Power equipment fault prediction method, system and device

Abstract

The application relates to the technical field of power system prediction, and discloses a power equipment fault prediction method, system and device, the method comprising the following steps: acquiring power equipment working condition data; extracting multi-dimensional features from the working condition data, wherein the multi-dimensional features comprise time domain features, frequency domain features, trend features, periodicity features and correlation features; and predicting power equipment faults based on the multi-dimensional features and a pre-constructed fault prediction model; the fault prediction model is constructed through integrated learning training based on multiple machine learning algorithms, and is optimized through a cross-validation and hyperparameter optimization algorithm, compared with a traditional periodical maintenance and post-maintenance mode, the application can effectively replace the above mode, avoid excessive maintenance, insufficient maintenance and power failure accidents caused by the inability to predict faults, thereby improving power system power supply reliability, reducing operation and maintenance costs, and improving operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of power system prediction technology, specifically to methods, systems, and devices for predicting power equipment faults. Background Technology

[0002] In power systems, the stable operation of electrical equipment is crucial to ensuring the reliability of power supply. However, during long-term operation, electrical equipment inevitably experiences failures due to various factors such as electrical stress, mechanical stress, and environmental factors. Traditional power equipment maintenance methods mainly involve periodic inspections and reactive repairs. Periodic inspections suffer from over-maintenance or under-maintenance, wasting manpower, material resources, and financial resources, and failing to detect potential equipment failures in a timely manner; reactive repairs, on the other hand, can lead to power outages, causing serious economic losses to social production and daily life. Summary of the Invention

[0003] This invention provides a method, system, and apparatus for predicting power equipment faults, in order to solve the problem that traditional maintenance methods for power equipment, which rely on regular inspections and post-incident repairs, cannot predict faults in advance.

[0004] In a first aspect, the present invention provides a method for predicting faults in power equipment, the method comprising: Acquire operating data of power equipment; Multidimensional features are extracted from operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features. Based on multidimensional features and a pre-built fault prediction model, power equipment faults are predicted. The fault prediction model is constructed by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0005] This invention provides a method for predicting power equipment faults. By acquiring operating data and extracting multi-dimensional features such as time domain, frequency domain, trend, period, and correlation, it integrates multiple machine learning algorithms for ensemble learning training and constructs a prediction model through cross-validation and hyperparameter optimization, thereby achieving early prediction of power equipment faults. Compared with traditional periodic inspections and reactive maintenance, this invention can effectively replace these methods, avoiding over-maintenance, under-maintenance, and power outages caused by the inability to predict faults, thus improving the reliability of power supply, reducing operation and maintenance costs, and increasing operation and maintenance efficiency.

[0006] In one optional implementation, the operating data includes power equipment operating parameters and power system basic data; Acquire operating data of power equipment, including: Multiple sensors deployed at various preset monitoring locations on the power equipment are used to collect the operating parameters of the power equipment in real time, and the power system's basic data is collected in real time through the power terminal; the operating parameters of the power equipment include vibration signals, temperature signals and current signals.

[0007] This invention utilizes multiple sensors deployed at various pre-set monitoring locations on power equipment to collect real-time operating parameters, including vibration, temperature, and current signals. Simultaneously, it collects basic power system data in real-time through power terminals, constructing a multi-dimensional operating condition data acquisition system covering both the equipment's own status and the system's operating environment. Compared to traditional methods relying on a single data source or manual inspection, this invention can more comprehensively and in real-time acquire multi-source information reflecting the equipment's health status, providing a sufficient data foundation for subsequent fault prediction. This improves the accuracy and timeliness of fault prediction, avoiding prediction errors caused by missing or delayed data.

[0008] In one alternative implementation, multidimensional features are extracted from the operating condition data, including: Time-frequency analysis is used to extract time-domain and frequency-domain features from vibration signals; Statistical analysis methods are used to extract trend and periodic features from temperature and current signals; Correlation analysis was used to extract correlation features between different types of signals from the operating data.

[0009] This invention employs time-frequency analysis to extract time-domain and frequency-domain features from vibration signals, comprehensively capturing instantaneous changes and frequency distribution information in equipment vibration signals. Statistical analysis extracts trend and periodic features from temperature and current signals, effectively reflecting the evolution and periodic patterns of equipment temperature and current over time. Furthermore, correlation analysis extracts correlation features between different types of signals from operating data, revealing the intrinsic relationships between vibration, temperature, and current signals. Compared to traditional methods that extract only a single type of feature or ignore correlations between signals, this invention extracts more comprehensive and richer multidimensional features, providing more discriminative and representative information input for fault prediction models, thereby improving the accuracy and reliability of fault prediction.

[0010] In one optional implementation, extracting multidimensional features from the operating condition data further includes: Principal component analysis or linear discriminant analysis are used to reduce the dimensionality of the extracted multidimensional features.

[0011] In this invention, principal component analysis or linear discriminant analysis is used to reduce the dimensionality of the extracted multidimensional features. This reduces feature dimensionality and data redundancy while preserving the main discriminative information of the original data. Compared to directly using high-dimensional original features for modeling and prediction, this invention effectively simplifies the input complexity of subsequent model training and reduces computational resource consumption through dimensionality reduction. It also avoids overfitting caused by excessively high feature dimensionality, thereby improving the training efficiency and generalization ability of the fault prediction model.

[0012] In one alternative implementation, the fault prediction model is constructed in the following manner: Acquire historical operating condition data and corresponding actual fault records; Extracting multidimensional historical features from historical operating data; Machine learning algorithms were used to train multidimensional historical features to obtain multiple candidate models; Multiple candidate models were evaluated using cross-validation, and hyperparameters were tuned using a hyperparameter optimization algorithm to obtain the fault prediction model.

[0013] In this invention, historical operating condition data and corresponding actual fault records are acquired, multi-dimensional historical features are extracted, and multiple candidate models are trained using various machine learning algorithms. These models are then evaluated and tuned using cross-validation and hyperparameter optimization algorithms to construct a fault prediction model. Compared to modeling methods using a single algorithm or without systematic optimization, this invention, through the combination of multi-algorithm training, cross-validation evaluation, and hyperparameter tuning, can select a prediction model with superior performance and stronger generalization ability, effectively improving the accuracy and stability of fault prediction.

[0014] In one optional implementation, a fault prediction model is obtained, comprising: Choose the model with the best evaluation index from multiple candidate models as the fault prediction model; or, integrate and learn multiple candidate models to obtain the fault prediction model.

[0015] This invention provides two flexible model construction paths: selecting the model with the best evaluation index from multiple candidate models, or integrating and learning multiple candidate models to obtain the fault prediction model. Compared to using a single model, this invention can both select the best-performing single model through selection and integrate the advantages of multiple models to improve prediction stability and accuracy. This allows for flexible selection of the optimal strategy based on actual application needs, effectively enhancing the reliability and adaptability of fault prediction.

[0016] In one optional implementation, power equipment faults are predicted based on multidimensional features and a pre-built fault prediction model, including: The extracted multidimensional features are input into the fault prediction model; The fault prediction model calculates the predicted results of power equipment faults, which include at least one of the following: fault probability, fault type, and predicted fault occurrence time range.

[0017] In one alternative implementation, predicting power equipment failures further includes: When the probability of failure exceeds a preset threshold, it is determined that there is a risk of failure in the power equipment, and the type of failure and the predicted time range of failure occurrence are obtained.

[0018] In this invention, by inputting extracted multidimensional features into a pre-built fault prediction model for calculation, a prediction result can be directly output, containing at least one of fault probability, fault type, and predicted fault occurrence time range. Compared to traditional methods that can only determine whether equipment is abnormal or only output a single fault indicator, this invention can provide richer prediction information, enabling maintenance personnel to not only understand the risk and probability of equipment failure, but also to identify the fault type and predict the approximate occurrence time. This provides a more sufficient basis for subsequent maintenance decisions, helps to arrange targeted maintenance in advance, and avoids sudden power outages.

[0019] In a second aspect, the present invention provides a power equipment fault prediction system, applied to the power equipment fault prediction method of the first aspect or any corresponding embodiment thereof, the system comprising: Data acquisition module, used to acquire operating data of power equipment; The data analysis and prediction module is used to extract multidimensional features from operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features. Based on the multidimensional features and a pre-built fault prediction model, the module predicts power equipment faults. The fault prediction model is built by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0020] Thirdly, the present invention provides a power equipment fault prediction device, the device comprising: The data acquisition module is used to acquire operating data of power equipment; The feature extraction module is used to extract multidimensional features from the operating condition data. These multidimensional features include time-domain features, frequency-domain features, trend features, periodic features, and correlation features. The fault prediction module is used to predict power equipment faults based on multidimensional features and a pre-built fault prediction model. The fault prediction model is built by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0021] Fourthly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the power equipment fault prediction method of the first aspect or any corresponding embodiment described above.

[0022] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the power equipment fault prediction method of the first aspect or any corresponding embodiment thereof.

[0023] In a sixth aspect, the present invention provides a computer program product, including computer instructions for causing a computer to execute the power equipment fault prediction method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0024] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first process of the power equipment fault prediction method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a second process for a power equipment fault prediction method according to an embodiment of the present invention; Figure 4 This is a structural block diagram of a power equipment fault prediction device according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0028] As an optional application scenario of this invention, such as Figure 1 As shown, the power equipment fault prediction system provided by the present invention is applied to the above-mentioned power equipment fault prediction method. The system includes: Data acquisition module, used to acquire operating data of power equipment; The data analysis and prediction module is used to extract multidimensional features from operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features. Based on the multidimensional features and a pre-built fault prediction model, the module predicts power equipment faults. The fault prediction model is built by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0029] Specifically, the data acquisition module uses various types of sensors, such as temperature sensors, vibration sensors, current sensors, voltage sensors, and humidity sensors, which are installed in key parts of the power equipment to collect the operating parameters of the power equipment in real time. At the same time, it uses smart power terminal devices such as smart meters and smart switches to collect basic power system data such as power consumption data, switch status, and power factor.

[0030] In one alternative implementation, the selection of sensor type, determination of installation location, and selection and installation of intelligent power terminal equipment are all based on the type of power equipment and monitoring requirements.

[0031] The data analysis and prediction module includes: The feature engineering submodule extracts time-domain and frequency-domain features from power equipment operating parameters and power system basic data, as well as the changing trends, periodicity, and correlation features of power equipment operating parameters. It also uses dimensionality reduction techniques such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to reduce the feature dimensions.

[0032] The machine learning model training submodule uses various machine learning algorithms, such as Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN), to train on historical operating data and establish a power equipment fault prediction model. For different power equipment and fault types, the optimal algorithm model can be selected or multiple candidate models can be fused using ensemble learning to obtain the final fault prediction model.

[0033] The model evaluation and optimization submodule uses metrics such as K-fold cross-validation, accuracy, recall, F1 score, and root mean square error (RMSE) to evaluate the trained fault prediction model, and employs hyperparameter tuning techniques such as grid search, random search, and genetic algorithms to optimize the hyperparameters of the fault prediction model. Among them, K-fold cross-validation is a model validation method that randomly divides the dataset into K equal parts, and alternately uses one part as the validation set and the remaining K-1 parts as the training set for K training and evaluation cycles.

[0034] The real-time prediction submodule processes the real-time collected operating parameters of power equipment and basic data of the power system, inputs them into the optimized fault prediction model, performs real-time fault prediction of power equipment, and outputs information such as the fault probability, fault type and fault occurrence time range of power equipment.

[0035] In one optional implementation, the feature engineering submodule uses the Python programming language in conjunction with data analysis libraries such as Scikitlearn and TensorFlow to write feature extraction programs; the machine learning model training submodule selects appropriate algorithms for different power equipment and fault types, and optimizes model parameters through multiple experiments; the model evaluation and optimization submodule uses cross-validation and hyperparameter tuning techniques to evaluate and optimize the model; and the real-time prediction submodule uses technologies such as Web services or message queues to transmit the prediction results to the user management and display module.

[0036] The system also includes: The data processing module specifically includes: The data cleaning submodule is used when the data acquisition module transmits the collected power equipment operating data to the data cleaning submodule of the data processing module in real time through communication technologies such as Wireless Sensor Network (WSN), Industrial Ethernet, and 4G / 5G. The data cleaning submodule uses outlier detection techniques based on statistical methods and machine learning algorithms, such as the Interquartile Range (IQR) method and the Isolation Forest algorithm, to clean the collected raw data and remove noise, outliers, and duplicate data.

[0037] The wireless sensor network uses appropriate wireless communication protocols such as ZigBee, Bluetooth, and LoRa, and sets the sensor node locations and communication parameters reasonably; the industrial Ethernet adopts a redundant network structure.

[0038] The data storage submodule uses relational databases such as MySQL and Oracle to store structured data, and non-relational databases such as MongoDB (Humongous Database) and HBase (Hadoop Database) to store unstructured and semi-structured data such as device operation logs and image data.

[0039] The data preprocessing submodule uses methods such as Z-score standardization and min-max normalization to standardize and normalize the operating data, so that different types of data have a unified dimension and scale, which is beneficial to subsequent data analysis and model training.

[0040] In one optional implementation, the data cleaning submodule uses the Python programming language in conjunction with data processing libraries such as Pandas and NumPy to write programs to perform data cleaning; the data storage submodule selects and configures the appropriate database management system according to the data type and characteristics; and the data preprocessing submodule uses the Python programming language and related data processing libraries to perform standardization and normalization operations.

[0041] The user management and display module specifically includes: The visualization interface submodule displays equipment operating status, forecast results, and early warning information through a web interface or mobile application, using charts such as line charts, bar charts, and pie charts, daily reports, monthly reports, and annual reports, as well as maps, and provides interactive operation functions.

[0042] The visualization interface submodule also uses front-end development technologies such as HTML (HyperText Markup Language), CSS (Cascading Style Sheets), and JavaScript, combined with the Echarts and D3.js visualization libraries, to develop web interfaces, or uses Android and iOS (iPhone Operating System) mobile application development technologies to develop mobile applications.

[0043] The user permission management submodule adopts a role-based access control (RBAC) model, setting different user permissions such as administrator permissions, operation and maintenance personnel permissions, and ordinary user permissions to ensure system data security and confidentiality.

[0044] The user permission management submodule also utilizes a database management system such as MySQL to store user information and role permission information. By creating user tables, role tables, permission tables, and user-role association tables, it implements permission allocation management.

[0045] The early warning management submodule sets different levels of early warning thresholds based on fault prediction results. When the probability of equipment failure exceeds the threshold, the early warning mechanism is automatically triggered. It provides multiple early warning methods such as SMS, email, system pop-ups, and voice alarms, and supports user-defined early warning thresholds and methods.

[0046] The early warning management submodule also utilizes SMS platforms such as Alibaba Cloud SMS service and Tencent Cloud SMS service, mail servers such as Sendmail and Postfix, and system pop-up technologies to implement early warning methods.

[0047] The maintenance decision support submodule utilizes technologies such as knowledge graphs and expert systems to provide maintenance personnel with maintenance plan suggestions, including maintenance time, maintenance methods, and required spare parts, based on equipment failure prediction results and historical maintenance records. It also tracks and evaluates the equipment after maintenance and updates maintenance records and equipment health status.

[0048] The maintenance decision support submodule is also used to build a knowledge graph for power equipment maintenance and to provide maintenance solution suggestions using an expert system, while tracking and evaluating maintenance results to update the knowledge graph and expert system.

[0049] The power equipment fault prediction system provided by this invention has the following workflow: First, the data acquisition module collects power equipment operating parameters, including vibration signals, temperature signals, and current signals, in real time through multiple sensors deployed at multiple preset monitoring locations on the power equipment. At the same time, it collects basic power system data in real time through the power terminal. Secondly, the data processing module cleans, stores, and preprocesses the collected operating data. Then, the feature engineering submodule in the data analysis and prediction module extracts multidimensional features from the processed operating data. This includes extracting time-domain and frequency-domain features from vibration signals using time-frequency analysis, extracting trend and periodic features from temperature and current signals using statistical analysis, extracting correlation features between different types of signals using correlation analysis, and using principal component analysis or linear discriminant analysis for dimensionality reduction. Next, the real-time prediction submodule in the data analysis and prediction module inputs the extracted multidimensional features into the pre-built fault prediction model for calculation. This model is constructed by acquiring historical working condition data and corresponding fault records, extracting multidimensional historical features, training multiple candidate models using various machine learning algorithms, and then fine-tuning them through cross-validation and hyperparameter optimization algorithms. Finally, the user management and display module receives the prediction results output by the model, which include at least one of the following: fault probability, fault type, and predicted fault occurrence time range. It then displays these results through a visual interface and triggers an early warning mechanism based on preset thresholds to provide maintenance decision support.

[0050] According to an embodiment of the present invention, a method for predicting power equipment faults is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0051] This embodiment provides a method for predicting power equipment faults, which can be used in the aforementioned power equipment fault prediction system. Figure 2 This is a flowchart of a power equipment fault prediction method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain power equipment operating condition data.

[0052] Specifically, by deploying multiple sensors at multiple preset monitoring locations on the power equipment, the operating parameters of the power equipment, including vibration signals, temperature signals, and current signals, are collected in real time. At the same time, basic data of the power system are collected in real time through the power terminal, thereby obtaining power equipment operating condition data that includes the operating parameters of the power equipment itself and the basic data of the power system.

[0053] Step S202: Extract multidimensional features from the operating data. The multidimensional features include time domain features, frequency domain features, trend features, periodic features, and correlation features.

[0054] Among them, time-domain characteristics refer to the statistical quantities directly extracted from the changes in the amplitude of the vibration signal over time, reflecting the instantaneous amplitude distribution of the signal.

[0055] Frequency domain characteristics refer to the spectral distribution information extracted after transforming the vibration signal to the frequency domain, reflecting the energy distribution of the signal at different frequencies.

[0056] Trend characteristics: The long-term trend of temperature and current signals over time is extracted to reflect the upward or downward trend of equipment parameters.

[0057] Periodicity refers to the regular repetitive patterns extracted from temperature and current signals, reflecting the periodic changes in the operating status of equipment.

[0058] Correlation characteristics refer to the degree of inter-correlation extracted from different types of signals, reflecting the intrinsic relationship between signals such as vibration, temperature, and current.

[0059] Specifically, in the data analysis and prediction module, the feature engineering submodule extracts multi-dimensional features from the working condition data processed by the data processing module: first, it uses time-frequency analysis to extract time-domain and frequency-domain features that reflect instantaneous changes and frequency distributions from the vibration signal; then, it uses statistical analysis to extract trend and periodic features that describe the evolution of the temperature and current signals; finally, it uses correlation analysis to extract correlation features from multiple types of signals.

[0060] Step S203: Based on multidimensional features and a pre-built fault prediction model, power equipment faults are predicted. The fault prediction model is constructed by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0061] Specifically, in the data analysis and prediction module, the real-time prediction submodule inputs the multi-dimensional features extracted by the feature engineering submodule into the pre-built fault prediction model for calculation, thereby outputting the prediction results of power equipment faults.

[0062] The fault prediction model is constructed as follows: First, historical operating condition data and corresponding actual fault records are acquired, and multi-dimensional historical features are extracted from them by the feature engineering submodule; then, multiple machine learning algorithms are used to train the multi-dimensional historical features to obtain multiple candidate models; finally, cross-validation is used to evaluate each candidate model, and hyperparameter optimization algorithm is used to fine-tune the hyperparameters to obtain the fault prediction model.

[0063] The power equipment fault prediction method provided in this embodiment acquires operating data and extracts multi-dimensional features such as time domain, frequency domain, trend, period, and correlation. It integrates multiple machine learning algorithms for ensemble learning training and combines cross-validation and hyperparameter optimization to construct a prediction model, achieving early prediction of power equipment faults. Compared to traditional periodic maintenance and reactive repair methods, this invention effectively replaces these methods, avoiding over-maintenance, under-maintenance, and power outages caused by the inability to predict faults. This improves the reliability of power supply, reduces operation and maintenance costs, and increases operation and maintenance efficiency.

[0064] This embodiment provides a method for predicting power equipment faults, which can be used in the aforementioned power equipment fault prediction system. Figure 3 This is a flowchart of a power equipment fault prediction method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Obtain power equipment operating condition data.

[0065] Specifically, step S301 includes: Step a: Real-time acquisition of power equipment operating parameters through multiple sensors deployed at multiple preset monitoring locations on the power equipment, and real-time acquisition of basic power system data through the power terminal; the power equipment operating parameters include vibration signals, temperature signals and current signals.

[0066] Specifically, by deploying multiple sensors at multiple preset monitoring locations on the power equipment, operating parameters of the power equipment, including vibration signals, temperature signals, and current signals, are collected in real time. At the same time, basic power system data, including power consumption data, switch status, and power factor, are collected in real time through the power terminal, thereby obtaining multi-condition data covering the equipment's own status and the system's operating environment.

[0067] Step S302: Extract multidimensional features from the operating data. The multidimensional features include time domain features, frequency domain features, trend features, periodic features, and correlation features.

[0068] Specifically, step S302 includes: Step S3021: Extract time-domain and frequency-domain features from the vibration signal using time-frequency analysis.

[0069] As an example, short-time Fourier transform or wavelet transform can be used to perform time-frequency analysis on vibration signals. The time-domain features include statistical quantities such as the mean, variance, peak value, and root mean square value of the vibration signal, reflecting the instantaneous amplitude distribution of the signal on the time axis. The frequency-domain features include parameters such as spectral energy, dominant frequency component, and frequency band energy distribution obtained through Fourier transform, describing the energy distribution characteristics of the signal at different frequencies. Through time-frequency analysis, fault-related information of the vibration signal can be captured simultaneously in both the time and frequency domains.

[0070] Step S3022: Statistical analysis is used to extract trend features and periodic features from the temperature and current signals.

[0071] As an example, a sliding window statistical method can be used to statistically analyze temperature and current signals. Trend characteristics include the slope of signal changes obtained through linear regression fitting, reflecting the long-term upward or downward trend of equipment temperature or current over time; periodic characteristics include parameters such as the period length and fluctuation amplitude of the signal extracted through autocorrelation analysis or periodogram methods, describing the regular repetitive patterns exhibited in the equipment's operating state. Through statistical analysis, the evolutionary patterns and periodic characteristics of equipment parameters over time can be effectively revealed.

[0072] Step S3023: Use correlation analysis to extract correlation features between different types of signals from the operating data.

[0073] As an example, Pearson correlation coefficient or cross-correlation analysis can be used to perform correlation analysis on different types of signals in operating condition data. Specifically, the correlation coefficients between vibration and temperature signals, vibration and current signals, and temperature and current signals are calculated to extract correlation features that reflect the degree of linear correlation or time delay relationship between different types of signals. Through correlation analysis, the inherent correlation between sensor signals such as vibration, temperature, and current can be revealed, providing a basis for multi-source information fusion in fault diagnosis.

[0074] Step S3024: Principal component analysis or linear discriminant analysis is used to reduce the dimensionality of the extracted multidimensional features.

[0075] As an example, principal component analysis (PCA) or linear discriminant analysis (LDA) can be used to reduce the dimensionality of the extracted multidimensional features. PCA projects the original high-dimensional features onto several orthogonal directions with the largest variance through linear transformation, preserving the main variation information of the data. LDA, on the other hand, seeks projection directions that maximize the ratio of inter-class divergence to intra-class divergence, maximizing the separability between different fault categories while reducing dimensionality. Dimensionality reduction removes redundant information between features, reduces feature dimensionality, thereby lowering the computational complexity of subsequent model training and mitigating overfitting problems that may result from excessively high dimensionality.

[0076] Step S303: Based on multidimensional features and a pre-built fault prediction model, power equipment faults are predicted. The fault prediction model is constructed by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0077] In one optional implementation, the fault prediction model is constructed as follows: acquiring historical operating condition data and corresponding actual fault records; extracting multi-dimensional historical features from the historical operating condition data; training the multi-dimensional historical features using machine learning algorithms to obtain multiple candidate models; evaluating the multiple candidate models using cross-validation; and then fine-tuning the hyperparameters using a hyperparameter optimization algorithm to obtain the fault prediction model.

[0078] The process of obtaining a fault prediction model includes: selecting the model with the best evaluation index from multiple candidate models as the fault prediction model; or, integrating and learning multiple candidate models to obtain a fault prediction model.

[0079] Specifically, the first step is to obtain historical operating condition data and corresponding actual fault records. The historical operating condition data includes power equipment operation data and power system basic data, while the actual fault records include tag information such as fault type and fault occurrence time. Then, multi-dimensional historical features are extracted from historical operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features. Next, various machine learning algorithms such as support vector machine, random forest, gradient boosting decision tree, long short-term memory network, and convolutional neural network are used to train the multi-dimensional historical features to obtain multiple candidate models. Then, K-fold cross-validation is used to evaluate each candidate model. That is, the historical data is divided into K parts, and one part is used as the validation set and the remaining K-1 parts are used as the training set for K training and validation cycles. The average performance index is calculated. At the same time, hyperparameter optimization algorithms (such as grid search, random search or genetic algorithm) are used to automatically search and tune the hyperparameters of the candidate models to find the optimal parameter combination. Finally, based on the evaluation results of cross-validation, the candidate model with the best performance is selected as the fault prediction model, or multiple candidate models are integrated and fused to obtain the final fault prediction model.

[0080] Specifically, step S303 includes: Step S3031: Input the extracted multidimensional features into the fault prediction model; the fault prediction model calculates and obtains the prediction result of the power equipment fault, which includes at least one of the fault probability, fault type and predicted fault occurrence time range.

[0081] Specifically, the multidimensional features extracted by the feature engineering submodule are used as input data and transmitted to the pre-built fault prediction model in the real-time prediction submodule via a data interface. The fault prediction model performs forward calculations based on the input multidimensional features and infers the current health status of the power equipment based on the parameter mapping relationships learned during the training phase. After the model calculation is completed, the prediction result is output. This prediction result includes at least one of the following: the probability value of equipment failure (e.g., a value between 0 and 1), the predicted fault type (e.g., overheating fault, mechanical wear fault, electrical fault, etc.), and the predicted fault occurrence time range (e.g., within the next 24 hours, within the next 3-5 days, etc.), providing maintenance personnel with a quantitative basis for fault risk assessment.

[0082] Step S3032: When the failure probability exceeds a preset threshold, it is determined that the power equipment has a failure risk, and the failure type and the predicted failure occurrence time range are obtained.

[0083] Specifically, the fault probability value output in step S3031 is compared with the system's preset warning threshold. The system has multiple levels of thresholds set according to different equipment types and fault severity (e.g., the yellow warning threshold is 0.5 (mild risk threshold), the orange warning threshold is 0.7 (medium risk threshold), and the red warning threshold is 0.9 (high risk threshold)).

[0084] When the failure probability value exceeds a preset threshold, the system automatically determines that the power equipment currently has a failure risk and further extracts the failure type and predicted failure occurrence time range from the model output. Simultaneously, the system triggers corresponding early warning mechanisms based on the threshold level exceeded (e.g., sending an email alert when exceeding the yellow threshold, and triggering both SMS and voice alarms when exceeding the red threshold). The determined failure risk information (failure type, predicted time range, and risk level) is then pushed to the user management and display module for visual presentation, enabling maintenance personnel to take timely and targeted repair measures.

[0085] The power equipment fault prediction method provided in this embodiment can accurately predict faults, arrange maintenance in advance, avoid power outages, and improve the reliability of the power system. It adopts on-demand maintenance, reduces unnecessary maintenance and spare parts delays, lowers operation and maintenance costs, and improves operation and maintenance efficiency through real-time monitoring, rapid prediction, and maintenance decision support. It achieves data-driven decision-making by analyzing massive amounts of data, customizes maintenance plans, and combines the modular design of the system to facilitate the integration of new power equipment functions, making it easy to upgrade and optimize, thus enhancing the competitiveness of the system and method.

[0086] As one or more specific application embodiments of the present invention, the power equipment fault prediction method provided by the present invention will be further described in detail in conjunction with a power equipment fault prediction system, as follows: The power equipment fault prediction system's various modules are activated. The data acquisition module begins operation, with various sensors closely monitoring key components of the power equipment. The collected equipment operating parameters and basic power system data gathered by intelligent power terminal devices are rapidly transmitted to the data processing module via adapted communication technology. The data processing module then starts. The data cleaning submodule uses a Python program, leveraging Pandas and NumPy libraries, to clean the raw data. The data storage submodule stores the data into the appropriate database based on its characteristics. The data preprocessing submodule uses Python and related libraries to standardize and normalize the data. The processed, valid data then enters the data analysis and prediction module. The feature engineering submodule, utilizing Python combined with Scikit-... The system utilizes the learn and TensorFlow libraries to extract key features and reduce dimensionality. The machine learning model training submodule selects appropriate algorithms to train models based on equipment and fault types. The model evaluation and optimization submodule uses techniques such as cross-validation to evaluate and fine-tune the model. The real-time prediction submodule inputs the processed data into the optimized model and outputs equipment fault-related information. The prediction results are transmitted to the user management and display module. The visualization interface submodule displays information through a web interface or mobile application in the form of intuitive charts and reports for easy user viewing. The user permission management submodule ensures secure access to the system for users with different permissions. The early warning management submodule issues early warnings based on preset thresholds, which can be customized by users. The maintenance decision support submodule uses knowledge graphs and expert systems to provide maintenance personnel with solution suggestions and tracks maintenance effectiveness to update the system, continuously ensuring the stable operation of power equipment and fault prediction.

[0087] This embodiment also provides a power equipment fault prediction device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0088] This embodiment provides a power equipment fault prediction device, such as... Figure 4 As shown, it includes: The data acquisition module 401 is used to acquire power equipment operating condition data.

[0089] The feature extraction module 402 is used to extract multidimensional features from the operating condition data. The multidimensional features include time domain features, frequency domain features, trend features, periodic features, and correlation features.

[0090] The fault prediction module 403 is used to predict power equipment faults based on multidimensional features and a pre-built fault prediction model. The fault prediction model is built by integrating and training multiple machine learning algorithms and then fine-tuning it through cross-validation and hyperparameter optimization algorithms.

[0091] In some optional implementations, the operating condition data includes power equipment operating parameters and power system basic data; the data acquisition module 401 includes: The data acquisition unit is used to collect power equipment operating parameters in real time through multiple sensors deployed at multiple preset monitoring locations on the power equipment, and to collect basic power system data in real time through the power terminal; the power equipment operating parameters include vibration signals, temperature signals and current signals.

[0092] In some alternative implementations, the feature extraction module 402 includes: The time-frequency feature extraction unit is used to extract time-domain and frequency-domain features from vibration signals using time-frequency analysis.

[0093] The trend and periodicity feature extraction unit is used to extract trend and periodicity features from temperature and current signals using statistical analysis.

[0094] The correlation feature extraction unit is used to extract correlation features between different types of signals from the operating data using correlation analysis. The dimensionality reduction unit is used to perform dimensionality reduction processing on the extracted multidimensional features using principal component analysis or linear discriminant analysis.

[0095] In some optional implementations, the fault prediction model is constructed as follows: acquiring historical operating condition data and corresponding actual fault records; extracting multi-dimensional historical features from the historical operating condition data; training the multi-dimensional historical features using machine learning algorithms to obtain multiple candidate models; evaluating the multiple candidate models using cross-validation; and then fine-tuning the hyperparameters using a hyperparameter optimization algorithm to obtain the fault prediction model.

[0096] The process of obtaining a fault prediction model includes: selecting the model with the best evaluation index from multiple candidate models as the fault prediction model; or, integrating and learning multiple candidate models to obtain a fault prediction model.

[0097] In some alternative implementations, the fault prediction module 403 includes: The fault prediction unit is used to input the extracted multidimensional features into the fault prediction model; the fault prediction model calculates and obtains the prediction result of the power equipment fault, which includes at least one of the fault probability, fault type and predicted fault occurrence time range.

[0098] In some optional implementations, the fault prediction module 403 further includes: The risk assessment unit is used to determine the risk of failure in power equipment when the failure probability exceeds a preset threshold, and to obtain the failure type and the predicted time range of failure occurrence.

[0099] The power equipment fault prediction device provided in this embodiment of the invention can execute the power equipment fault prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0100] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0101] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0102] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0103] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the power equipment fault prediction method of the embodiments of the present invention.

[0104] Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0105] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the power equipment fault prediction method shown in the above embodiments is implemented.

[0106] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0107] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting faults in power equipment, characterized in that, The method includes: Acquire operating data of power equipment; Multidimensional features are extracted from the operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features. Based on the multidimensional features and the pre-built fault prediction model, power equipment faults are predicted; the fault prediction model is constructed based on the integrated learning and training of multiple machine learning algorithms, and is then optimized through cross-validation and hyperparameter optimization algorithms.

2. The method according to claim 1, characterized in that, The operating data includes power equipment operating parameters and power system basic data; The acquisition of power equipment operating condition data includes: Multiple sensors deployed at various preset monitoring locations on the power equipment are used to collect the operating parameters of the power equipment in real time, and the power system's basic data is collected in real time through the power terminal; the operating parameters of the power equipment include vibration signals, temperature signals and current signals.

3. The method according to claim 2, characterized in that, Extracting multidimensional features from the operating condition data includes: Time-frequency analysis is used to extract time-domain and frequency-domain features from the vibration signal; Statistical analysis is used to extract trend and periodic features from the temperature and current signals; Correlation analysis is used to extract correlation features between different types of signals from the operating data.

4. The method according to claim 3, characterized in that, Extracting multidimensional features from the operating condition data also includes: Principal component analysis or linear discriminant analysis are used to reduce the dimensionality of the extracted multidimensional features.

5. The method according to claim 1, characterized in that, The fault prediction model is constructed in the following manner: Acquire historical operating condition data and corresponding actual fault records; Extract multidimensional historical features from the historical operating data; Machine learning algorithms are used to train the multidimensional historical features to obtain multiple candidate models; The multiple candidate models are evaluated using cross-validation, and the hyperparameters are tuned using a hyperparameter optimization algorithm to obtain the fault prediction model.

6. The method according to claim 5, characterized in that, The fault prediction model is obtained by: Select the model with the optimal evaluation index from the multiple candidate models as the fault prediction model; or... The multiple candidate models are integrated and fused to obtain the fault prediction model.

7. The method according to claim 1, characterized in that, The method of predicting power equipment faults based on the multidimensional features and the pre-built fault prediction model includes: The extracted multidimensional features are input into the fault prediction model; The fault prediction model calculates the predicted fault results of the power equipment, which include at least one of the following: fault probability, fault type, and predicted fault occurrence time range.

8. The method according to claim 7, characterized in that, Predicting power equipment failures also includes: When the failure probability exceeds a preset threshold, it is determined that the power equipment has a failure risk, and the failure type and the predicted failure occurrence time range are obtained.

9. A power equipment fault prediction system, characterized in that, The system applied to the power equipment fault prediction method according to any one of claims 1 to 8 includes: Data acquisition module, used to acquire operating data of power equipment; The data analysis and prediction module is used to extract multi-dimensional features from the operating data, including time-domain features, frequency-domain features, trend features, periodic features, and correlation features; based on the multi-dimensional features and a pre-built fault prediction model, the module predicts power equipment faults; the fault prediction model is constructed based on the integrated learning and training of multiple machine learning algorithms, and is then optimized through cross-validation and hyperparameter optimization algorithms.

10. A power equipment fault prediction device, characterized in that, The device includes: The data acquisition module is used to acquire operating data of power equipment; The feature extraction module is used to extract multidimensional features from the operating condition data. The multidimensional features include time-domain features, frequency-domain features, trend features, periodic features, and correlation features. The fault prediction module is used to predict power equipment faults based on the multidimensional features and the pre-built fault prediction model. The fault prediction model is constructed based on the integrated learning and training of multiple machine learning algorithms and after cross-validation and hyperparameter optimization algorithm tuning.

Images