A high-voltage circuit breaker residual life intelligent evaluation system

Abstract

The application relates to the technical field of circuit breaker life evaluation, in particular to an intelligent residual life evaluation system for a high-voltage circuit breaker, which comprises an operation data acquisition module, a degradation degree feature extraction module and a residual life calculation module; the operation data acquisition module is used for collecting operation data of the high-voltage circuit breaker; the degradation degree feature extraction module is used for extracting degradation degree features from the processed operation data; the residual life calculation module is used for constructing a residual life prediction model by using the degradation degree features and adopting a generalized regression neural network algorithm, and the residual life of the high-voltage circuit breaker is calculated; and the early warning and maintenance strategy module is used for formulating a predictive maintenance strategy for the high-voltage circuit breaker based on the residual life calculation result, combining a residual life probability density function and an expected loss rate of the circuit breaker, and balancing economy and reliability; the application can reflect the degradation process of the circuit breaker in real time, predict potential equipment faults in a timely manner, and avoid the delay and inaccuracy caused by traditional experience judgment and regular inspection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of circuit breaker life assessment, and particularly relates to a high-voltage circuit breaker residual life intelligent assessment system. BACKGROUND

[0002] At present, the traditional high-voltage circuit breaker health state assessment method mainly depends on regular maintenance and experience judgment. These methods usually cannot accurately reflect the running state of the circuit breaker in real time, and are difficult to cope with the complex situation of the gradual degradation of the circuit breaker. Regular maintenance is usually based on a set time interval, which may lead to premature or late maintenance decisions, thereby increasing the maintenance cost and the risk of equipment failure. In addition, the traditional method cannot dynamically adjust according to the actual operation data under different working conditions, thereby limiting the accuracy of the residual life prediction.

[0003] In recent years, with the development of intelligent sensing technology and big data analysis technology, online monitoring methods based on sensor data have gradually become a trend. These methods can assess the health state of the circuit breaker by collecting real-time operation data (such as current, voltage, temperature, and pressure) of the high-voltage circuit breaker and combining data analysis models. However, the existing monitoring methods based on sensor data still face several challenges:

[0004] Real-time dynamic assessment of degradation state: Although existing systems can collect real-time data, how to dynamically combine these real-time data with the complex changes in the device degradation process and accurately predict the residual life is still a difficult problem.

[0005] Balancing between economy and reliability: Existing maintenance strategies usually focus on one aspect of cost or safety, and lack an optimized solution that comprehensively considers the device residual life and maintenance cost, fault risk, and other factors.

[0006] Therefore, there is an urgent need for a new method that can dynamically assess the health state of the high-voltage circuit breaker based on real-time monitoring data and accurately predict the residual life, while taking into account economy and reliability, and proposing a reasonable predictive maintenance strategy. SUMMARY

[0007] Based on the above purpose, the present application provides a high-voltage circuit breaker residual life intelligent assessment system.

[0008] A high-voltage circuit breaker residual life intelligent assessment system, comprising:

[0009] An operation data acquisition module: collects the operation data of the high-voltage circuit breaker, including current data, voltage data, temperature data, and pressure data, and normalizes the collected high-voltage circuit breaker state data;

[0010] The degradation degree feature extraction module extracts the degradation degree features from the processed operation data, including current fluctuation amplitude features and voltage frequency features, temperature change rate, and pressure abnormal fluctuation features.

[0011] The residual life calculation module: using the degradation degree features, adopts the generalized regression neural network (GRNN) algorithm to construct a residual life prediction model, combines the real-time updated working state, evaluates the state of the circuit breaker, and calculates the residual life thereof;

[0012] The early warning and maintenance strategy module: based on the residual life calculation result, combines the residual life probability density function and the expected loss rate of the circuit breaker, formulates the predictive maintenance strategy of the high-voltage circuit breaker, and balances the economy and reliability.

[0013] Further, the operation data acquisition module arranges current sensors, voltage sensors, temperature sensors, and pressure sensors (oil-immersed circuit breakers) on the key components of the high-voltage circuit breaker, collects current data, voltage data, temperature data, and pressure data of the circuit breaker in different working states in real time, normalizes the collected operation data, and transmits the operation data to the degradation degree feature extraction module.

[0014] Further, the degradation degree feature extraction module specifically includes:

[0015] The current fluctuation amplitude feature extraction unit is configured to extract the fluctuation amplitude features from the processed current data, including calculating the standard deviation σ I of the current signal, calculating the peak-to-peak difference ΔI peak of the current signal based on a sliding window, and evaluating the load fluctuation of the circuit breaker in different operating states.

[0016] The voltage fluctuation frequency feature extraction unit is configured to extract the frequency features from the processed voltage data, including analyzing the frequency spectrum of the voltage signal by Fourier transform, calculating the frequency distribution and power spectral density of the voltage signal, and evaluating the frequency component P(f) of the voltage fluctuation.

[0017] The temperature change rate feature extraction unit is configured to extract the temperature change rate features from the processed temperature data, by calculating the change rate of the temperature signal in a unit time (such as first-order difference), and analyzing the thermal load change of the circuit breaker under different working conditions.

[0018] The pressure abnormal fluctuation feature extraction unit is configured to extract the abnormal fluctuation features from the processed pressure data, using abnormal fluctuation detection to identify the abnormal fluctuation ΔP in the pressure data, and thereby evaluating the leakage and pressure instability of the equipment.

[0019] Further, the standard deviation of the current signal in the current fluctuation amplitude feature extraction unit is calculated as:

[0020] Where I i represents the current data at the i-th moment, represents the average value of the current data, and N is the total number of sampling data points; the standard deviation σ I describes the fluctuation amplitude of the current signal, and a large standard deviation indicates that the current fluctuation is large, indicating that the load fluctuation or fault of the device occurs;

[0021] The peak-to-peak difference is calculated as: ΔI peak = I max - I min , where I max is the maximum value of the current signal, and I min is the minimum value of the current signal, and ΔI peak measures the maximum fluctuation range of the current signal.

[0022] Further, the Fourier transform analysis in the voltage fluctuation frequency feature extraction unit is represented as:

[0023] Where x n is the voltage data at the n-th moment, X(f) is the frequency domain component at frequency f, N is the number of sampling points, and f is the frequency. The Fourier transform converts the voltage signal from the time domain to the frequency domain, and analyzes its frequency components;

[0024] The power spectral density is represented as: P(f) = |X(f)| 2 , where P(f) is the power spectral density at frequency f, and X(f) is the frequency domain component after Fourier transform.

[0025] Further, in the temperature change rate feature extraction unit:

[0026] The change rate is calculated as: Where T(t) represents the temperature data at the t-th moment, Δt is the time interval, usually the sampling period, and the temperature change rate describes how fast the temperature changes over time.

[0027] Further, in the pressure abnormal fluctuation feature extraction unit, a deviation detection method is included:

[0028] Where P i represents the pressure data at the i-th moment, represents the average value of the pressure data, and N is the total number of sampling data points; the standard deviation σ P of the pressure indicates the amplitude of the pressure fluctuation;

[0029] The abnormal fluctuation detection is represented as: Wherein, a is a threshold coefficient, set as 3, indicating that the pressure data exceeding the range of 3 times of standard deviation is an abnormal fluctuation, P i is the pressure data at the i-th moment, is the average value of the pressure data, and P is the standard deviation of the pressure data.

[0030] Further, the construction of the remaining life prediction model specifically includes:

[0031] Degradation feature input unit: input the degradation degree features, including the standard deviation of the current signal I , the peak-peak difference of the current signal abnormal , the frequency component of the voltage fluctuation P(f), the temperature change rate and the abnormal fluctuation of the pressure data ΔP to the generalized regression neural network GRNN model for remaining life prediction;

[0032] Generalized regression neural network GRNN model: based on the degradation degree features, the remaining life prediction model is constructed by a supervised learning method, and the training process of the GRNN model includes the following:

[0033] Using historical data set for training, the historical data set includes the actual working data of the circuit breaker and the corresponding remaining life data;

[0034] In the training, according to the input degradation features I , ΔI peak , P(f), ΔP and the corresponding remaining life, the mapping learning is carried out, the weights and parameters in the GRNN model are adjusted, so that the model can predict the remaining life of the circuit breaker under the given feature input;

[0035] In the prediction process, the GRNN model outputs a remaining life value according to the input degradation features, which represents the health status of the circuit breaker and the expected time of failure;

[0036] Real-time state updating unit: real-time acquisition of the current working state of the circuit breaker, including current, voltage, temperature and pressure data, and real-time degradation features calculated by the degradation feature extraction module are taken as the input of the GRNN model to update the remaining life prediction;

[0037] Remaining life calculation unit: according to the output of the GRNN model, the remaining life under the current working state is calculated, and through comprehensive analysis of the health status of the circuit breaker, the estimated value of the remaining life is provided.

[0038] Further, the input of the generalized regression neural network GRNN model is the degradation feature, and the output is the remaining life R. The generalized regression neural network GRNN model processes the input features, and the predicted output value is represented as:

[0039] wherein, is the predicted value of the model, represents the remaining life of the circuit breaker, N is the number of samples of the training data set, and M is the dimension of the input features, which is 5 here (including σ I , ΔI peak , P(f), ΔP), is the feature vector of the ith sample, y i is the true remaining life value corresponding to the ith sample, σ is the width parameter of the Gaussian kernel function, which controls the smoothness of the model, x ij is the jth feature of the ith sample.

[0040] Further, the early warning and maintenance strategy module specifically includes:

[0041] A remaining life probability density function generation unit generates a probability density function f lifespan (t) of the remaining life of the high-voltage circuit breaker based on the remaining life calculation result, which represents the probability distribution of the remaining life of the circuit breaker at different time points t. The probability density function f lifespan (t) is fitted by using the maximum likelihood estimation method through statistical analysis of historical data combined with the output of the prediction model, to obtain the probability distribution of the remaining life of the circuit breaker.

[0042] An expected loss rate calculation unit calculates the expected loss rate L lifespan based on the remaining life probability density function f exp of the high-voltage circuit breaker and the maintenance cost, which is the expected economic loss due to equipment failure or failure to maintain in time. The calculation formula of the expected loss rate is:

[0043] wherein, t0 is the current time, T is the expected service life of the circuit breaker, C repair is the cost of each maintenance, R(t) is the frequency or probability of inspection or replacement at time point t, C failure is the repair cost after failure, and P failure (t) is the probability of failure of the circuit breaker at time t.

[0044] A maintenance strategy optimization unit optimizes the maintenance strategy based on the expected loss rate L exp in combination with the remaining life probability density function f lifespan(t), formulate a predictive maintenance strategy for high-voltage circuit breakers. The predictive maintenance strategy balances economy and reliability by optimizing the objective function:

[0045] min τ (L exp (τ))subject toP failure (τ)≤, where τ represents the choice of maintenance timing (such as the time threshold for delayed maintenance), L exp (τ) is the expected loss rate at maintenance opportunity τ, P failure (τ) represents the probability of circuit breaker failure at maintenance opportunity τ, which is an acceptable failure risk threshold (such as a certain maximum failure probability).

[0046] Beneficial effects of the present invention:

[0047] The present invention introduces multi-dimensional degradation characteristics (including current fluctuation amplitude, frequency component, temperature change rate and pressure fluctuation characteristics) and combines them with the generalized regression neural network (GRNN) algorithm to construct a remaining life prediction model for high-voltage circuit breakers. Compared with traditional methods, this method can more accurately evaluate the status of high-voltage circuit breakers based on real-time working status, accurately calculate the remaining life, reflect the degradation process of the circuit breaker in real time, and promptly predict potential equipment failures, avoiding the delays and inaccuracies caused by traditional empirical judgment and periodic inspections.

[0048] The present invention adopts the method of independently analyzing degradation characteristics, avoiding the information loss and errors that may be caused by weighted fusion. In the process of extracting the fluctuation characteristics of current, voltage, temperature and pressure, the impact of each feature on the health status of the circuit breaker is deeply analyzed through multiple dimensions such as standard deviation, peak difference, and frequency component. This ensures the independence and comprehensiveness of each feature information, can more carefully capture the degradation trend of the circuit breaker under different operating conditions, and improve the reliability and accuracy of degradation analysis. In particular, when facing multiple potential failure modes, different degradation processes can be more effectively distinguished.

[0049] This invention combines the remaining life probability density function of high-voltage circuit breakers with the expected loss rate to propose a predictive maintenance strategy based on real-world data and failure risk assessment. This strategy not only considers the potential failure risk of the equipment but also comprehensively balances economic efficiency and reliability. By calculating the expected loss rate, optimal maintenance timing can be selected in actual maintenance, avoiding premature or late maintenance interventions, significantly reducing maintenance costs, and maximizing the lifespan of the equipment, ensuring the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only belong to the present application, and other drawings can also be obtained by those skilled in the art without creative effort.

[0051] Fig. 1 The evaluation system of the embodiment of the present application is shown in the figure.

[0052] Fig. 2 The construction of the residual life prediction model of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail in combination with specific embodiments.

[0054] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by those skilled in the art to which the present application belongs. The terms "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. The terms "include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, and do not exclude other elements or objects. The terms "connect" or "connect" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. The terms "up", "down", "left", "right" and the like are only used to represent relative positional relationship, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0055] As shown in Figs. 1-2 A residual life intelligent evaluation system of high-voltage circuit breaker, comprising:

[0056] An operation data acquisition module: collecting operation data of the high-voltage circuit breaker, including current data, voltage data, temperature data and pressure data, and normalizing the collected high-voltage circuit breaker state data;

[0057] A degradation degree feature extraction module: extracting degradation degree features from the processed operation data, including current fluctuation amplitude features and voltage frequency features, temperature change rate and pressure abnormal fluctuation features;

[0058] A residual life calculation module: using the degradation degree features, adopting a generalized regression neural network (GRNN) algorithm to construct a residual life prediction model, combining real-time updated working state to evaluate the state of the circuit breaker and calculate its residual life;

[0059] Early warning and maintenance strategy module: Based on the remaining life calculation results, combined with the remaining life probability density function and expected loss rate of the circuit breaker, a predictive maintenance strategy for the high-voltage circuit breaker is formulated to balance economy and reliability, provide operators with intelligent maintenance or replacement suggestions, and optimize maintenance cycles and operation strategies.

[0060] The operation data acquisition module arranges current sensors, voltage sensors, temperature sensors and pressure sensors (oil-immersed circuit breakers) on the key components of the high-voltage circuit breaker to collect the current data, voltage data, temperature data and pressure data of the circuit breaker under different working conditions in real time, normalizes the collected operation data, and transmits it to the degradation degree feature extraction module.

[0061] The degradation degree feature extraction module specifically includes:

[0062] The current fluctuation amplitude feature extraction unit is used to extract the fluctuation amplitude feature from the processed current data, including calculating the standard deviation σ of the current signal I , calculate the peak-to-peak difference ΔI of the current signal based on the sliding window peak , to evaluate the load fluctuation of the circuit breaker under different operating conditions;

[0063] a voltage fluctuation frequency feature extraction unit, configured to extract frequency features from the processed voltage data, including analyzing the frequency spectrum in the voltage signal through Fourier transform, calculating the frequency distribution of the voltage signal and its power spectrum density, and thereby evaluating the frequency component P(f) of the voltage fluctuation;

[0064] The temperature change rate feature extraction unit is used to extract the temperature change rate feature from the processed temperature data by calculating the change rate of the temperature signal per unit time. (such as first-order difference), analyze the thermal load changes of the circuit breaker under different working conditions;

[0065] The pressure abnormal fluctuation feature extraction unit is used to extract abnormal fluctuation features from the processed pressure data, and use abnormal fluctuation detection to identify abnormal fluctuations ΔP in the pressure data, so as to evaluate the leakage and pressure instability of the equipment.

[0066] The standard deviation of the current signal in the current fluctuation amplitude feature extraction unit is calculated as:

[0067] Among them, I i represents the current data at the i-th moment, represents the average value of current data, N is the total number of sampling data points; standard deviation σ IThe fluctuation amplitude of the current signal is described, and a large standard deviation indicates that the current fluctuation is large, indicating that the load fluctuation of the device or the fault occurs;

[0068] The peak-to-peak difference is calculated as: ΔI peak = I max - I min , where I max is the maximum value of the current signal, I min is the minimum value of the current signal, and ΔI peak measures the maximum fluctuation range of the current signal, which can reflect the current impact when the load changes or the switch operates.

[0069] The Fourier transform analysis in the voltage fluctuation frequency feature extraction unit is represented as:

[0070] where x n is the voltage data at the nth time, X(f) is the frequency domain component at frequency f, N is the number of sampling points, and f is the frequency. The Fourier transform converts the voltage signal from the time domain to the frequency domain, analyzes its frequency components, and the low-frequency components in the frequency spectrum are usually related to steady-state voltage fluctuations, while high-frequency components indicate transient faults or system instability.

[0071] The power spectral density is represented as: P(f) = |X(f)| 2 , where P(f) is the power spectral density at frequency f, and X(f) is the frequency domain component after Fourier transform. The power spectral density (PSD) characterizes the frequency distribution of the voltage signal, and analyzing the power distribution of different frequency components helps to evaluate the stability of the device and the electrical load fluctuation.

[0072] In the temperature change rate feature extraction unit:

[0073] The change rate is calculated as: where T(t) represents the temperature data at the tth time, Δt is the time interval, usually the sampling period, and the temperature change rate describes the speed of temperature change over time, and a large change rate usually indicates that the temperature fluctuation is large, which is the result of load change or poor heat dissipation of the device;

[0074] It also includes cumulative temperature change calculation: where T(t) is the temperature data at the tth time, and N is the total number of sampling points.

[0075] In the pressure abnormal fluctuation feature extraction unit, the deviation detection method is included:

[0076] where P i represents the pressure data at the ith time, The average value of the pressure data, N is the total number of sampling data points, and the standard deviation of the pressure is σ P The amplitude of the pressure fluctuation, if the pressure fluctuation is too large, it indicates that the gas leaks or the system pressure is unstable, thereby affecting the arc extinguishing performance of the circuit breaker;

[0077] The abnormal fluctuation detection is represented as: Wherein, α is the threshold coefficient, which is set to 3, indicating that the pressure data exceeding 3 times the standard deviation is an abnormal fluctuation, P i is the pressure data at the i-th moment, is the average value of the pressure data, and σ P is the standard deviation of the pressure data.

[0078] The construction of the residual life prediction model specifically includes:

[0079] Degradation feature input unit: input the degradation degree features, including the standard deviation σ I of the current signal, the peak-to-peak difference ΔP abnormal of the current signal, the frequency component P(f) of the voltage fluctuation, the temperature change rate and the abnormal fluctuation ΔP in the pressure data to the generalized regression neural network GRNN model for residual life prediction;

[0080] Generalized regression neural network GRNN model: based on the degradation degree features, the residual life prediction model is constructed by a supervised learning method, and the training process of the GRNN model includes the following:

[0081] Using historical data set for training, the historical data set includes the actual working data of the circuit breaker and the corresponding residual life data;

[0082] In the training, according to the input degradation features σ I , ΔI peak , P(f), ΔP and the corresponding residual life, the mapping learning is carried out, the weights and parameters in the GRNN model are adjusted, so that the model can predict the residual life of the circuit breaker under the given feature input;

[0083] In the prediction process, the GRNN model outputs a residual life value according to the input degradation features, which represents the health status of the circuit breaker and the expected time of failure;

[0084] Real-time state updating unit: real-time acquisition of the current working state of the circuit breaker, including current, voltage, temperature and pressure data, and real-time degradation features calculated by the degradation feature extraction module, as the input of the GRNN model, to update the residual life prediction;

[0085] The remaining life calculation unit calculates the remaining life under the current working state according to the output of the GRNN model, and provides an estimated value of the remaining life by comprehensively analyzing the health state of the circuit breaker.

[0086] The input of the generalized regression neural network GRNN model is the degradation feature, and the output is the remaining life R. The generalized regression neural network GRNN model processes the input features and predicts the output value, which is expressed as:

[0087] wherein, is the predicted value of the model, represents the remaining life of the circuit breaker, N is the number of samples of the training data set, and M is the dimension of the input feature, which is 5 here (including σ I , ΔI peak , P(f), ΔP), is the feature vector of the i-th sample, y i is the real remaining life value corresponding to the i-th sample, σ is the width parameter of the Gaussian kernel function, which controls the smoothness of the model, x ij is the j-th feature of the i-th sample.

[0088] GRNN training process:

[0089] First, prepare the degradation features and the corresponding real remaining life data as the training data set;

[0090] Calculate the Gaussian kernel function: for each sample, calculate the distance between the input features and other samples in the training set, and use the Gaussian kernel function for weighting;

[0091] Model training: use the least squares method to train the model to find suitable weights (by adjusting σ and other hyperparameters) to minimize the error between the predicted value and the real value.

[0092] Remaining life prediction: after the model is trained, for the real-time acquired circuit breaker state (including real-time current, temperature, pressure, etc.), as input, the GRNN model will output a predicted remaining life value according to the learned pattern, which represents the current circuit breaker's remaining effective working time or the expected time to failure.

[0093] Real-time state update and life prediction: by periodically or continuously acquiring the real-time running data of the circuit breaker (such as current, voltage, temperature, pressure, etc.), the above feature extraction module will update the degradation feature value and input it into the already trained GRNN model to generate the latest remaining life prediction value, so that the health status of the circuit breaker can be monitored in real time and support for subsequent maintenance decisions can be provided.

[0094] The early warning and maintenance strategy module specifically includes:

[0095] A remaining life probability density function generation unit generates a probability density function f of the remaining life of the high-voltage circuit breaker based on the remaining life calculation result lifespan (t), representing the probability distribution of the remaining life of the circuit breaker at different time points t, the probability density function f lifespan (t) is fitted by statistical analysis of historical data, combined with the output of the prediction model, using the maximum likelihood estimation method, to obtain the probability distribution of the remaining life of the circuit breaker.

[0096] An expected loss rate calculation unit calculates the expected loss rate L lifespan (t) and the maintenance cost of the high-voltage circuit breaker, according to the probability density function f exp of the remaining life, which is the expected economic loss due to equipment failure or failure to maintain in a timely manner, and the calculation formula of the expected loss rate is:

[0097] Where t0 is the current time, T is the expected life of the circuit breaker, C repair is the cost of each maintenance, R(t) is the frequency or probability of maintenance or replacement at time t, C failure is the repair cost after failure, P failure (t) is the probability of failure of the circuit breaker at time t.

[0098] A maintenance strategy optimization unit formulates a predictive maintenance strategy for the high-voltage circuit breaker based on the expected loss rate L exp and the probability density function f lifespan (t) of the remaining life, the predictive maintenance strategy balances economy and reliability by optimizing the objective function:

[0099] min τ (L exp (τ))subject toP failure (τ)≤, where τ represents the selection of maintenance timing (such as the time threshold for delayed maintenance), L exp (τ) is the expected loss rate at maintenance timing τ, P failure (τ) represents the probability of failure of the circuit breaker at maintenance timing τ, and is the acceptable failure risk threshold (such as a certain maximum failure probability).

[0100] It also includes an intelligent maintenance decision unit, which provides intelligent maintenance or replacement recommendations for operating personnel based on the optimal maintenance timing τ opt output by the maintenance strategy optimization unit, guiding the actual maintenance cycle and operation strategy.

[0101] Those skilled in the art should understand that the above discussion of any embodiment is merely exemplary in nature and is not intended to imply that the present application is limited to these examples; any of the above embodiments or technical features among different embodiments can be combined, and steps can be implemented in any order, under the idea of the present application, and there are many other changes of different aspects of the present application as described above, which are not provided in details for the sake of brevity.

[0102] The present application is intended to cover all such alternatives, modifications, and variations as fall within the broad scope of the claims. Accordingly, any and all such alternatives, modifications and variations should be included within the scope of the present application.

Claims

1. A high-voltage circuit breaker residual life intelligent evaluation system, characterized in that, The application relates to a high-voltage circuit breaker remaining life prediction method and device. The operation data acquisition module collects the operation data of the high-voltage circuit breaker, including current data, voltage data, temperature data and pressure data, and normalizes the collected high-voltage circuit breaker operation data; The degradation degree feature extraction module extracts the degradation degree features from the processed operation data, including current fluctuation amplitude features, voltage frequency features, temperature change rate and pressure abnormal fluctuation features; The remaining life calculation module uses the degradation degree features, adopts a generalized regression neural network algorithm, constructs a remaining life prediction model, combines real-time updated working states, evaluates the state of the circuit breaker and calculates the remaining life thereof; The early warning and maintenance strategy module formulates a predictive maintenance strategy for the high-voltage circuit breaker based on the remaining life calculation result, combines the remaining life probability density function and the expected loss rate of the circuit breaker and balances the economy and reliability.

2. The system of claim 1, wherein, The operation data acquisition module arranges current sensors, voltage sensors, temperature sensors and pressure sensors on key components of the high-voltage circuit breaker, collects current data, voltage data, temperature data and pressure data of the circuit breaker under different working states in real time, normalizes the collected operation data, and transmits the normalized operation data to the degradation degree feature extraction module.

3. The system of claim 1, wherein the system further comprises a plurality of sensors configured to measure a plurality of parameters of the circuit breaker. The degradation degree feature extraction module specifically comprises: a current fluctuation amplitude feature extraction unit for extracting a fluctuation amplitude feature from the processed current data, including calculating the standard deviation of the current signal a peak-to-peak difference of the current signal based on a sliding window to evaluate the load fluctuation of the circuit breaker under different operating states; a voltage fluctuation frequency feature extraction unit for extracting frequency features from the processed voltage data, including analyzing the frequency spectrum in the voltage signal by Fourier transform, calculating the frequency distribution of the voltage signal and its power spectral density to assess the frequency components of the voltage fluctuation ; a temperature change rate feature extraction unit configured to extract a temperature change rate feature from the processed temperature data by calculating a change rate of the temperature signal in a unit time , analyze the thermal load change of the circuit breaker under different working conditions; The pressure abnormal fluctuation feature extraction unit is configured to extract abnormal fluctuation features from the processed pressure data, and identify abnormal fluctuations in the pressure data by using abnormal fluctuation detection Thus, the presence of leaks and pressure instability in the equipment can be assessed.

4. The system of claim 3, wherein the system further comprises a plurality of sensors configured to measure the operating parameters of the circuit breaker. The standard deviation of the current signal in the current fluctuation amplitude feature extraction unit is calculated as: wherein, represents the current data at the time instant, represents the average value of the current data, is the total number of sampled data points; the standard deviation describes the fluctuation amplitude of the current signal, a large standard deviation indicates a large current fluctuation, indicating a load fluctuation or a fault occurrence of the device; The peak-to-peak difference is calculated as: wherein is the maximum value of the current signal, is the minimum value of the current signal, measures the range of extreme fluctuations of the current signal.

5. The system of claim 4, wherein the system further comprises a plurality of sensors configured to measure the operating parameters of the circuit breaker. The Fourier transform analysis in the voltage fluctuation frequency feature extraction unit is expressed as: wherein, is the voltage data at the time instant, is the frequency frequency domain component at the frequency, is the number of sampling points, is the frequency, the Fourier transform converts the voltage signal from the time domain to the frequency domain, analyzing its frequency components; The power spectral density is represented as: wherein, is the frequency at which the power spectral density, is the frequency domain component after Fourier transformation.

6. The system of claim 5, wherein the system further comprises a plurality of sensors configured to measure the operating parameters of the circuit breaker. In the temperature change rate feature extraction unit: The rate of change is calculated as: where, denotes the temperature data at the time instant, is the time interval, typically the sampling period, and the rate of change of temperature describes how fast the temperature changes over time.

7. The system of claim 6, wherein the system further comprises a plurality of sensors configured to measure the operating parameters of the circuit breaker. In the pressure abnormal fluctuation feature extraction unit, the bias detection method comprises: wherein, represents the pressure data at the time t, represents the average of the pressure data, is the total number of sampled data points, the standard deviation of the pressure represents the amplitude of the pressure fluctuations; The abnormal fluctuation detection is represented as: wherein, is a threshold coefficient, is the pressure data at the moment in time, is the average value of the pressure data, is the standard deviation of the pressure data.

8. The system of claim 7, wherein the system further comprises a plurality of sensors configured to measure the operating parameters of the circuit breaker. The remaining life prediction model specifically comprises: degradation feature input unit: inputs the degradation degree features, including the standard deviation of the current signal , the peak-to-peak difference of the current signal , the frequency component of the voltage fluctuation , the temperature change rate , and the abnormal fluctuation in the pressure data as inputs to the generalized regression neural network (GRNN) model for residual life prediction; The generalized regression neural network (GRNN) model: based on the degradation degree features, a remaining life prediction model is constructed through a supervised learning method, and the training process of the GRNN model comprises the following: The historical data set is used for training, and the historical data set comprises actual working data and corresponding remaining life data of the circuit breaker; In the training, according to the input degradation features With the corresponding remaining life mapping learning, adjusting the weight and parameter in the GRNN model, so that the model can predict the remaining life of the circuit breaker under the given feature input; In the prediction process, the GRNN model outputs a remaining life value according to the input degradation features, and the value represents the health state of the circuit breaker and the expected time of failure; The real-time state updating unit acquires the current working state of the circuit breaker in real time, including current, voltage, temperature and pressure data, and real-time degradation features calculated through the degradation feature extraction module are used as the input of the GRNN model to update the remaining life prediction; The remaining life calculation unit calculates the remaining life under the current working state according to the output of the GRNN model, and provides an estimated value of the remaining life by comprehensively analyzing the health state of the circuit breaker.

9. The system of claim 8, wherein the system further comprises a plurality of sensors configured to measure the operating conditions of the circuit breaker. The generalized regression neural network GRNN model input is a degradation feature, and the output is a remaining life The generalized regression neural network GRNN model processes the input features, and the predicted output value is expressed as: wherein, is the predicted value of the model, representing the remaining life of the circuit breaker, is the number of samples in the training dataset, is the dimension of the input feature, is the feature vector of the i-th sample, is the feature vector of the i-th sample, is the true residual life value corresponding to the i-th sample, is the true residual life value corresponding to the i-th sample, is the width parameter of the Gaussian kernel function, controlling the smoothness of the model, is the j-th feature of the i-th sample, is the j-th feature of the i-th sample, is the j-th feature of the i-th sample.

Images