Dielectric loss intelligent prediction and fault diagnosis system combined with current analysis

Abstract

The invention discloses a dielectric loss intelligent prediction and fault diagnosis system combined with current analysis, and relates to the technical field of electrical equipment insulation detection and fault diagnosis, and the system comprises a data collection module which is used for collecting current data of equipment in different temperature and humidity environments in real time, the system comprises a multi-dimensional current transmission characteristic database, a model training module used for training a neural network model based on the multi-dimensional current transmission characteristic database to obtain compensation parameters for different environment conditions, and a data analysis module used for inputting collected actual current data into the trained neural network model for real-time analysis, the fault diagnosis and early warning module is used for determining the dielectric loss condition of the equipment based on the corrected current characteristics and carrying out fault diagnosis and early warning. According to the dielectric loss intelligent prediction and fault diagnosis system combined with current analysis, the problem of dielectric loss misdiagnosis caused by environmental factors is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of insulation detection and fault diagnosis of power equipment, and particularly to an intelligent prediction and fault diagnosis system for dielectric loss combined with current analysis. Background Art

[0002] In the insulation condition assessment of power equipment, the dielectric loss factor (tanδ) is an important parameter for measuring the health status of insulating media. However, the test results of dielectric loss are affected by external environmental factors, especially the changes in temperature and humidity. These factors can cause changes in current transmission characteristics, thereby introducing measurement errors. For example, an increase in temperature may enhance the conductivity of the insulating medium, resulting in an increase in the dielectric loss value, while an increase in humidity may cause an increase in surface leakage current, thus affecting the accuracy of dielectric loss measurement. Existing dielectric loss test methods usually do not consider the real-time compensation of these environmental factors, leading to measurement errors under different environmental conditions and even possible misdiagnosis of the equipment insulation status.

[0003] Currently, for the compensation of environmental impacts on dielectric loss testing, some studies use fixed temperature and humidity correction curves for compensation. However, this method is usually based on static experimental data and cannot adapt to the dynamic changes under complex working conditions. In addition, most traditional dielectric loss test equipment is based on fixed measurement parameters and it is difficult to achieve adaptive regulation of environmental factors. As a result, under extreme environments such as high humidity and high temperature, the measurement errors are further amplified, affecting the reliability of test results. Therefore, there is an urgent need for an intelligent prediction and fault diagnosis system combined with current analysis to achieve compensation regulation for the differences in current transmission characteristics under different temperature and humidity environments, improve the accuracy of dielectric loss measurement, and reduce the misdiagnosis rate. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent prediction and fault diagnosis system for dielectric loss combined with current analysis to compensate and regulate the differences in current transmission characteristics under different temperature and humidity environments to solve the problem of misdiagnosis of dielectric loss caused by environmental factors.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An intelligent prediction and fault diagnosis system for dielectric loss combined with current analysis, the system includes: A data acquisition module, which is used to collect current data of the equipment in different temperature and humidity environments in real time and construct a multi-dimensional current transmission characteristic database; A model training module connected to the data acquisition module, which is used to train a neural network model based on the multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions. When training the neural network, the signal of the current is modeled so that the neural network model learns the probability relationship of the feature changes before and after the current. The specific formula is: ; Among them, A n represents the current value of the current characteristic at the current moment, and A n－1 represents the current value of the current characteristic at the previous moment. P(A n ∣A n－1 ) represents the probability of the current current characteristic state transitioning to the next current characteristic state. q represents the learning weight between current characteristics, and n represents the time series index at the current moment; The data analysis module connected to the model training module is used to input the collected actual current data into the trained neural network model for real-time analysis and output the corrected current characteristics, including determining the contribution value of the current characteristics to fault diagnosis, statistically analyzing the calculation time required for each characteristic, and calculating the efficiency of each current characteristic. The specific formula is: B = C / T; Among them, B represents the efficiency of the current characteristic, C represents the contribution value of the current characteristic to fault diagnosis, and T represents the calculation time required for the current characteristic; The fault diagnosis and early warning module connected to the data analysis module is used to determine the dielectric loss condition of the device based on the corrected current characteristics and perform fault diagnosis and early warning.

[0006] Preferably, the data acquisition module collects the current data of the device in different temperature and humidity environments in real time, constructs a multi-dimensional current transmission characteristic database including the current signals of the operating state of the acquisition device, simultaneously monitors the fluctuation range of the environmental temperature and humidity, and calculates the signal strength of the collected signals. The specific formula is: D = D 0 / g; Among them, D represents the current intensity of the collected current characteristic, D 0 represents the original intensity of the device signal source, and g represents the fluctuation range of temperature and humidity.

[0007] Preferably, the fault diagnosis and early warning module determines the dielectric loss condition of the device based on the corrected current characteristics and performs fault diagnosis and early warning, including determining the fault occurrence probability and potential loss corresponding to the currently detected current characteristics, and calculating the priority of each fault. The specific formula is: H = Q×U; Among them, H represents the priority value of the current fault early warning, Q represents the fault occurrence probability corresponding to the currently detected current characteristics, and U represents the degree of loss that the fault may cause; All potential faults are sorted in descending order according to the priority value of the fault early warning, and warning signals are sent out in turn.

[0008] Preferably, the data acquisition module includes a temperature and humidity sensor and a current sensor, which are used to synchronously collect environmental temperature and humidity data and current data, and store the collected data in the multi-dimensional current transmission characteristic database.

[0009] Preferably, the data analysis module inputs the collected actual current data into the trained neural network model for real-time analysis, and outputs the contribution value C of the current feature to fault diagnosis in the corrected current feature. The calculation formula is: C = ∣V 1 −V 2 ∣×k; Wherein, C represents the contribution value of the current feature to fault diagnosis, V 1 represents the value of the current feature in the fault state, V 2 represents the value of the current feature in the normal state, and k represents the normalization coefficient.

[0010] Preferably, the data acquisition module collects the current data of the device in different temperature and humidity environments in real time, and constructs the calculation formula of the temperature and humidity fluctuation amplitude g in the multi-dimensional current transmission characteristic database as: ; Wherein, g represents the temperature and humidity fluctuation amplitude, L max represents the maximum humidity recorded in the current monitoring period, L min represents the minimum humidity recorded in the current monitoring period, M max represents the maximum temperature recorded in the current monitoring period, M min represents the minimum temperature recorded in the current monitoring period.

[0011] Preferably, the fault diagnosis and early warning module determines the dielectric loss condition of the device based on the corrected current feature, and the calculation formula of the fault occurrence probability Q corresponding to the current detected current feature in the fault diagnosis and early warning is: Q = N 1 / N 2 ; Wherein, Q represents the fault occurrence probability corresponding to the current detected current feature, N 1 represents the number of times the current feature value causes a fault in the historical data, N 2 represents the total number of times the feature value appears in the historical data.

[0012] Preferably, the fault diagnosis and early warning module determines the dielectric loss condition of the device based on the corrected current feature, and the calculation formula of the loss degree U that the fault may cause in the fault diagnosis and early warning is U = Y 1 +Y 2 ; Wherein, U represents the loss degree that the fault may cause, Y 1 represents the maintenance cost of the equipment fault, Y 2 represents the downtime loss cost caused by the fault.

[0013] Preferably, the model training module trains a neural network model based on a multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions, and further includes: Selecting a training sample set from the multi-dimensional current transmission characteristic database through K-fold cross-validation; Setting the learning rate w to satisfy 0 < w ≤ 0.01, where w represents the learning rate; In each training iteration cycle, if the loss function loss satisfies the following condition, stop training: loss ≤ λ, where loss represents the loss function and λ represents the threshold of the loss function.

[0014] Preferably, the expression form of the loss function loss in the model training module training a neural network model based on a multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions is: loss = ; where loss represents the loss function, F represents the number of training samples, represents the true value, represents the model prediction value, and i represents the index of the current training sample.

[0015] It can be seen from the above technical solutions that the present invention has the following beneficial effects: The intelligent prediction and fault diagnosis system for dielectric loss combined with current analysis collects the current data of the equipment in different temperature and humidity environments in real time through the data acquisition module, and constructs a multi-dimensional current transmission characteristic database. The model training module trains a neural network model based on the multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions. The data analysis module inputs the collected actual current data into the trained neural network model for real-time analysis and outputs the corrected current characteristics. The fault diagnosis and early warning module determines the dielectric loss status of the equipment based on the corrected current characteristics and conducts fault diagnosis and early warning. It can collect environmental parameters such as temperature and humidity in real time, dynamically compensate for the changes in current transmission characteristics under different environmental conditions, significantly improve the accuracy of the dielectric loss factor test results, can adaptively adjust the compensation parameters, overcome the limitations of the traditional fixed temperature and humidity correction curve, effectively respond to the dynamic environmental changes under complex working conditions, can perform real-time correction for the enhanced conductivity of the insulating medium caused by temperature increase and the change of surface leakage current caused by humidity increase, reduce the measurement error introduced by environmental factors, avoid misdiagnosis of the equipment insulation status caused by test deviation, significantly improve the adaptability of the measuring equipment under variable environmental conditions and the reliability of the test results, realize the real-time evaluation of the equipment insulation status and fault early warning, comprehensively improve the intelligent level of equipment operation status evaluation, realize the real-time perception and automatic compensation of environmental factors, simplify the calibration process of traditional test equipment, reduce the operation complexity, improve the test efficiency, and compensate and regulate the differences in current transmission characteristics under different temperature and humidity environments to solve the problem of dielectric loss misdiagnosis caused by environmental factors. Description of the Drawings

[0016] Figure 1 This is the connection diagram of the system modules of the present invention. Detailed Embodiments

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] As Figure 1 shown, the present invention provides a technical solution: an intelligent prediction and fault diagnosis system for dielectric loss combined with current analysis, the system includes: A data acquisition module, which is used to collect the current data of the equipment in different temperature and humidity environments in real time and construct a multi-dimensional current transmission characteristic database; A model training module connected to the data acquisition module, which is used to train a neural network model based on a multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions. When training the neural network, it models the current signal so that the neural network model learns the probability relationship of the current feature changes before and after. The specific formula is: ; where, A n represents the current feature value of the current moment, and A n－1 represents the current feature value of the previous moment. P(A n |A n－1 ) represents the probability of the current current feature state transitioning to the next current feature state. q represents the learning weight between current features, and n represents the time series index of the current moment; A data analysis module connected to the model training module, which is used to input the collected actual current data into the trained neural network model for real-time analysis and output the corrected current features. It includes determining the contribution value of the current features to fault diagnosis, statistically analyzing the calculation time required for each feature, and calculating the efficiency of each current feature. The specific formula is: B = C / T; where, B represents the efficiency of the current feature, C represents the contribution value of the current feature to fault diagnosis, and T represents the calculation time required for the current feature; A fault diagnosis and early warning module connected to the data analysis module, which is used to determine the dielectric loss condition of the equipment based on the corrected current features and perform fault diagnosis and early warning.

[0019] This system collects the current data during the operation of the equipment in real time through the data acquisition module and covers different temperature and humidity environments to build a multi-dimensional current transmission characteristic database. Through this database, the model training module uses the neural network algorithm to model the change trend of the current features and learn the probability relationship of the changes before and after. Specifically, this probability relationship is represented by the formula where q is the weight factor in the neural network learning process, indicating that the current feature value A n and the previous feature value A n－1Dependency relationship. After the model is trained, it can automatically compensate for the error of current feature data in different environments. Subsequently, the data analysis module inputs the actually collected current data into the trained model, calculates the contribution value of each feature to fault diagnosis, and calculates the efficiency B of each current feature based on the formula B = C / T. This analysis provides a reliable quantitative basis for fault diagnosis. Finally, the fault diagnosis and early warning module compares the corrected current feature values with the thresholds, quickly evaluates the dielectric loss state of the equipment, generates a diagnostic report, and issues an early warning signal under abnormal conditions to ensure the efficient operation and safety monitoring of the equipment. The present invention combines current analysis technology and neural network modeling technology, significantly improving the accuracy of dielectric loss prediction and the efficiency of fault diagnosis. First, the data acquisition module covers multi-dimensional environmental conditions to collect more comprehensive current data, ensuring the reliability of model training and analysis. Second, the neural network model can accurately learn the transformation law of the current feature state, eliminate the influence of environmental conditions on feature data based on compensation parameters, thereby improving the adaptability and robustness of the model. Third, the data analysis module quantitatively analyzes the contribution of different current features to fault diagnosis through the feature efficiency calculation formula B = C / T, providing data support for subsequent optimization. Finally, the fault diagnosis and early warning module realizes the full-process automation from current feature correction to dielectric loss state evaluation, and has a fast response ability, effectively reducing the risk of equipment failure and improving the safety and reliability of system operation. The overall system has significant advantages such as high efficiency, intelligence, and strong adaptability, and has broad application prospects in industrial equipment monitoring and fault early warning.

[0020] The data acquisition module collects the current data of the equipment in different temperature and humidity environments in real time, and constructs a multi-dimensional current transmission characteristic database including the current signals of the operating state of the acquisition equipment, while monitoring the fluctuation range of environmental temperature and humidity, and calculating the signal strength of the collected signals. The specific formula is: D = D 0 / g; where D represents the current feature strength collected currently, D 0 represents the original strength of the equipment signal source, and g represents the fluctuation range of temperature and humidity.

[0021] In the actual operation of this system, the data acquisition module monitors the current signal during the operation of the equipment in real time, and at the same time introduces the fluctuation range of temperature and humidity as a key environmental parameter for monitoring. In the data acquisition stage, the module obtains the original signal strength D 0 during the operation of the equipment and the real-time change data of environmental temperature and humidity. Through the formula: D = D 0Dynamically calculate the current feature intensity D and associate it with the environmental conditions. The system uses this data to construct a multi-dimensional current transmission characteristic database and comprehensively models the current features with temperature and humidity parameters. In this way, the collected signals not only reflect the operating state of the device but also contain the response characteristics to environmental factors, providing high-precision data support for subsequent model training and fault diagnosis. This implementation effectively improves the system's adaptability to environmental changes and makes the calculation of the current signal intensity more accurate by introducing the monitoring and dynamic adjustment of the temperature and humidity fluctuation amplitude g. Specifically, the system can capture the current features under the operating state of the device in real time and construct a multi-dimensional current transmission characteristic database in combination with environmental parameters. This method not only improves the comprehensiveness of data collection but also reduces the interference of environmental conditions on the signal intensity. By dynamically calculating D = D 0 / g, the accuracy and repeatability of the collected signal intensity are ensured, providing a reliable basis for subsequent model training, dielectric loss prediction, and fault diagnosis. In addition, this implementation enables the system to operate efficiently under complex working conditions, significantly enhancing the robustness and diagnostic accuracy of the system.

[0022] The fault diagnosis and early warning module determines the dielectric loss condition of the device based on the corrected current features and conducts fault diagnosis and early warning, including determining the probability of occurrence and potential losses of the faults corresponding to the currently detected current features, and calculating the priority of each fault. The specific formula is: H = Q × U; where H represents the priority value of the current fault early warning, Q represents the probability of occurrence of the fault corresponding to the currently detected current features, and U represents the degree of loss that the fault may cause; Sort all potential faults in descending order according to the priority value of the fault early warning and trigger the warning signals in sequence.

[0023] During actual operation, the fault diagnosis and early warning module calculates the probability of fault occurrence Q corresponding to each current feature by analyzing the corrected current eigenvalue. This value is generated based on the changing trend of the current feature obtained during the model training process and the correlation of specific fault modes. At the same time, the system also combines the operating state of the equipment and historical data to evaluate the potential loss degree U that each fault may cause. Through the formula H = Q × U, the system quantifies the priority of each fault and generates a priority value H reflecting the urgency of the fault. The system sorts all possible faults according to the priority value, arranging the fault handling order from largest to smallest. Based on the fault priority sorting, the module gives the early warning signal to the most urgent fault first, notifies the user or automatically activates corresponding protection measures, such as equipment load reduction or shutdown, to ensure operation safety. In addition, the system can continuously and dynamically update the priority value H, and adjust the fault handling order in a timely manner when environmental parameters or equipment states change, so as to achieve efficient and reliable fault management and intelligent early warning. This implementation method realizes the quantitative calculation of fault priority by introducing the probability of fault occurrence Q and potential loss U corresponding to the current feature, and has significant advantages compared with the traditional single fault monitoring method. First of all, the system can dynamically evaluate the fault correlation and potential impact of the current feature, so as to achieve more accurate fault priority division. Secondly, through the sorting mechanism of the priority value H, the system can identify the most urgent fault in time and send out the early warning signal quickly, greatly reducing the secondary losses caused by the delayed response to the fault. In addition, this module can adjust the priority order according to real-time data, significantly improving the adaptability and intelligent level of fault diagnosis and early warning. Generally speaking, this system can quickly and accurately evaluate and manage the equipment state under complex working conditions, significantly improving the safety and reliability of the system operation, and reducing the operation and maintenance costs at the same time.

[0024] The data acquisition module includes a temperature and humidity sensor and a current sensor, which are used to synchronously collect ambient temperature and humidity data and current data, and store the collected data in a multi-dimensional current transmission characteristic database. Through the integrated temperature and humidity sensor and current sensor, the data acquisition module monitors the temperature and humidity data of the environment and the current characteristic values under the operating state of the device in real time. In actual operation, the temperature and humidity sensor is responsible for recording the dynamic changes of environmental parameters, which may have a significant impact on the current transmission characteristics of the device. At the same time, the current sensor accurately captures the current data during the operation of the device, including the amplitude, frequency, and waveform characteristics of the current. The acquisition module synchronizes these two types of data in time to ensure that the current characteristics and environmental parameters can be accurately matched, thus forming a highly correlated data set. Subsequently, the module stores the synchronously collected data in a multi-dimensional current transmission characteristic database, providing comprehensive and reliable data support for subsequent model training, feature analysis, and fault diagnosis. This real-time synchronous acquisition mechanism not only improves the accuracy of the data but also significantly reduces the analysis errors caused by inconsistent data time series. Through the integration of the temperature and humidity sensor and the current sensor, this embodiment realizes the synchronous acquisition of environmental parameters and current data, and has the following remarkable advantages: The synchronous acquisition mechanism ensures the time consistency of the ambient temperature and humidity data and the current characteristic data, providing high-precision data support for subsequent modeling and analysis; The data acquisition module can capture the impact of environmental parameters on the current transmission characteristics, significantly improving the adaptability of the system under complex working conditions; Through the construction of a multi-dimensional current transmission characteristic database, the system has the ability to comprehensively describe the operating state of the device, providing rich reference data for the training of neural network models and fault diagnosis; The automated and structured management of data storage reduces the errors of manual operations, improves the data utilization efficiency, and lays a foundation for realizing intelligent prediction and diagnosis; By comprehensively using environmental parameters and current data, the accuracy of dielectric loss monitoring and the reliability of fault prediction are significantly improved.

[0025] The data analysis module inputs the collected actual current data into the trained neural network model for real-time analysis, and outputs the contribution value C of the current characteristic to fault diagnosis in the corrected current characteristic. The calculation formula of the contribution value C of the current characteristic to fault diagnosis is: C = ∣V 1 −V 2 ∣×k; Among them, C represents the contribution value of the current characteristic to fault diagnosis, V 1 represents the value of the current characteristic in the fault state, V 2 represents the value of the current characteristic in the normal state, and k represents the normalization coefficient.

[0026] The data analysis module dynamically analyzes these eigenvalue by receiving the current eigenvalues from the data acquisition module in real time and using a trained neural network model. During this process, the module first corrects the currently acquired current eigenvalues to eliminate environmental interference or measurement errors, obtaining the corrected current eigenvalues. Then, by comparing the values V 1 of the corrected current eigenvalues in the fault state and the values V 2 in the normal state, the change amplitude ∣V 1 -V 2 ∣ of the current eigenvalues is calculated. To facilitate the comparison and evaluation between different features, the system introduces a normalization coefficient k to make the calculation results of the contribution value C universal and comparable. Finally, the system outputs the calculated current feature contribution value C for use by the fault diagnosis module. This real-time calculation and output mechanism ensures that the fault diagnosis process can obtain accurate feature contribution information in a timely manner, thus significantly improving the efficiency and accuracy of diagnosis. This implementation method quantifies the contribution of eigenvalues to fault diagnosis by calculating the numerical difference of current features between the fault state and the normal state. Compared with traditional methods, this method has the following advantages: introducing the normalization coefficient k realizes the standardization of feature contribution values, facilitating the comparison and comprehensive evaluation of different features in fault diagnosis; the real-time analysis and correction functions ensure the timeliness and accuracy of data, greatly enhancing the system's adaptability to dynamic working conditions; using the quantization result of the feature contribution value C can more accurately identify the current features most sensitive to faults, optimizing the diagnosis process; providing a reliable basis for subsequent feature optimization and model improvement, improving the overall robustness and diagnosis accuracy of the system; dynamically outputting the corrected current features and their contribution values makes the system more reliable and practical in complex environments.

[0027] The data acquisition module collects the current data of the device in different temperature and humidity environments in real time, and the calculation formula for the temperature and humidity fluctuation amplitude g in the multi-dimensional current transmission characteristic database is: ; where g represents the temperature and humidity fluctuation amplitude, L max represents the maximum humidity recorded during the current monitoring period, L min represents the minimum humidity recorded during the current monitoring period, M max represents the maximum temperature recorded during the current monitoring period, M min represents the minimum temperature recorded during the current monitoring period.

[0028] The system monitors the environmental parameters during the operation of the device in real time through the data acquisition module, specifically including the dynamic changes in temperature and humidity. The module records the maximum and minimum values of temperature and humidity respectively in each monitoring cycle. Through the above formula, the system calculates the fluctuation range g of temperature and humidity. This value combines the comprehensive influence of temperature and humidity changes and can accurately reflect the fluctuation of environmental conditions. The calculated result g is stored in the multi-dimensional current transmission characteristic database, forming a corresponding relationship with the current data collected in real time. This fluctuation range is used to correct the environmental influence in the current characteristic value and provide key parameter support for subsequent model training and fault diagnosis. In addition, through this method, the system can accurately capture the dynamic change characteristics of the device operation environment, laying a foundation for improving the monitoring and diagnosis accuracy. It accurately reflects the changes in temperature and humidity, realizing the quantitative evaluation of environmental fluctuations; the fluctuation range g is stored in the multi-dimensional current transmission characteristic database synchronously with the current data, providing reliable data support for modeling the influence of the environment on the device performance; the system can dynamically adjust the current characteristic correction model according to the size of the fluctuation range, reducing the interference of environmental fluctuations on the diagnosis result and improving the robustness and adaptability of the diagnosis; combined with the real-time monitoring of the environmental fluctuation range, it provides personalized support for device fault diagnosis and status prediction under different working conditions; compared with the single environmental parameter monitoring method, this implementation method comprehensively considers the changes in temperature and humidity, improving the comprehensiveness and accuracy of environmental monitoring and data analysis.

[0029] The fault diagnosis and early warning module determines the dielectric loss condition of the device based on the corrected current characteristics and conducts fault diagnosis and early warning. The calculation formula for the fault occurrence probability Q corresponding to the currently detected current characteristics is: Q = N 1 / N 2 ; where Q represents the fault occurrence probability corresponding to the currently detected current characteristics, N 1 represents the number of times the current characteristic value caused a fault in the historical data, and N 2 represents the total number of times this characteristic value appears in the historical data.

[0030] The fault diagnosis and early warning module calculates the fault occurrence probability Q corresponding to the current current characteristic value through statistical analysis of historical data. In actual operation, the module first sorts and classifies the historical data, and matches each current characteristic value with its corresponding fault record. For a certain current characteristic value, it counts the number of times N 1 that this characteristic value caused a fault in the historical data and the total number of times N 2 that this characteristic value appears. Through the formula Q = N 1 / N 2, the system calculates the probability of failure Q corresponding to this eigenvalue. This probability value reflects the correlation between the current characteristics and the failure, providing a basis for quantitative assessment of the dielectric loss condition of the equipment. The module ranks the possible failure risks of the equipment according to the value of Q and gives priority to dealing with the high-probability failure characteristics. At the same time, the system can dynamically update N according to the real-time collected data 1 and N 2 , thereby continuously optimizing and adjusting the failure probability model to ensure the timeliness and accuracy of the diagnosis results. By statistically analyzing the relationship between current characteristics and failures using historical data, the scientificity and credibility of failure diagnosis are effectively improved; dynamically updating the probability of failure occurrence enables the system to adapt to changes in the equipment operation status and environmental conditions in a timely manner, enhancing the real-time and accuracy of diagnosis; ranking the characteristics according to the failure probability optimizes the priority management of failure warnings to ensure that high-risk failures can be processed in a timely manner; reducing the dependence on manual experience and realizing intelligent failure diagnosis through big data analysis improve the diagnosis efficiency; integrating the information of the failure occurrence probability Q provides data support for further optimizing the failure model and feature screening.

[0031] The failure diagnosis and warning module determines the dielectric loss condition of the equipment based on the corrected current characteristics and calculates the formula for the degree of loss U that may be caused by the failure in the failure diagnosis and warning as U = Y 1 + Y 2 ; where U represents the degree of loss that may be caused by the failure, and Y 1 represents the repair cost of the equipment failure, and Y 2 represents the cost of downtime loss caused by the failure.

[0032] During operation, the failure diagnosis and warning module calculates the overall degree of loss U of the failure by comprehensively evaluating the direct and indirect economic losses that may be caused by the equipment failure. Among them, the module first evaluates the repair cost Y 1 of a specific type of failure according to the equipment historical repair records and market repair cost standards. This value includes material replacement costs, labor repair costs, and related management costs. Subsequently, the module calculates the cost of downtime loss Y 2 caused by the failure according to the operation characteristics of the equipment, production plan, and the impact of downtime on production. This part is mainly obtained by estimating the production capacity loss and indirect economic impact during the downtime (such as compensation costs for customer delivery delays, etc.). Finally, through the formula U = Y 1 + Y 2, the module sums up the two parts of the cost to obtain the overall loss degree of the fault. This value can be used to optimize the equipment maintenance strategy and prioritize fault warnings, thereby achieving economic - oriented fault management. By comprehensively considering the direct repair cost and the indirect downtime loss cost, the economic impact of faults is comprehensively quantified; providing a fault priority assessment based on economic benefits provides a scientific basis for equipment managers to optimize the allocation of maintenance resources; dynamically updating the repair cost Y 1 and the downtime loss cost Y 2 , which can timely reflect the changes in market prices and production conditions, ensuring the real - time and accuracy of the calculation results. Reducing the limitations of traditional fault diagnosis that only relies on technical parameters, introducing economic factors into the fault diagnosis process, improving the practicality and decision - making support ability of the system; by quantifying the fault loss degree U, enhancing the system's perception and management ability of fault risks, and contributing to improving the economic benefits of equipment operation.

[0033] The model training module trains a neural network model based on the multi - dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions, and also includes: Selecting a training sample set from the multi - dimensional current transmission characteristic database through K - fold cross - validation; Setting the learning rate w to satisfy 0 < w ≤ 0.01, where w represents the learning rate; In each training iteration cycle, if the loss function loss satisfies the following conditions, then stop training: loss ≤ λ, where loss represents the loss function and λ represents the threshold of the loss function.

[0034] The model training module selects data from the multi - dimensional current transmission characteristic database for training, and uses the K - fold cross - validation technique to optimize the allocation of the training sample set, ensuring that each training covers enough sample features to improve the generalization ability of the model. In each training iteration, the model adjusts the weight parameters according to the set learning rate w to ensure that the weight update step size is appropriate and the training process converges smoothly. At the same time, the system calculates the loss function loss to evaluate the error of the model in each iteration. When the value of the loss function is less than or equal to the set threshold λ, the training is automatically stopped. This mechanism based on dynamic error monitoring and precise stopping enables the model to ensure high - precision output while converging quickly, thereby generating accurate parameters that can be compensated for different environmental conditions. Combining these optimization strategies, the model can more accurately adapt to the changes in current characteristics under different environmental conditions, generate reliable compensation parameters, provide accurate and efficient algorithm support for equipment fault diagnosis and prediction, and lay a solid foundation for intelligent management in industrial application scenarios.

[0035] The model training module trains a neural network model based on a multi-dimensional current transmission characteristic database to obtain the expression form of the loss function loss in the compensation parameters for different environmental conditions: loss = ; where loss represents the loss function, F represents the number of training samples, represents the true value, represents the model prediction value, and i represents the index of the current training sample.

[0036] The model training module uses the multi-dimensional current transmission characteristic database to optimize and train the neural network model for the current characteristics under different environmental conditions. During the training process, the system calculates the loss function loss according to the above formula, and its value represents the mean square error between the model prediction value and the corresponding true value. For each training iteration, the loss function loss is the core index to measure the current model prediction accuracy. By taking the number of training samples F as the denominator, the system realizes the normalization processing of all sample errors, so that loss can reflect the overall error level. After each iteration, the model adjusts the weights of the neural network through an optimization algorithm (such as the gradient descent method) to reduce the value of loss. When loss converges to the set threshold or meets the stop condition, the training terminates, and finally the model compensation parameters that meet the accuracy requirements are generated. This method ensures that the model can efficiently learn the change law of current characteristics and adapt to various environmental conditions. By introducing the loss function loss in the form of mean square error in this embodiment, the accuracy and robustness of model training are significantly improved. Specifically, the loss function can effectively measure the difference between the prediction value and the true value, and the model parameters are optimized by minimizing loss. In addition, the normalization processing ensures the adaptability of training to different-scale sample sets and avoids the interference of sample size differences on the model training effect. Combined with the dynamic adjustment mechanism of the loss function, the system can automatically adapt to the characteristic data under different environmental conditions, generate high-precision compensation parameters, and provide strong technical support for equipment dielectric loss prediction and fault diagnosis. Compared with the traditional method, this method not only improves the training efficiency of the model, but also greatly improves the reliability of prediction and diagnosis, laying a technical foundation for intelligent management in complex industrial application scenarios.

[0037] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dielectric loss intelligent prediction and fault diagnosis system combined with current analysis, characterized in that: The system comprises: Data acquisition module, used to collect the current data of the equipment in different temperature and humidity environments in real time, and build a multi-dimensional current transmission characteristics database; A model training module connected to the data acquisition module is used to train a neural network model based on a multi-dimensional current transmission characteristic database to obtain compensation parameters for different environments; The data analysis module connected to the model training module is used to input the collected actual current data into the trained neural network model for real-time analysis and output the corrected current characteristics, including determining the contribution value of the current characteristics to fault diagnosis, statistically analyzing the calculation time required for each characteristic, and calculating the efficiency of each current characteristic. The specific formula is: B = C / T; Among them, B represents the efficiency of current characteristics, C represents the contribution value of current characteristics to fault diagnosis, and T represents the calculation time required for current characteristics; The fault diagnosis and early warning module connected to the data analysis module is used to determine the dielectric loss status of the equipment based on the corrected current characteristics, and to perform fault diagnosis and early warning.

2. According to claim 1, a dielectric loss intelligent prediction and fault diagnosis system combined with current analysis is characterized in that: The data acquisition module collects the current data of the device in different temperature and humidity environments in real time, and constructs a multi-dimensional current transmission characteristic database, including: collecting the current signal of the device operation status, monitoring the fluctuation amplitude of the ambient temperature and humidity, and calculating the collected signal strength. The specific formula is: D = D0 / g; Among them, D represents the current characteristic intensity currently collected, D0 represents the original intensity of the device signal source, and g represents the fluctuation amplitude of temperature and humidity.

3. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 1 is characterized in that: Training a neural network model based on a multi-dimensional current transmission characteristic database to obtain compensation parameters for different environments includes: when training the neural network, modeling the current signal so that the neural network model learns the probability relationship between the changes in the current characteristics before and after. The specific formula is: ; Among them, A n Indicates the current characteristic value at the current moment, A n－1 Indicates the current characteristic value at the previous moment, P(A n ∣A n－1 ) represents the probability of the current characteristic state transitioning to the next current characteristic state, q represents the learning weight between current characteristics, and n represents the time series index at the current moment.

4. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 1 is characterized by: The fault diagnosis and early warning module determines the dielectric loss status of the equipment based on the corrected current characteristics, and performs fault diagnosis and early warning, including: determining the fault probability and potential loss corresponding to the current characteristics currently detected, and calculating the priority of each fault, the specific formula is: H = Q × U; Among them, H represents the priority value of the current fault warning, Q represents the fault occurrence probability corresponding to the current characteristics currently detected, and U represents the degree of loss that may be caused by the fault; All potential faults are sorted in descending order according to the fault warning priority value, and warning signals are triggered in sequence.

5. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 1 is characterized in that: The data analysis module inputs the collected actual current data into the trained neural network model for real-time analysis, and outputs the contribution value of the current feature to fault diagnosis in the corrected current feature. The calculation formula of the contribution value C of the current feature to fault diagnosis is: C = |V1-V2| × k; Among them, C represents the contribution value of the current feature to fault diagnosis, V1 represents the value of the current feature in the fault state, V2 represents the value of the current feature in the normal state, and k represents the normalization coefficient.

6. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 1 is characterized by: The data acquisition module collects the current data of the device in different temperature and humidity environments in real time, and constructs the calculation formula of the temperature and humidity fluctuation amplitude g in the multi-dimensional current transmission characteristic database as follows: ; Among them, g represents the fluctuation range of temperature and humidity, L max Indicates the maximum humidity value recorded in the current monitoring period, L min Indicates the minimum humidity value recorded in the current monitoring period, M max Indicates the maximum temperature recorded in the current monitoring period, M min Indicates the minimum temperature recorded in the current monitoring period.

7. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 4 is characterized by: The fault diagnosis and early warning module determines the dielectric loss status of the equipment based on the corrected current characteristics, and performs fault diagnosis and early warning. The calculation formula of the fault occurrence probability Q corresponding to the current characteristics currently detected is: Q=N1 / N2; Among them, Q represents the fault occurrence probability corresponding to the currently detected current feature, N1 represents the number of times the current feature value causes a fault in the historical data, and N2 represents the total number of times the feature value appears in the historical data.

8. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 4 is characterized by: The fault diagnosis and early warning module determines the dielectric loss condition of the equipment based on the corrected current characteristics, and the calculation formula for the loss degree U that may be caused by faults during fault diagnosis and early warning is U = Y1 + Y2; Among them, U represents the loss degree that may be caused by faults, Y1 represents the maintenance cost of equipment faults, and Y2 represents the downtime loss cost caused by faults.

9. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 3 is characterized by: The model training module trains a neural network model based on the multi-dimensional current transmission characteristic database to obtain compensation parameters for different environmental conditions, and also includes: Selecting a training sample set from the multi-dimensional current transmission characteristic database through K-fold cross-validation; Setting the learning rate w to satisfy 0 < w ≤ 0.01, where w represents the learning rate; In each training iteration cycle, if the loss function loss satisfies the following conditions, stop training: loss ≤ λ, where loss represents the loss function and λ represents the threshold of the loss function.

10. The dielectric loss intelligent prediction and fault diagnosis system combined with current analysis according to claim 9, characterized in that: The model training module trains the neural network model based on the multi-dimensional current transmission characteristic database to obtain the compensation parameters for different environmental conditions. The loss function loss is expressed as: loss = ; Among them, loss represents the loss function, F represents the number of training samples, represents the true value, Represents the model prediction value, and i represents the index of the current training sample.

Images