Intelligent multi-dimensional fault prediction and diagnosis system for power transformation equipment

Abstract

The invention relates to the technical field of power transformation equipment fault prediction, in particular to a power transformation equipment-oriented intelligent multi-dimensional fault prediction and diagnosis system, which comprises a data acquisition module used for acquiring real-time operation data of power transformation equipment; the fault multi-dimensional prediction module is used for establishing a fault prediction model, and the fault prediction model analyzes real-time fault data according to the real-time operation data of the power transformation equipment; the fault diagnosis module is used for analyzing a high-probability time period of fault occurrence according to the fault probability and extracting real-time operation data of the high-probability time period; performing fault detection on the real-time operation data in the high-probability time period by adopting a modulus maximum detection algorithm based on wavelet transform, and generating a fault diagnosis result; and the decision support module is used for generating a fault maintenance strategy according to the fault diagnosis result, the fault position, the fault type and the fault trend. By adopting the scheme, multi-dimensional analysis of the fault data can be realized, and the fault detection accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation equipment fault prediction, and in particular relates to an intelligent fault multi-dimensional prediction and diagnosis system for substation equipment. BACKGROUND

[0002] As the core hub of the power system, substation equipment bears the key functions of voltage transformation, power distribution and system protection, and its reliability directly affects the safe and stable operation of the power grid. Substation equipment includes transformers, high-voltage switchgear, distribution devices, compensation devices, relay protection devices and automation equipment, etc. The transformer is the core of the substation equipment, which uses the principle of electromagnetic induction to realize the increase or decrease of alternating voltage to meet the power demand of different users. The high-voltage switchgear is responsible for turning on or off the current in the circuit to realize the switching and protection of the line. The distribution device is responsible for distributing the power output by the transformer to each power line. The compensation device is used to improve the power factor of the power system and improve the system operation efficiency. With the rapid development of the power system and the advancement of the intelligent process, the running state of the substation equipment as an important part of the power system is directly related to the safe and stable operation of the power system. However, due to the influence of electrical, mechanical, environmental and other factors, various faults such as insulation aging, overheating and mechanical wear will inevitably occur in the long-term operation of the substation equipment. Therefore, during the operation of the substation equipment, various possible faults need to be predicted and diagnosed.

[0003] Although the existing substation equipment fault monitoring system can obtain a large amount of operation data, most of the systems only focus on a single fault indicator, such as temperature threshold alarm, and lack deep analysis of fault data, resulting in insufficient support for operation and maintenance decision information, and thus causing excessive maintenance or insufficient maintenance. In addition, the conventional threshold detection has insufficient sensitivity to early weak faults, which may lead to the evolution of weak faults into breakdown, explosion and other malignant accidents. Therefore, it is urgent to provide an intelligent fault multi-dimensional prediction and diagnosis system for substation equipment, which can realize multi-dimensional analysis of fault data and improve the fault detection accuracy. SUMMARY

[0004] The present application provides an intelligent fault multi-dimensional prediction and diagnosis system for substation equipment, which can realize multi-dimensional analysis of fault data and improve the fault detection accuracy.

[0005] In order to achieve the above purpose, the present application provides the following technical scheme:

[0006] An intelligent fault multi-dimensional prediction and diagnosis system for substation equipment, comprising a data acquisition module, a fault multi-dimensional prediction module, a fault diagnosis module and a decision support module.

[0007] The data collection module is configured to collect real-time operation data of the power transformation equipment.

[0008] The fault multi-dimensional prediction module is configured to establish a fault prediction model, which is configured to analyze real-time fault data according to the real-time operation data of the power transformation equipment, wherein the real-time fault data comprises a fault location, a fault type, a fault probability and a fault trend.

[0009] The fault diagnosis module is configured to analyze a high-probability period of fault occurrence according to the fault probability and extract real-time operation data of the high-probability period, and to perform fault detection on the real-time operation data of the high-probability period by using a modulus maximum value detection algorithm based on wavelet transform and generate a fault diagnosis result.

[0010] The decision support module is configured to generate a fault maintenance strategy according to the fault diagnosis result, the fault location, the fault type and the fault trend.

[0011] Further, the system further comprises a data preprocessing module configured to perform cleaning, denoising and normalization processing on the real-time operation data.

[0012] Further, the fault multi-dimensional prediction module comprises a historical data management module, a feature screening module, a model construction module, a model training module and a prediction analysis module.

[0013] The historical data management module is configured to store historical operation data and corresponding historical fault data of the power transformation equipment.

[0014] The feature screening module is configured to evaluate the importance of the historical operation data by using a ReliefF algorithm and generate an importance evaluation result, and generate a training feature set according to the importance evaluation result.

[0015] The model construction module is configured to establish a fault prediction model.

[0016] The model training module is configured to train the fault prediction model by using the training feature set.

[0017] The prediction analysis module is configured to analyze real-time fault data according to the real-time operation data of the power transformation equipment by using the trained fault prediction model.

[0018] Further, the fault multi-dimensional prediction module further comprises a prediction result output module.

[0019] The prediction result output module is configured to generate visual early warning information according to the real-time fault data.

[0020] Further, the prediction result output module is configured to generate a fault heat map according to the fault position and the fault probability, and to generate a fault trend map according to the fault trend.

[0021] Further, the fault maintenance strategy includes one or more of repair time, repair personnel and spare part demand.

[0022] Further, the fault maintenance strategy includes one or more of repair time, repair personnel and spare part demand.

[0023] Further, the fault maintenance strategy includes one or more of repair time, repair personnel and spare part demand.

[0024] The principles and advantages of the present application are that:

[0025] 1. In the present scheme, the real-time operation data is analyzed by the fault multi-dimensional prediction module to generate fault probability, fault position, fault type and fault trend, so that the operation and maintenance personnel can obtain clear early warning information before the actual occurrence of the fault, the operation and maintenance work is moved forward, and the active defense capability of the power grid operation is greatly improved. In addition, based on the fault probability, the real-time operation data of the high-probability period is accurately extracted, the further analysis range is significantly reduced, and the calculation load is reduced. Specifically, for the operation data of the high-probability period, a modulus maximum value detection algorithm based on wavelet transform is used for fault detection. The wavelet transform has excellent time-frequency localization characteristics, can effectively capture singular points in the signal such as transient state, impact, etc., and the modulus maximum value detection can accurately locate the fault feature points such as partial discharge, arc, voltage and current mutation, is very sensitive to early and weak fault signals, greatly improves the sensitivity, accuracy and positioning accuracy of fault detection, and improves the reliability of the fault diagnosis result.

[0026] 2. The decision support module integrates the real-time fault data of the fault multi-dimensional prediction module and the accurate result of the fault diagnosis module, intelligently generates an operable fault maintenance strategy, including the optimal repair time window, the required skilled personnel and the key spare part demand. Provides a scientific basis for operation and maintenance scheduling, and avoids blindness and resource waste.

[0027] 3. The prediction data is converted into intuitive visual warning information. Specifically, the fault heat map can intuitively display the fault risk level of different position equipment, and the fault trend map can clearly present the possible path of fault development. The man-machine interaction efficiency of information is improved, so that the operation and maintenance personnel can quickly master the overall situation and make better decisions.

[0028] 4. The ReliefF algorithm is used to evaluate and screen the importance of the massive historical operation data, and a training feature set is generated. The redundant and irrelevant features are removed, so that the fault prediction model constructed is more concise, more efficient, has stronger generalization ability, and the prediction result is more reliable. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 The logical block diagram of an embodiment of the intelligent fault multi-dimensional prediction and diagnosis system for power transformation equipment of the application is shown. DETAILED DESCRIPTION

[0030] The following will be further described in detail through specific embodiments:

[0031] Embodiment 1:

[0032] Embodiment 1 is basically as shown in the accompanying drawings: Figure 1

[0033] An intelligent fault multi-dimensional prediction and diagnosis system for power transformation equipment, as shown in the accompanying drawings, comprises a data acquisition module, a data preprocessing module, a fault multi-dimensional prediction module, a fault diagnosis module, a decision support module, a fault alarm module and a display module. Figure 1

[0034] The data acquisition module is used to acquire real-time operation data of the power transformation equipment. The operation data includes voltage, current and temperature data, and the acquired real-time operation data is transmitted to the data preprocessing module synchronously. Specifically, the electrical parameters and state information of the power transformation equipment are monitored in real time through current sensors, voltage sensors and temperature sensors; the sensor signals are collected, converted and preliminarily processed by a data acquisition terminal; and the collected data is transmitted to a data center or a server through a communication network. The real-time and accuracy of the data are ensured, and the basic data support is provided for subsequent data preprocessing, fault prediction and diagnosis.

[0035] The data preprocessing module is used to clean, denoise and normalize the real-time operation data to ensure the quality and consistency of the data, and then pass the processed data to the fault multi-dimensional prediction module and the fault diagnosis module. In this scheme, a low-pass filter is used to filter the real-time operation data, that is, the received signals, and the calculation formula is as follows:

[0036]

[0037] In the formula, y[n] represents the value of the filtered output signal at time index n, n and k are both time indexes, h[k] is the coefficient value of the filter impulse response at position k, x[.] represents the original input signal of the filtering process, and N represents the order of the filter.

[0038] ​​In this scheme, the minimum-maximum normalization method is used to normalize the signal amplitude, which is convenient for subsequent processing and analysis. The calculation formula is as follows:

[0039]

[0040] In the formula, z[n] is the value of the normalized signal at time index n, x[n] represents the original input signal after normalization processing, min(x) is the minimum value in the signal, max(x) is the maximum value in the signal, and n is the time index.

[0041] The fault multi-dimensional prediction module is configured to establish a fault prediction model, which analyzes real-time fault data according to real-time operation data of the power transformation equipment, wherein the real-time fault data includes fault location, fault type, fault probability and fault trend; the fault multi-dimensional prediction module includes a historical data management module, a feature screening module, a model construction module, a model training module, a prediction analysis module and a prediction result output module.

[0042] The historical data management module is configured to store, manage and maintain historical operation data and corresponding historical fault data of the power transformation equipment. The module supports data classification, indexing and query functions, facilitating users to quickly retrieve device operation conditions or historical fault information in a specific time period. In addition, the historical data management module also has data cleaning and preprocessing functions, which can eliminate outliers and fill in missing data to ensure data quality and consistency. Through regular backup and recovery mechanisms, the module ensures the security and reliability of the data. Historical data not only provides training and validation data sets for real-time prediction analysis module, but also provides a solid data foundation for subsequent fault diagnosis, trend analysis and decision support. Overall, the historical data management module is an important data support platform for realizing intelligent fault prediction and diagnosis of power transformation equipment.

[0043] The feature screening module is configured to screen features that have important influence on fault prediction and diagnosis from the preprocessed historical operation data. Specifically, the importance of the historical operation data is evaluated by using a ReliefF algorithm to generate an importance evaluation result, and a training feature set is generated according to the importance evaluation result. The evaluation method is as follows:

[0044] The feature dimension is extracted from the historical operation data, and the i-th feature dimension is denoted as Xi. The system traverses all samples to find the nearest neighbor samples of the same class and different classes for each sample j. The feature difference diff(Xi,j,H) and diff(Xi,j,M) are calculated, wherein diff(Xi,j,H) represents the difference between sample j and the nearest neighbor of the same class in feature dimension Xi, and diff(Xi,j,M) represents the difference between sample j and the nearest neighbor of different classes in feature dimension Xi. The feature weight is calculated by using the following formula:

[0045]

[0046] In the formula, W(Xi) represents the weight of feature Xi, and m represents the number of samples.

[0047] The features are sorted in descending order of weight to generate importance assessment results. The top k features with the highest weights from these importance assessment results are selected to form the training feature set for model training. Through feature selection and optimization, redundant features can be effectively reduced, improving model training efficiency and prediction accuracy.

[0048] The model building module is used to establish a fault prediction model; specifically, it uses the Support Vector Machine (SVM) machine learning algorithm to build the fault prediction model. The model training module is used to train the fault prediction model using a training feature set; specifically, it finds the optimal classification hyperplane using the training data in the training feature set, i.e., it solves for the optimal parameter β such that the distance from each support vector to the separating hyperplane is no less than 1, and there are no misclassifications; this process involves solving a quadratic programming problem, ultimately obtaining the model parameters. During training, the system uses methods such as cross-validation to optimize the model parameters and improve the model's generalization ability. After training, the model can be used for classification prediction of new data to identify whether there are faults in substation equipment.

[0049] The kernel function uses a Gaussian kernel, and the calculation formula is as follows:

[0050]

[0051] In the formula, K(xi,xj) is the Gaussian kernel function, representing the similarity measure of samples xi and xj in the high-dimensional feature space. xi and xj represent the feature vectors of samples i and j, respectively. exp(·) represents the exponential function, ∥·∥ represents the norm operator, and σ represents the hyperparameter. Through this formula, the system maps the input data to a high-dimensional space, finds the optimal classification hyperplane, and achieves fault classification.

[0052] The predictive analysis module is used to analyze real-time fault data based on the real-time operating data of the substation equipment using a trained fault prediction model. Specifically, the predictive analysis module uses a logistic regression algorithm to perform predictive analysis on the real-time operating data using the trained fault prediction model. The specific calculation steps include: for the real-time input feature vector x, based on the model parameters β0, β1, ..., βn, the prediction formula is:

[0053]

[0054] In the formula, P(Y=1|x) represents the probability of the fault occurring, x represents the feature set of a single sample, and β iThe model parameters are represented. Through this formula, the system calculates the failure probability corresponding to the real-time data to judge the equipment state.

[0055] In this embodiment, the failure location and failure trend are predicted by artificial intelligence, taking the real-time operation parameters, failure probability and failure classification as inputs of the input layer, and taking the failure location and failure trend as outputs of the output layer, and specifically adopting a BP neural network model.

[0056] The prediction result output module is configured to generate visual warning information according to the real-time failure data, and present the analysis results of the failure multi-dimensional prediction module to the user in an intuitive and easy-to-understand manner. Specifically, according to the failure type, failure location and failure probability, a report, a chart or a visual interface is generated; according to the failure location and failure probability, a failure heat map is generated to intuitively display the failure risk degree of different equipment or parts through color depth; according to the failure trend, a failure trend chart is generated to show the change of the failure probability over time. In addition, the module also supports generating a detailed failure report listing specific failure diagnosis results, possible failure causes and recommended maintenance suggestions. The user can easily view this information through the system interface to facilitate quick decision-making and take appropriate maintenance measures. The design of the prediction result output module focuses on user experience to ensure accurate information transmission and efficient use.

[0057] The failure diagnosis module is configured to analyze high-probability time periods of failure occurrence according to the failure probability, and extract real-time operation data of the high-probability time periods; specifically, a time period within N minutes before the failure probability is higher than a probability threshold is recorded as a high-probability time period, and real-time operation data of the high-probability time period is extracted. The failure diagnosis module is also configured to use a modulus maximum value detection algorithm based on wavelet transform to perform failure detection on the real-time operation data of the high-probability time period, and generate a failure diagnosis result; specifically:

[0058] The modulus maximum value of the real-time operation data of the high-probability time period is generated by performing wavelet transform, and the calculation formula is as follows:

[0059] M g，k = max(|W g,k |)

[0060] In the formula, W g,k is a wavelet coefficient of a scale g and a position k, M g，k represents a modulus maximum value of the wavelet coefficient of the scale g and the position k.

[0061] The modulus maximum value is screened to remove noise and other non-failure caused modulus maximum values, and to retain possible failure features; by setting a threshold, modulus maximum values exceeding the threshold are screened out; first, a dynamic threshold T g of the scale g is calculated, and the calculation formula is as follows:

[0062]

[0063] where T g is the dynamic threshold of scale g, and σ g is the noise standard deviation estimate of scale g (estimated by the median of wavelet coefficients), and N is the signal length.

[0064] The screening condition is: retaining the modulus maxima satisfying M g，k > T g , and eliminating invalid modulus maxima caused by noise.

[0065] The screened modulus maxima are subjected to fault feature identification, which needs to satisfy the following fault feature conditions simultaneously:

[0066] 1. Amplitude condition:

[0067] M g，k > A threshold

[0068] where A threshold is a preset fault amplitude threshold.

[0069] 2. Duration condition:

[0070] D g，k > D threshold

[0071] where D g，k is the duration scale number of the modulus maxima M g，k in multiple scales, and D threshold is a preset fault duration threshold.

[0072] Fault detection decision: according to the identified fault features, a fault detection decision is made. Specifically, if there is at least one modulus maxima satisfying the fault feature conditions, it is determined that the equipment has a fault, otherwise, it is determined that the equipment has no fault.

[0073] The fault diagnosis module realizes the detection of the fault of the power transformation equipment through the steps of extracting the modulus maxima of the wavelet coefficients, screening the modulus maxima, identifying the fault features, and making a fault detection decision; the calculation formula and parameter definition are adjusted and optimized according to specific applications and signal characteristics; the algorithm utilizes the multi-scale analysis capability of the wavelet transform to effectively detect the fault features in the power transformation equipment signals, and provides accurate information for subsequent fault positioning.

[0074] The decision support module is configured to generate a fault maintenance strategy based on the fault diagnosis result, fault location, fault type, and fault trend. The fault maintenance strategy includes one or more of repair time, repair personnel, and spare part requirements. In this embodiment, the operation content includes: analyzing the fault diagnosis result to determine the severity and impact of the fault; generating repair recommendations and spare part requirements based on historical repair data and a maintenance knowledge base; developing a repair plan, including repair time, repair personnel, and spare part requirements; and providing fault warnings and preventive maintenance recommendations to reduce the probability of failure. The decision support module helps users make scientific and reasonable maintenance decisions through these functions, improving the efficiency and reliability of the equipment.

[0075] The fault alarm module is configured to generate fault alarm information based on the real-time fault data and the fault diagnosis result.

[0076] The display module is configured to display the real-time operation data, real-time fault data, fault diagnosis result, and fault maintenance strategy. The display module is an interactive interface between the system and the user, responsible for displaying system information and receiving user operations. The operation content includes: displaying real-time operation data, real-time fault data, fault diagnosis result, and fault maintenance strategy; receiving user queries and instructions, such as parameter settings, report generation, and system configuration; designing an intuitive and friendly interface to improve user experience. The user interface module enables effective communication between the user and the system through these functions, allowing the system to better serve user needs.

[0077] In this scheme, by integrating data collection, preprocessing, multi-dimensional prediction, diagnosis, and decision support modules, a closed-loop processing flow for fault prediction and diagnosis of power transformation equipment is constructed. The fault multi-dimensional prediction module relies on high-quality historical data provided by the historical data management module, combined with precise feature selection by the feature selection module, and advanced modeling capabilities of the model construction and training module (using support vector machine algorithm), to efficiently generate multi-dimensional prediction results including fault probability, type, location, and trend. These prediction results provide the basis for the fault alarm module, and also serve as key inputs to drive the fault diagnosis module for accurate fault location and feature analysis (such as using the modulus maximum detection algorithm based on wavelet transform). Overall, the system improves the accuracy and timeliness of fault prediction through multi-dimensional data fusion and intelligent algorithm collaboration, providing strong technical support for preventive maintenance and rapid fault response of power transformation equipment.

[0078] In addition, after receiving the multi-dimensional prediction result, the fault diagnosis module in the scheme uses a wavelet transform-based modulus maximum value detection algorithm to analyze the running state in depth, accurately locates the fault and identifies its type, and generates a diagnosis result based on the high-probability period data. The decision support module then generates a maintenance strategy including repair recommendations, spare parts requirements, and maintenance plans based on the diagnosis result, fault location, type, and trend. This complete technical chain from multi-dimensional prediction, accurate diagnosis to scientific decision not only enhances the depth and positioning accuracy of fault analysis, but also greatly improves the scientificity and foresight of maintenance decisions. Users can intuitively view the system running state, analysis results, and maintenance strategies through the display module and perform necessary operations. In summary, the system realizes comprehensive intelligent monitoring and proactive maintenance of substation equipment faults, effectively improving equipment operation reliability and operation and maintenance efficiency.

[0079] Embodiment 2:

[0080] Embodiment 2 has the same basic principles as Embodiment 1, except that in Embodiment 2, a system management module is further included, which is responsible for the configuration, monitoring, and management of the entire system. Its running content includes: user management, setting user permissions and roles to ensure system security; system configuration, adjusting system parameters according to user needs and environmental changes; data management, responsible for data backup, recovery, and migration; log management, recording system running status and operation logs for fault troubleshooting and system optimization; monitoring and alarm, real-time monitoring of system running status, and timely alarm for abnormalities. Through these management functions, the system management module ensures stable operation and efficient management of the system.

[0081] The above is only an embodiment of the present application, and common knowledge of specific structures and properties in the scheme is not described in detail. Those skilled in the art know all ordinary technical knowledge in the field of the application before the filing date or the priority date, can know all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Those skilled in the art can improve and implement the scheme based on their own abilities under the guidance of this application, and some typical known structures or methods should not be an obstacle to the implementation of the present application. It should be noted that for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should be considered within the scope of protection of the present application. The scope of protection of the present application should be subject to the content of its claims, and the specific embodiments in the specification can be used to explain the content of the claims.

Claims

1. A smart fault multi-dimensional prediction and diagnosis system for power transformation equipment, characterized in that: The system comprises a data acquisition module, a fault multi-dimensional prediction module, a fault diagnosis module and a decision support module. The data acquisition module is configured to acquire real-time operation data of the power transformation equipment. The fault multi-dimensional prediction module is configured to establish a fault prediction model. The fault prediction model is configured to analyze real-time fault data according to the real-time operation data of the power transformation equipment. The real-time fault data comprises a fault location, a fault type, a fault probability and a fault trend.

2. The intelligent fault multi-dimensional prediction and diagnosis system for power transformation equipment according to claim 1, characterized in that: The fault diagnosis module is configured to analyze a high-probability period of fault occurrence according to the fault probability and extract real-time operation data of the high-probability period.

3. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 1, characterized in that: The fault diagnosis module is configured to perform fault detection on the real-time operation data of the high-probability period by using a modulus maximum value detection algorithm based on wavelet transform and generate a fault diagnosis result. The decision support module is configured to generate a fault maintenance strategy according to the fault diagnosis result, the fault location, the fault type and the fault trend. The system further comprises a data preprocessing module configured to perform cleaning, denoising and normalization processing on the real-time operation data. The fault multi-dimensional prediction module comprises a historical data management module, a feature screening module, a model construction module, a model training module and a prediction analysis module. The historical data management module is configured to store historical operation data and corresponding historical fault data of the power transformation equipment. The feature screening module is configured to evaluate the importance of the historical operation data by using a ReliefF algorithm and generate an importance evaluation result.

4. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 1, characterized in that: The feature screening module is configured to generate a training feature set according to the importance evaluation result. The model construction module is configured to establish a fault prediction model.

5. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 4, characterized in that: The model training module is configured to train the fault prediction model by using the training feature set. The prediction analysis module is configured to analyze real-time fault data according to the real-time operation data of the power transformation equipment by using the trained fault prediction model.

6. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 1, characterized in that: The fault multi-dimensional prediction module further comprises a prediction result output module.

7. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 1, characterized in that: The prediction result output module is configured to generate visual warning information according to the real-time fault data.

8. The transformer equipment oriented intelligent fault multi-dimensional prediction and diagnosis system according to claim 1, characterized in that: The prediction result output module is configured to generate a fault heat map according to the fault location and the fault probability. The prediction result output module is further configured to generate a fault trend graph according to the fault trend. The fault maintenance strategy comprises one or more of a repair time, repair personnel and spare parts demand. The system further comprises a fault alarm module configured to generate fault alarm information according to the real-time fault data and the fault diagnosis result. The system further comprises a display module configured to display the real-time operation data, the real-time fault data, the fault diagnosis result and the fault maintenance strategy.

Images