one embodiment, the baseline model is a pre-trained mathematical model that has learned how the leakage current varies with temperature and humidity when the device is well-insulated and has a clean surface. The theoretical current is the ideal leakage current value predicted by inputting the normalized real-time temperature and humidity into the baseline model.

[0036] The baseline model is a support vector regression model, also known as the SVR model; the kernel function of the support vector regression model is a radial basis function, which is used to map the nonlinear relationship between real-time temperature, real-time humidity and leakage current.

[0037] Understandably, in outdoor high-voltage environments, leakage current exhibits a significant nonlinear relationship with temperature and humidity. Support vector regression models utilize kernel functions to map low-dimensional temperature and humidity data to a high-dimensional feature space, thereby achieving linear regression. Radial basis function kernel functions have a strong ability to handle nonlinear mappings and can accurately characterize the complex surface between environmental parameters and insulation performance.

[0038] In one embodiment, the loss function of the support vector regression model is an insensitive loss function. The process of constructing the benchmark model includes: selecting historical data of the device in a healthy state as a training set; training the benchmark model using the training set until the insensitive loss function between the output of the benchmark model and the leakage current in the healthy state is less than a preset difference, thereby obtaining the benchmark model.

[0039] A single historical data point includes historical temperature, historical humidity, and the corresponding historical measured current. During the construction of the training set, the historical database needs to be cleaned, removing data from time periods with clear fault records, retaining only data from periods when the equipment is operating well and its insulation performance is normal. Simultaneously, to ensure the generalization ability of the baseline model, the training set should cover typical meteorological conditions throughout the four seasons, including the complete temperature and humidity range from dry to humid (such as the plum rain season).

[0040] Historical temperature and humidity data are selected as input features, and corresponding historical measured current is used as the target label. These are input into a support vector regression model for training. The core objective during training is to find a regression hyperplane such that most training samples fall within a tolerance band around this hyperplane. The insensitive loss function satisfies the following relationship: ; In the formula, for and Insensitive loss values ​​between; Training samples in the training set Corresponding historical measured current; The baseline model is based on the training samples The output result; This is the insensitivity coefficient, i.e., the preset tolerance band width. Insensitivity coefficient It can be set to 0.01.

[0041] Thus, the baseline model built using the support vector regression model can accurately predict the current theoretical current from complex environmental changes, providing a model basis for subsequent removal of environmental impacts.

[0042] S103, determine the confidence weight of the current moment based on the real-time humidity change trend, wherein the confidence weight of the humidity rising phase is less than the confidence weight of the humidity falling phase.

[0043] In one embodiment, the confidence weight is a value between 0 and 1, used to quantify the degree of confidence in the output of the benchmark model at the current moment. Because insulating materials exhibit a hygroscopic hysteresis effect—that is, when humidity rises rapidly, the moisture balance inside the insulating material lags behind the air humidity—the measured current change lags behind the theoretical prediction. In this case, the reference value of the benchmark model decreases, and a smaller confidence weight should be assigned. Furthermore, at the same relative humidity, the equilibrium moisture content of the desorption process (corresponding to the dehumidification process during the humidity decrease phase) is usually higher than that of the hygroscopic process. That is, when humidity decreases or stabilizes, the moisture balance inside the insulating material tends to synchronize with the air humidity, and a larger confidence weight should be assigned in this case.

[0044] To address the asymmetry in the moisture absorption and desiccation processes, a formula for calculating the confidence weights is constructed. Satisfying the relation: ; In the formula, The confidence weight at the current moment. This is the sensitivity coefficient. This represents the real-time humidity trend, specifically the difference between the current humidity and the humidity at the previous moment.

[0045] Understandably, when humidity is in the rising phase, it corresponds to the moisture absorption process of the insulating material. If the value is positive, then in the denominator Significantly increased, leading to The level rapidly decreases and approaches zero. In subsequent updates to the contamination drift, the residual of the current baseline model is automatically ignored to prevent misjudging normal physical hysteresis as anomaly. When humidity is in a decreasing or stable phase, corresponding to the dehumidification or stabilization process of the insulation material,... When the value is negative or zero, the denominator decreases, making Maintaining a high level close to 1, the residuals of the current benchmark model are fully utilized for state calibration to update the amount of contamination drift.

[0046] In another embodiment, considering that disconnecting switch insulators made of different materials have different microporous structures, their adsorption and desorption characteristics for moisture vary significantly. To further improve the accuracy of the confidence weight, the sensitivity coefficient... It is determined by the moisture hysteresis value of the insulating material.

[0047] Specifically, sensitivity coefficient Hygroscopic hysteresis value of insulating materials Positive correlation; sensitivity coefficient Satisfying the relation: ; In the formula, Based on sensitivity, For material impact factors, This represents the moisture hysteresis value of the insulating material; the material influence factor can be set to 1, and the basic sensitivity can be set to 0.5. The moisture hysteresis value... It is the maximum normalized value of the difference between the moisture content of the material when it desorbs from a saturated state to equilibrium and the moisture content when it absorbs moisture from a dry state to equilibrium at the standard test temperature.

[0048] Understandably, when the insulating material has a large hygroscopic hysteresis value... When using composite insulators, the sensitivity coefficient As it increases, at the same rate of change in humidity... Below, a larger sensitivity coefficient value makes Faster growth leads to higher confidence weights The confidence weights rapidly approach 0, meaning they respond more quickly to real-time humidity trends.

[0049] Thus, by introducing asymmetric confidence weights based on humidity change trends and adaptively adjusting them in conjunction with the hygroscopic hysteresis characteristics of the material, spurious biases caused by the hygroscopic hysteresis effect of the insulation material are effectively eliminated, and the adaptability of the monitoring system under extreme weather changes is greatly improved.

[0050] S104, calculate the difference between the measured current and the theoretical current to obtain the prediction residual, and calculate the fluctuation coefficient of the prediction residual within a preset time window.

[0051] In one embodiment, the predicted residual reflects the degree to which the measured value deviates from the theoretical baseline. A fluctuation coefficient is used to statistically analyze the degree of oscillation of the predicted residual over a period of time. This fluctuation coefficient distinguishes between contamination accumulation and sudden interference. The current increase caused by genuine surface contamination accumulation is typically slow and smooth; while electromagnetic interference or poor contact often manifests as high-frequency, violent oscillations.

[0052] First, calculate the prediction residual at the current time. The following relation is satisfied: ;in, For the measured current, The theoretical current output by the baseline model is used. The variance of the predicted residual within a preset time window is calculated, and the fluctuation coefficient is calculated accordingly. Satisfying the relation: ; In the formula, This is the volatility sensitivity coefficient. To predict the variance of the residuals within a preset time window.

[0053] Understandably, variance This directly reflects the dispersion of the prediction residual within the preset time window. When the prediction residual experiences severe oscillations due to sudden disturbances... Increased, leading to fluctuation coefficient Approaching 0 is equivalent to turning off the gating switch for subsequent updates of contamination drift; conversely, when the predicted residual changes smoothly, Smaller Approaching 1 allows for capturing slowly changing trends to accurately update dirt drift levels.

[0054] For example, the length of the preset time window can be set to 60 seconds, and the volatility sensitivity coefficient... The possible value is 10.

[0055] Thus, by constructing a gated switch by calculating the fluctuation coefficient, high-frequency interference signals can be effectively filtered out, ensuring that the update of the pollution drift amount can accurately reflect the smooth changes in insulation performance.

[0056] S105, the prediction residual is weighted based on the confidence weight and the fluctuation coefficient, and the contamination drift is iteratively updated. The contamination drift is used to characterize the slow current increment caused by surface contamination.

[0057] In one embodiment, the amount of pollution drift is a dynamic variable representing the long-term increase in leakage current caused by factors such as the accumulation of dust and salt deposits on the insulator surface. Since pollution accumulation is an extremely slow process, a weighted iterative approach is needed to gradually track this change.

[0058] Dirt drift The update satisfies the following relation: ; In the formula, The amount of filth drift at the current moment. The amount of filth drifted in the previous moment. To preset the learning rate, As confidence weights, For fluctuation coefficient, This represents the prediction residual at the current moment.

[0059] Understandably, through and The update rate of the combined control of dirt drift can only be achieved when the environment is in a non-hygroscopic state and the signal is stable and free from interference. and Only when both are relatively large is the prediction residual at the current moment considered to truly reflect the pollution level, and the pollution drift amount at the previous moment is updated accordingly to obtain the pollution drift amount at the current moment. Preset learning rate. This determines the update speed; for example, a value of 0.01 can be used to ensure that changes in the drift amount are smooth enough.

[0060] Thus, during real-time monitoring of insulation performance, each sampling point corresponds to a measured current, a theoretical current, and a pollution drift amount. Please refer to [link / reference needed]. Figure 2 This is a schematic diagram of the measured current, theoretical current, and pollution drift at each time sampling point according to the embodiments of this application.

[0061] In this way, by updating the amount of contamination drift through a dual-weighted iterative process, the current component caused by surface contamination can be accurately separated from the leakage current, thus avoiding misjudging it as a sudden fault.

[0062] S106, calculate the difference between the measured current and the sum of the theoretical current and the amount of pollution drift to obtain a fault determination value. If the fault determination value is greater than the alarm threshold, determine that there is an insulation fault.

[0063] In one embodiment, the fault determination value refers to the abnormal current component that cannot be explained by temperature, humidity, environmental factors, and surface contamination, accurately reflecting the fault characteristics of the insulation performance. The fault determination value satisfies the following relationship: ; In the formula, This is the fault determination value. For the measured current, For theoretical current, This represents the amount of filth drift at the current moment.

[0064] Understandably, in calculating the fault determination value, the measured current is decomposed into three parts: the theoretical current determined by temperature and humidity, the amount of dirt drift determined by dirt accumulation, and the fault current caused by insulation defects. The fault determination value corresponds to the fault current caused by insulation defects and directly reflects the true health status of the insulation material. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the fault determination values ​​at each time sampling point according to the embodiments of this application.

[0065] The alarm threshold can be set to 0.5mA. When the fault judgment value is greater than the alarm threshold, it indicates that there is a risk of insulation deterioration or breakdown, and an alarm signal will be issued immediately.

[0066] In this way, by peeling away the environmental and pollution effects layer by layer, pure fault characteristics are extracted from complex monitoring data, enabling real-time determination of insulation faults in the secondary circuit of disconnecting switches with high sensitivity and low false alarm rate.

[0067] According to a second aspect of this application, this application also provides a real-time monitoring system for the insulation performance of the secondary circuit of a disconnecting switch. Figure 4 This is a structural block diagram of a real-time monitoring system for the insulation performance of a secondary circuit of a disconnector switch, according to an embodiment of this application. Figure 4 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a real-time monitoring method for the insulation performance of the secondary circuit of a disconnector switch according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0068] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch, characterized in that, The monitoring method includes: collecting real-time temperature, real-time humidity, and measured current of the secondary circuit inside the disconnector switch housing; inputting the real-time temperature and real-time humidity into a pre-constructed benchmark model, and outputting the theoretical current under the current environment, wherein the benchmark model characterizes the mapping relationship between temperature, humidity, and leakage current of the disconnector switch in a clean state, and the theoretical current is the ideal leakage current value predicted after inputting the normalized real-time temperature and real-time humidity into the benchmark model; determining the confidence weight of the current moment based on the changing trend of the real-time humidity, wherein the confidence weight of the humidity rising phase is less than the confidence weight of the humidity falling phase; calculating the difference between the measured current and the theoretical current to obtain the prediction residual, and calculating the fluctuation coefficient of the prediction residual within a preset time window, including: calculating the fluctuation coefficient of the prediction residual within the preset time window. The variance within the window; calculating the product of a preset gating sensitivity and the variance, where the fluctuation coefficient is negatively correlated with the product; weighting the predicted residual based on the confidence weight and the fluctuation coefficient, and iteratively updating the contamination drift amount, including: calculating the comprehensive update gain, where the comprehensive update gain is the product of the confidence weight and the fluctuation coefficient; calculating the difference between the predicted residual and the contamination drift amount at the previous moment; adding the contamination drift amount at the previous moment to the product of the comprehensive update gain and the difference to obtain the contamination drift amount at the current moment, where the contamination drift amount is used to characterize the slow current increment caused by surface contamination; calculating the difference between the measured current and the sum of the theoretical current and the contamination drift amount to obtain a fault determination value, and determining that an insulation fault exists when the fault determination value is greater than the alarm threshold.

2. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 1, characterized in that, After collecting real-time temperature, real-time humidity, and measured current of the secondary circuit inside the disconnector switch box, the monitoring method further includes: performing a moving average filter on the real-time temperature, real-time humidity, and measured current to eliminate random noise; and performing a time-series alignment operation on the real-time temperature, real-time humidity, and measured current after standardization to form time-series data with a unified timestamp.

3. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 1, characterized in that, The baseline model is a support vector regression model. The baseline model uses normalized real-time temperature and real-time humidity as input features and outputs the leakage current corresponding to the input features.

4. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 3, characterized in that, The baseline model uses radial basis functions as kernel functions to map the nonlinear relationship between real-time temperature, real-time humidity and leakage current.

5. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 3, characterized in that, The process of constructing the benchmark model includes: selecting historical data of disconnecting switches in a healthy state as a training set; training the benchmark model using the training set until the insensitive loss function between the output of the benchmark model and the leakage current in a healthy state is less than a preset difference, thus obtaining the benchmark model.

6. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 1, characterized in that, The step of determining the confidence weight for the current moment based on the real-time humidity change trend includes: taking the difference between the real-time humidity and the humidity at the previous moment as the real-time humidity change trend; the confidence weight is negatively correlated with the product of a preset sensitivity coefficient and the change trend.

7. The method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector switch according to claim 6, characterized in that, The sensitivity coefficient is positively correlated with the moisture hysteresis value of the insulating material.

8. A real-time monitoring system for the insulation performance of the secondary circuit of a disconnector switch, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a method for real-time monitoring of the insulation performance of the secondary circuit of a disconnector according to any one of claims 1 to 7.

Images