Power transformation equipment fault prediction system and method based on multi-modal data

Abstract

The invention discloses a power transformation equipment fault prediction system and method based on multi-modal data, and the system comprises a data collection module which is used for obtaining the multi-modal data of power transformation equipment at the current moment based on a plurality of sensors; the data preprocessing module is used for performing preprocessing and standardization operation based on the multi-modal data at the current moment to obtain standardized data at the current moment; the feature extraction module is used for performing feature extraction based on the standardized data at the current moment to obtain a feature vector at the current moment; the real-time detection and prediction module is used for performing fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result; according to the invention, the obtained multi-modal data can be used for monitoring the power transformation equipment at different angles and dimensions, wider fault modes and abnormal conditions are covered, and the operation and maintenance efficiency is improved; according to the method, high-dimensional data fusion is carried out, so that the accuracy and reliability of fault recognition can be remarkably improved, and the situations of false detection and missing detection are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of equipment failure monitoring, and in particular to a system and method for predicting failure of substation equipment based on multimodal data. Background Art

[0002] With the rapid development of modern power systems, power equipment, as the cornerstone of the power grid, its operating status is directly related to the stability and safety of the entire power system. However, in actual operation, power equipment is often affected by various factors such as the environment, load changes, and material aging, resulting in frequent failures. These failures will not only cause huge economic losses, but may also cause serious consequences such as large-scale power outages, affecting the normal order of social life. Therefore, how to effectively monitor and prevent power equipment failures and ensure the safe and stable operation of the power system has become an important issue that needs to be urgently solved in the power industry.

[0003] The emergence of intelligent inspection and fault prediction technology provides a new solution to this problem. By using advanced data acquisition and analysis technology, combined with intelligent algorithms, real-time monitoring of the status of power equipment and fault warning can be achieved, so that measures can be taken in advance before the fault occurs to avoid or reduce the impact of the fault on the power system. The advantage of this technology is that it can significantly improve the inspection efficiency, reduce the labor intensity of manual inspection, and improve the accuracy and timeliness of fault prediction, providing a strong guarantee for the safe operation of the power system. However, although certain progress has been made in the field of operation and maintenance and fault diagnosis of power transmission and transformation equipment, most methods are still limited to single-mode data processing, especially relying on binocular cameras and machine vision technology, and only detecting faults through a single feature of images. This single-dimensional information acquisition method significantly limits the comprehensiveness and accuracy of fault identification. Due to the lack of multi-dimensional reference conditions, it is often difficult to fully reflect the true state of equipment failures, which increases the risk of false detection and missed detection, affecting the operation and maintenance efficiency and the reliability of the power grid.

[0004] Most methods are still limited to single-mode data processing, especially those that rely on binocular cameras and machine vision technology, and only use images as a single feature for fault detection. This single-dimensional information acquisition method significantly limits the comprehensiveness and accuracy of fault identification. Due to the lack of multi-dimensional reference conditions, it is often difficult to fully reflect the true state of equipment failures, which in turn increases the risk of false detection and missed detection, affecting operation and maintenance efficiency and grid reliability. Summary of the invention

[0005] In order to solve the problem that the existing technology is difficult to fully reflect the true state of equipment failure due to the lack of multi-dimensional data, this application proposes a substation equipment failure prediction system based on multi-modal data, including:

[0006] A data collection module is used to obtain multimodal data of the substation equipment at the current moment based on multiple sensors;

[0007] A data preprocessing module, used to perform preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment;

[0008] A feature extraction module, used to extract features based on the standardized data at the current moment to obtain a feature vector at the current moment;

[0009] A real-time detection and prediction module, used to perform fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result;

[0010] The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

[0011] Preferably, the system further comprises a fault prediction model training module, wherein the fault prediction model training module is specifically used for:

[0012] The fusion feature calculation formula is used to calculate the feature vector of the historical moment and the standardized data of the historical moment to obtain the fusion feature of the historical moment;

[0013] The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state.

[0014] Preferably, the relationship between the fusion features and the equipment status in the fault prediction model training module is as follows:

[0015] f(P α )=sign(w*P α +b),P α ∈{P1,P2,…,P m}

[0016] Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

[0017] Preferably, the data preprocessing module is specifically used for:

[0018] Performing cleaning, formatting, and normalization operations based on the multimodal data at the current moment to obtain preprocessed data at the current moment;

[0019] Calculating the mean and standard deviation of the preprocessed data at the current moment to obtain the mean and standard deviation of the preprocessed data at the current moment;

[0020] A standardization operation is performed based on the mean and standard deviation of the preprocessed data at the current moment to obtain standardized data at the current moment that are unified into the same magnitude.

[0021] Preferably, the feature extraction module is specifically used for:

[0022] The information features in the standardized data at the current moment are extracted by using target detection and image classification technology, and the information features are converted into feature vectors at the current moment.

[0023] Preferably, the real-time detection and prediction module is specifically used for:

[0024] Calculate the fusion feature at the current moment based on the feature vector at the current moment and the standardized data at the current moment using a fusion feature calculation formula to obtain the fusion feature at the current moment;

[0025] The fault prediction model is input based on the fusion features at the current moment to perform prediction, and the device state at the next moment after the current moment is output as a fault prediction result.

[0026] Based on the same application concept, this application also proposes a substation equipment fault prediction method based on multimodal data, including:

[0027] Acquire multimodal data of substation equipment at the current moment based on multiple sensors;

[0028] Performing preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment;

[0029] Perform feature extraction based on the standardized data at the current moment to obtain a feature vector at the current moment;

[0030] Based on the feature vector at the current moment and in combination with a pre-trained fault prediction model, a fault prediction is performed to obtain a fault prediction result;

[0031] The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

[0032] Preferably, the training process of the fault prediction model includes: calculating based on the feature vector of the historical moment and the standardized data of the historical moment using a fusion feature calculation formula to obtain the fusion feature of the historical moment;

[0033] The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state.

[0034] Preferably, the relationship between the fusion feature and the device state is as follows: f(P α )=sign(w*

[0035] P α +b),P α ∈{P1,P2,…,P m}

[0036] Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

[0037] Preferably, the preprocessing and standardization operations are performed based on the multimodal data at the current moment to obtain the standardized data at the current moment, including:

[0038] Performing cleaning, formatting, and normalization operations based on the multimodal data at the current moment to obtain preprocessed data at the current moment;

[0039] Calculating the mean and standard deviation of the preprocessed data at the current moment to obtain the mean and standard deviation of the preprocessed data at the current moment;

[0040] A standardization operation is performed based on the mean and standard deviation of the preprocessed data at the current moment to obtain standardized data at the current moment that are unified into the same magnitude.

[0041] Preferably, the extracting features based on the standardized data at the current moment to obtain the feature vector at the current moment includes:

[0042] The information features in the standardized data at the current moment are extracted by using target detection and image classification technology, and the information features are converted into feature vectors at the current moment.

[0043] Preferably, the fault prediction of the substation equipment is performed based on the feature vector at the current moment in combination with a pre-built fault prediction model to obtain a fault prediction result, including:

[0044] Calculate the fusion feature at the current moment based on the feature vector at the current moment and the standardized data at the current moment using a fusion feature calculation formula to obtain the fusion feature at the current moment;

[0045] The fault prediction model is input based on the fusion features at the current moment to perform prediction, and the device state at the next moment after the current moment is output as a fault prediction result.

[0046] Compared with the prior art, the beneficial effects of this application are:

[0047] A substation fault prediction system and method based on multimodal data, comprising: a data collection module, used to obtain multimodal data of the substation at the current moment based on multiple sensors; a data preprocessing module, used to perform preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment; a feature extraction module, used to perform feature extraction based on the standardized data at the current moment to obtain a feature vector at the current moment; a real-time detection and prediction module, used to perform fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result; wherein the fault prediction model is obtained by training a support vector machine model using fusion features calculated based on feature vectors at historical moments and standardized data as input data and device states corresponding to the next moment of the historical moment as output data; the multimodal data obtained in the present application can monitor substation equipment at different angles and dimensions, cover a wider range of fault modes and abnormal conditions, and improve operation and maintenance efficiency; the high-dimensional data fusion performed in the present application can significantly improve the accuracy and reliability of fault identification and reduce false detection and missed detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a structural diagram of a substation fault prediction system based on multimodal data of the present application;

[0049] Figure 2 A flow chart of a method for predicting faults of substation equipment based on multimodal data of the present application;

[0050] Figure 3 The flow chart of the substation equipment for multimodal data analysis of this application;

[0051] Figure 4 This is a flow chart of image noise removal based on wavelet transform in this application;

[0052] Figure 5This is a flowchart of equipment operation status monitoring and fault classification based on the YOLO algorithm in this application. DETAILED DESCRIPTION

[0053] This application proposes a substation equipment fault prediction system and method based on multimodal data, aiming to build a more comprehensive and accurate equipment status assessment framework by integrating diversified information from different sensors and data sources, including but not limited to multimodal data such as images, sounds, and temperatures. Through advanced data analysis and machine learning algorithms, the system can deeply explore the inherent connections and complementary advantages between each modal data, and realize multi-dimensional, in-depth monitoring and diagnosis of equipment health. By integrating precise detection, efficient classification, feature matching and data fusion technology, the complex high-dimensional data in the power system can be effectively processed, and a multi-scale time analysis model can be constructed. The model combines efficient data preprocessing with deep learning to achieve real-time monitoring and accurate fault prediction of substation equipment, providing technical support for the safe and stable operation of the power grid.

[0054] In order to better understand the present application, the contents of the present application are further described below in conjunction with the accompanying drawings and embodiments of the specification.

[0055] Embodiment 1:

[0056] A substation equipment fault prediction system based on multimodal data, the structure is as follows Figure 1 As shown, including:

[0057] A data collection module is used to obtain multimodal data of the substation equipment at the current moment based on multiple sensors;

[0058] A data preprocessing module, used to perform preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment;

[0059] A feature extraction module, used to extract features based on the standardized data at the current moment to obtain a feature vector at the current moment;

[0060] A real-time detection and prediction module, used to perform fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result;

[0061] The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

[0062] It also includes a fault prediction model training module, which is specifically used to:

[0063] The fusion feature calculation formula is used to calculate the feature vector of the historical moment and the standardized data of the historical moment to obtain the fusion feature of the historical moment;

[0064] The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state.

[0065] The data and features extracted in the data preprocessing module and feature extraction module are subjected to principal component analysis (PCA) to efficiently map multidimensional features into a few principal components to achieve data dimensionality reduction. These principal components, as independent linear combinations of the original features, are not only mutually orthogonal, but also greatly retain the original information, simplifying the data structure and accelerating subsequent analysis, retaining key clues and an efficient foundation for subsequent in-depth exploration.

[0066] Support vector machine (SVM) is selected as the prediction model, and the fusion features obtained after efficient dimensionality reduction are input into the model to capture the complex relationship between feature data and the actual state of substation equipment.

[0067] The relationship between the fusion features and the equipment status in the fault prediction model training module is as follows:

[0068] f(P α )=sign(w*P α +b),P α ∈{P1,P2,…,P m}

[0069] Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

[0070] Data collection module, specifically used for:

[0071] Gather diverse, widely distributed and high-dimensional data captured by various sensors from the power system, including temperature sensors, binocular cameras, current and voltage sensors, etc. This module strictly guarantees the integrity, accuracy and timeliness of the data, laying a solid data foundation for subsequent data processing and in-depth analysis. The data preprocessing module is specifically used for:

[0072] Performing cleaning, formatting, and normalization operations based on the multimodal data at the current moment to obtain preprocessed data at the current moment;

[0073] Calculating the mean and standard deviation of the preprocessed data at the current moment to obtain the mean and standard deviation of the preprocessed data at the current moment;

[0074] A standardization operation is performed based on the mean and standard deviation of the preprocessed data at the current moment to obtain standardized data at the current moment that are unified into the same magnitude.

[0075] Responsible for cleaning, formatting, and normalizing the raw data collected by the data collection module. This step aims to eliminate noise, outliers, missing values, and other problems in the data and improve data quality; at the same time, data from different sources and in different formats are uniformly converted into a format suitable for subsequent processing and analysis. The preprocessed data is more standardized and consistent, laying a good foundation for subsequent feature extraction and model training.

[0076] Feature extraction module, specifically used for:

[0077] The information features in the standardized data at the current moment are extracted by using target detection and image classification technology, and the information features are converted into feature vectors at the current moment.

[0078] Responsible for extracting valuable information features for fault prediction from the data preprocessed by the data preprocessing module using advanced technologies such as target detection and image classification. These features may include equipment operating status parameters, abnormal areas in the image, etc. Through feature extraction, the module converts raw data into more representative feature vectors.

[0079] Real-time detection and prediction module, specifically used for:

[0080] Calculate the fusion feature at the current moment based on the feature vector at the current moment and the standardized data at the current moment using a fusion feature calculation formula to obtain the fusion feature at the current moment;

[0081] The fault prediction model is input based on the fusion features at the current moment to perform prediction, and the device state at the next moment after the current moment is output as a fault prediction result.

[0082] Responsible for deploying the trained fault prediction model to the actual system, and performing real-time monitoring and fault prediction of the status of the substation equipment. This module receives new data from the data acquisition module, and inputs it into the model after preprocessing and feature extraction. The model outputs the fault prediction results based on the input data. At the same time, the module also has a real-time feedback mechanism, which can adjust the monitoring strategy or trigger the early warning mechanism in time according to the prediction results.

[0083] It also includes an output and feedback module for outputting the fault prediction results to operation and maintenance personnel or related systems in an appropriate form.

[0084] Multi-dimensional information fusion: The system integrates multiple sensors and data sources to obtain all-round information on equipment operation. Through comprehensive analysis and fusion of multi-modal data, the system can more comprehensively understand the equipment status, significantly improve the accuracy and reliability of fault identification, and reduce false detection and missed detection.

[0085] Real-time status monitoring: The system can monitor the operating status of the equipment in real time, detect potential faults in time and issue early warnings. This helps operation and maintenance personnel take measures in advance to prevent the expansion of faults and ensure the safe and stable operation of the power grid.

[0086] Flexibility and scalability: The system supports the access of multiple sensors and data sources, and has high flexibility and scalability. With the continuous development of technology and the emergence of new sensors, the system can easily integrate new data sources and analysis methods, and continuously improve its performance and application scope.

[0087] Embodiment 2:

[0088] A method for predicting faults of substation equipment based on multimodal data. The specific process is as follows: Figure 2 As shown, including:

[0089] Step 1, obtaining multimodal data of the substation equipment at the current moment based on multiple sensors;

[0090] Step 2, performing preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment;

[0091] Step 3, performing feature extraction based on the standardized data at the current moment to obtain a feature vector at the current moment;

[0092] Step 4, performing fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result;

[0093] The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

[0094] Combine the following Figure 3 The flowchart shown further illustrates this embodiment:

[0095] Before step 1, the training process of the fault prediction model is also included:

[0096] Based on the feature vector of the historical moment and the standardized data of the historical moment, the fusion feature calculation formula is used to calculate and obtain the fusion features of the historical moment, including:

[0097] Through principal component analysis PCA, we can directly extract the preprocessed data Y:Y1,Y2,…,Y n Efficient dimensionality reduction simplifies model complexity and accelerates the training process.

[0098] The fused feature P can be obtained from the original feature Y by the following formula:

[0099] P=YV

[0100] Where P is the fusion feature P1, P2, ..., P k , k≤n, k is the number of new features after fusion, P1 is the first fusion feature, P2 is the second fusion feature, P k is the kth fusion feature; Y is the original feature after standardization Y1, Y2,…, Y n , n is the number of original features, Y1 is the first original feature, Y2 is the second original feature, Y n is the nth original feature; V is the v×k eigenvector matrix, and its columns are the eigenvectors corresponding to the eigenvalue decomposition of the covariance matrix of Y.

[0101] The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state, including: using the fusion feature P obtained after dimensionality reduction as the input of the support vector machine model SVM, and whether the device state is faulty as the output, to construct a model reflecting the complex mapping relationship between the feature and the device state, that is, to obtain w and b, as shown below:

[0102] f(P α )=sign(w*P α +b),P α ∈{P1,P2,…,P m}

[0103] Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

[0104] The dataset Q contains m samples, each sample has k features and a label 0 for no fault and 1 for fault. Its structure is Q = {(P1, τ1), (P2, τ2), …, (P m ,τ m )};

[0105] τ={τ1,τ2,…,τ m} is a label set containing m samples, each τ i is a scalar value of {-1,1}. -1 means no fault, 1 means fault.

[0106] sign() is the sign function. If (w*P α +b) is greater than 0, it returns 1; if less than 0, it returns -1.

[0107] In step 1, the multimodal data of the substation equipment at the current moment is obtained based on multiple sensors, specifically including:

[0108] Collect multi-source feature data: Use a combination of multiple sensors such as temperature sensors, ordinary cameras, and current and voltage sensors to monitor key components in the power system in real time. The temperature sensor is responsible for the temperature of the equipment to be tested, while the ordinary camera is used to capture images of key components such as switches and display lights in the power system. At the same time, the current and voltage sensors monitor the circuit status in real time.

[0109] In step 2, preprocessing and standardization operations are performed based on the multimodal data at the current moment to obtain standardized data at the current moment, specifically including:

[0110] The process of removing image noise based on wavelet transform is as follows Figure 4 As shown, the method includes inputting an original image, performing N-level wavelet decomposition on the original image signal to obtain high-frequency components and low-frequency components, performing threshold processing on the high-frequency components, and performing wavelet reconstruction with the processed components to obtain a denoised image.

[0111] The data from different sensors are unified to the same level based on the standardized Z-Score, including:

[0112] For each sensor data sequence X i :x i1 ,x i2 ,…,x im Unified into

[0113] In the formula, X i is the data sequence of the ith sensor, x i1 is the data of the first sample at the i-th sensor, x i2 is the data of the second sample at the i-th sensor, x im is the data of the mth sample at the i-th sensor; m is the number of samples; y ij is the data of the jth sample after standardization on the i-th sensor; x ij is the original data of the jth sample on the i-th sensor; For the data sequence x i1 ,x i2 ,…,x im The average value of s i For the data sequence x i1 ,x i2 ,…,x im The standard deviation of Y i is the standardized data sequence Y of the i-th sensor i1 ,Y i2 ,…,Y im .

[0114] The data is standardized based on the mean and standard deviation of the original data, so that data of different magnitudes can be unified into the same magnitude, ensuring the comparability of the data.

[0115] The 3σ principle removes anomalies, including:

[0116] The data sequence X of each sensor i After the above steps, the obtained data sequence Y i It obeys the standard normal distribution with a mean of 0 and a standard deviation of 1. Outliers are removed according to the following formula:

[0117]

[0118] In the formula, Z i is the abnormal judgment data of the i-th sensor data; y ij is the data of the j-th sample after standardization on the i-th sensor; For the data sequence x i1 ,x i2 ,…,x im The average value of i For the data sequence x i1 ,x i2 ,…,x im If Z i >3 ij Those judged as outliers are removed.

[0119] In step 3, feature extraction is performed based on the standardized data at the current moment to obtain the feature vector at the current moment, which specifically includes:

[0120] The process of equipment operation status monitoring and fault classification based on YOLO algorithm is as follows: Figure 5As shown, it includes: data collection, data labeling, selecting the YOLO model for training and deployment, importing real-time photos into the trained model for device and status recognition, and outputting whether a fault occurs based on the recognized status.

[0121] In step 4, fault prediction is performed based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result, which specifically includes:

[0122] After real-time sensor data is processed, it is input into the constructed model to instantly predict the device status, as shown below:

[0123] f(P β )=sign(w*P β +b)

[0124] Where P β is the feature vector returned by the sensor; f(P β ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term.

[0125] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0126] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0127] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0129] The above are merely embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application are included in the scope of the claims of the present application to be approved.

Claims

1. A substation equipment fault prediction system based on multimodal data, characterized in that: include: A data collection module is used to obtain multimodal data of the substation equipment at the current moment based on multiple sensors; A data preprocessing module, used to perform preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment; A feature extraction module, used to extract features based on the standardized data at the current moment to obtain a feature vector at the current moment; A real-time detection and prediction module, used to perform fault prediction based on the feature vector at the current moment in combination with a pre-trained fault prediction model to obtain a fault prediction result; The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

2. The system according to claim 1, characterized in that The system further includes a fault prediction model training module, wherein the fault prediction model training module is specifically used to: The fusion feature calculation formula is used to calculate the feature vector of the historical moment and the standardized data of the historical moment to obtain the fusion feature of the historical moment; The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state.

3. The system according to claim 2, characterized in that The relationship between the fusion features and the equipment status in the fault prediction model training module is as follows: f(P α )=sign(w*P α +b),P α ∈{P1,P2,…,P m } Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

4. The system according to claim 1, characterized in that The data preprocessing module is specifically used for: Performing cleaning, formatting, and normalization operations based on the multimodal data at the current moment to obtain preprocessed data at the current moment; Calculating the mean and standard deviation of the preprocessed data at the current moment to obtain the mean and standard deviation of the preprocessed data at the current moment; A standardization operation is performed based on the mean and standard deviation of the preprocessed data at the current moment to obtain standardized data at the current moment that are unified into the same magnitude.

5. The system according to claim 1, characterized in that The feature extraction module is specifically used for: The information features in the standardized data at the current moment are extracted by using target detection and image classification technology, and the information features are converted into feature vectors at the current moment.

6. The system according to claim 1, characterized in that The real-time detection and prediction module is specifically used for: Calculate the fusion feature at the current moment based on the feature vector at the current moment and the standardized data at the current moment using a fusion feature calculation formula to obtain the fusion feature at the current moment; The fault prediction model is input based on the fusion features at the current moment to perform prediction, and the device state at the next moment after the current moment is output as a fault prediction result.

7. A method for predicting faults of substation equipment based on multimodal data, characterized in that: include: Acquire multimodal data of substation equipment at the current moment based on multiple sensors; Performing preprocessing and standardization operations based on the multimodal data at the current moment to obtain standardized data at the current moment; Perform feature extraction based on the standardized data at the current moment to obtain a feature vector at the current moment; Based on the feature vector at the current moment and in combination with a pre-trained fault prediction model, a fault prediction is performed to obtain a fault prediction result; The fault prediction model is obtained by training a support vector machine model based on the fusion features calculated based on the feature vector of the historical moment and the standardized data as input data and the device state corresponding to the next moment of the historical moment as output data.

8. The method according to claim 7, characterized in that The training process of the fault prediction model includes: calculating the fusion features of the historical moments by using a fusion feature calculation formula based on the feature vectors of the historical moments and the standardized data of the historical moments; The support vector machine model is trained based on the fusion feature of the historical moment as input data and the device state corresponding to the fusion feature of the next moment of the historical moment as output data to determine the relationship between the fusion feature and the device state.

9. The method according to claim 8, characterized in that The relationship between the fusion feature and the device state is as follows: f(P α )=sign(w*P α +b),P α ∈{P1,P2,…,P m } Where P α is the fusion feature of the αth sample; f(P α ) is the device state; sign() is the sign function; w is the weight vector; b is the bias term; P1 is the fusion feature of the first sample; P2 is the fusion feature of the second sample; P m is the fusion feature of the mth sample.

10. The method according to claim 7, characterized in that The preprocessing and standardization operations are performed based on the multimodal data at the current moment to obtain the standardized data at the current moment, including: Performing cleaning, formatting, and normalization operations based on the multimodal data at the current moment to obtain preprocessed data at the current moment; Calculating the mean and standard deviation of the preprocessed data at the current moment to obtain the mean and standard deviation of the preprocessed data at the current moment; A standardization operation is performed based on the mean and standard deviation of the preprocessed data at the current moment to obtain standardized data at the current moment that are unified into the same magnitude.

11. The method according to claim 7, characterized in that The extracting features based on the standardized data at the current moment to obtain the feature vector at the current moment includes: The information features in the standardized data at the current moment are extracted by using target detection and image classification technology, and the information features are converted into feature vectors at the current moment.

12. The method according to claim 7, characterized in that The method of performing fault prediction on the substation equipment based on the feature vector at the current moment in combination with a pre-built fault prediction model to obtain a fault prediction result includes: Calculate the fusion feature at the current moment based on the feature vector at the current moment and the standardized data at the current moment using a fusion feature calculation formula to obtain the fusion feature at the current moment; The fault prediction model is input based on the fusion features at the current moment to perform prediction, and the device state at the next moment after the current moment is output as a fault prediction result.

Images