Power equipment health monitoring method and system fusing spatial-temporal characteristics and semantic information

Abstract

The embodiment of the invention provides a power equipment health monitoring method and system fusing spatial-temporal characteristics and semantic information, and belongs to the technical field of power equipment health monitoring. The power equipment health monitoring method comprises the steps that historical multi-source data of power equipment are acquired, the historical multi-source data are preprocessed, and the historical multi-source data comprise time sequence data, spatial data and semantic data; extracting time sequence features, spatial features and semantic features of the preprocessed historical multi-source data; carrying out feature fusion on the time sequence features, the spatial features and the semantic features; constructing and training a health state evaluation model according to the fusion features; therefore, the health state of the current power equipment can be rapidly and accurately evaluated based on the real-time data of the current power equipment, and the reliability and safety of the operation of the power equipment are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment health monitoring technology, and specifically to a power equipment health monitoring method and system that integrates spatiotemporal features and semantic information. Background Technology

[0002] The health status of power equipment (such as transformers, GIS switchgear, cables, etc.) directly affects the stability of the power grid and the reliability of power supply.

[0003] Traditional power equipment health maintenance strategies rely on manual experience and fixed-cycle maintenance, which have the following problems: 1. Low data utilization, relying only on local sensor data (such as temperature and vibration), lacking systematic analysis of the equipment's operating status throughout its entire life cycle; 2. Single maintenance strategy, not combined with the actual operating conditions of the equipment, leading to over-maintenance or under-maintenance; 3. Delayed response, unable to detect equipment abnormalities in real time and make quick decisions, affecting fault handling efficiency.

[0004] In the process of realizing this invention, the inventors of this application discovered that the above-mentioned solutions in the prior art have the defect of poor accuracy in monitoring the health of power equipment. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for monitoring the health of power equipment that integrates spatiotemporal features and semantic information. This method and system for monitoring the health of power equipment that integrates spatiotemporal features and semantic information has the function of high monitoring accuracy of power equipment.

[0006] To achieve the above objectives, one embodiment of the present invention provides a method for monitoring the health of power equipment by fusing spatiotemporal features and semantic information, comprising: Historical multi-source data of power equipment is acquired, and the historical multi-source data is preprocessed, wherein the historical multi-source data includes time-series data, spatial data, and semantic data; Extract the temporal, spatial, and semantic features of the preprocessed historical multi-source data; The temporal features, spatial features, and semantic features are fused together. A health status assessment model is constructed and trained based on the fusion features; Obtain real-time data of the power equipment currently described; The current health status of the power equipment is obtained based on the real-time data and the health status assessment model.

[0007] Optionally, preprocessing the historical multi-source data includes: The historical multi-source data is cleaned. The cleaned historical multi-source data is then standardized.

[0008] Optionally, extracting the temporal, spatial, and semantic features of the preprocessed historical multi-source data includes: using a long short-term memory network and / or extracting temporal features from the temporal data.

[0009] Optionally, extracting the temporal, spatial, and semantic features of the preprocessed historical multi-source data further includes: using convolutional neural networks and / or feature encoding to extract spatial features from the spatial data.

[0010] Optionally, extracting the temporal, spatial, and semantic features of the preprocessed historical multi-source data further includes: extracting semantic features from the semantic data using the BERT model and / or LDA algorithm.

[0011] Optionally, feature fusion of the temporal features, spatial features, and semantic features includes: Obtain the feature weights for each feature; The fusion features are obtained according to formula (1). (1) in, As a feature of fusion, For the first Feature weights of each feature The number of features, and , For the first One characteristic, The number is an integer.

[0012] Optionally, obtaining the feature weights for each feature includes: The feature weights of each feature are obtained according to formula (2). (2) in, For the first Each parameter weight, The number is an integer.

[0013] Optionally, constructing and training a health status assessment model based on the fusion features includes: Obtain the equipment health index according to formula (3). (3) in, For equipment health index, For the fusion features described above, As a health characteristic; The remaining lifespan of the power equipment is predicted using a survival analysis model and the current equipment health index.

[0014] On the other hand, the present invention also provides a power equipment health monitoring system that integrates spatiotemporal features and semantic information, comprising: The data acquisition module is connected to the power system and is used to collect real-time data from each power device in the power system. The controller, connected to the data acquisition module, is used to execute any of the above-described power equipment health monitoring methods.

[0015] In another aspect, the present invention also provides a computer-readable storage medium storing instructions for being read by a machine to cause the machine to perform any of the above-described power equipment health monitoring methods.

[0016] Through the above technical solutions, the power equipment health monitoring method and system that integrates spatiotemporal features and semantic information provided by the present invention extracts and fuses features from historical multi-source data of power equipment, and constructs and trains a health status assessment model based on the fused features. Thus, it can quickly and accurately assess the health status of the current power equipment based on the real-time data of the current power equipment, thereby effectively improving the reliability and safety of power equipment operation.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a power equipment health monitoring method that integrates spatiotemporal features and semantic information according to an embodiment of the present invention; Figure 2 This is a flowchart of the preprocessing of historical multi-source data in a power equipment health monitoring method that integrates spatiotemporal features and semantic information according to an embodiment of the present invention. Figure 3 This is a flowchart of feature fusion in a power equipment health monitoring method that integrates spatiotemporal features and semantic information according to an embodiment of the present invention; Figure 4 This is a flowchart of constructing a health assessment model in a power equipment health monitoring method that integrates spatiotemporal features and semantic information according to an embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0020] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0021] Figure 1 This is a flowchart of a power equipment health monitoring method that integrates spatiotemporal features and semantic information according to an embodiment of the present invention. Figure 1 The method for monitoring the health of electrical equipment may include: In step S1, historical multi-source data of the power equipment is acquired and preprocessed. This historical multi-source data includes time-series data, spatial data, and semantic data. Specifically, time-series data may include voltage, current, etc.; spatial data may include infrared thermal imaging images, etc.; and semantic data may include maintenance text, etc. Further, the acquisition of time-series data may include collecting real-time parameters of the equipment such as voltage, current, temperature, and vibration (sampling frequency 50Hz~1kHz); the acquisition of spatial data can be achieved by obtaining surface temperature distribution images of the equipment using an infrared thermal imager (resolution 640×480 pixels); and the acquisition of semantic data may include integrating historical maintenance records, fault logs, and expert knowledge base text (such as PDF and Excel files).

[0022] In step S2, the temporal features, spatial features, and semantic features of the preprocessed historical multi-source data are extracted.

[0023] In step S3, feature fusion is performed on temporal features, spatial features, and semantic features.

[0024] In step S4, a health status assessment model is constructed and trained based on the fusion features.

[0025] In step S5, real-time data of the current power equipment is acquired. This real-time data is multi-source data, including time-series data, spatial data, and semantic data.

[0026] In step S6, the current health status of the power equipment is obtained based on real-time data and the health status assessment model.

[0027] In steps S1 to S6, historical multi-source data of the power equipment is first acquired, and then features are extracted from the data from different sources. Next, feature fusion is performed on the features from different sources to train the health status assessment model. Simultaneously, real-time data of the current power equipment is acquired, and features are extracted and fused from the temporal, spatial, and semantic data within this real-time data. The fused features are then input into the trained health status assessment model to obtain the current health status of the power equipment.

[0028] Traditional power equipment health monitoring relies on simple analysis of localized data, resulting in a single and inaccurate health assessment method. In this embodiment of the invention, a multi-source data feature fusion approach is used to evaluate the health status of power equipment, effectively improving the convenience and accuracy of health status assessment, thereby ensuring the reliability and safety of power equipment operation.

[0029] In this embodiment of the invention, after obtaining historical multi-source data of the power equipment, it is necessary to preprocess the historical multi-source data. Specific preprocessing steps may include, for example: Figure 2 As shown, in Figure 2 The method for monitoring the health of electrical equipment may also include: In step S10, the historical multi-source data is cleaned.

[0030] In step S11, the cleaned historical multi-source data is standardized.

[0031] In steps S10 and S11, a sliding window method can be used to eliminate temporal data drift, and wavelet transform (Daubechies-4) can be used to remove noise. Simultaneously, histogram equalization and normalization can be performed on the infrared image to unify the brightness range (0~255). Furthermore, natural language processing (NLP) techniques can be used to segment the text data, remove stop words, and extract stems.

[0032] In this embodiment of the invention, the extraction of time-series features from time-series data may include using a Long Short-Term Memory (LSTM) network to model voltage and current time-series data and extract long-term dependent features, such as harmonic distortion rate and load switching coefficient. Frequency domain analysis may also be performed, using Fast Fourier Transform (FFT) to extract harmonic energy distribution and identify the risk of converter valve overheating.

[0033] In this embodiment of the invention, the extraction of spatial features from spatial data may include using a convolutional neural network (CNN) to extract local features from infrared thermal imaging images, such as temperature hotspots and edge textures. Alternatively, feature encoding may be employed, using a ResNet-50 pre-trained model to extract high-level semantic features of the image, such as areas of equipment aging.

[0034] In this embodiment of the invention, the extraction of semantic features from semantic data may include using the BERT model to perform semantic analysis on the operation and maintenance text and generate context-related vectors, such as fault mode keyword embedding. Alternatively, topic modeling may be employed, using the LDA algorithm to mine implicit topics in the text, such as "insulation breakdown" and "mechanical wear".

[0035] In this embodiment of the invention, after obtaining the temporal features, spatial features, and semantic features, it is necessary to fuse the three features. Specifically, the fusion steps can be as follows: Figure 3 As shown. Specifically, in Figure 3 The method for monitoring the health of electrical equipment may also include: In step S30, the feature weights of each feature are obtained. The feature weights of each feature can be obtained according to formula (2). (2) in, For the first Each parameter weight, The number is an integer.

[0036] In step S31, the fusion features are obtained according to formula (1). (1) in, As a feature of fusion, For the first Feature weights of each feature The number of features, and , For the first One characteristic, The number is an integer.

[0037] In this embodiment of the invention, after obtaining the fusion features, a health status assessment model can be constructed and trained based on the fusion features. Specifically, the construction steps can be as follows: Figure 4 As shown. Specifically, in Figure 4 The method for monitoring the health of electrical equipment may also include: In step S40, the equipment health index is obtained according to formula (3). (3) in, For equipment health index, For the present The fusion characteristics of moments for Health characteristics at any given time.

[0038] In step S41, the remaining lifespan of the power equipment is predicted using a survival analysis model and the current equipment health index. The current equipment health index serves as an input covariate for the survival analysis (Cox) model to predict the remaining lifespan (RUL) of the power equipment, and the feedback is continuously optimized based on the prediction results. A periodic maintenance plan is generated based on the DHI (Digital Health Index) prediction of the remaining lifespan. When a sudden fault is detected, such as insulation breakdown, the circuit breaker is immediately tripped and a backup device is switched on.

[0039] In this embodiment of the invention, the invention may further include a reinforcement learning decision model. Specifically, the reinforcement learning decision model may include a state space, an action space, and a reward function. Specifically, the state space defines equipment health indices, environmental parameters (such as temperature and humidity), and maintenance resource constraints (such as the number of maintenance personnel). The action space includes maintenance operations (such as replacing parts and cleaning) and adjustments to operating parameters (such as voltage limits). The reward function comprehensively considers improvements in equipment reliability, reductions in maintenance costs, and reductions in downtime, employing a multi-objective optimization function. Specifically, Double DQN (DDQN) is used to reduce overestimation problems, Prioritized Experience Replay is introduced to accelerate convergence, and a multi-objective reward function is designed to balance equipment reliability, maintenance costs, and downtime.

[0040] In this embodiment of the invention, the invention may further include a visual interface and a feedback loop mechanism. Specifically, the visual interface provides a heatmap of equipment health status, a fault risk warning pop-up, and a maintenance task dashboard, supporting user-defined scenario thresholds (such as the threshold for DHI-triggered maintenance) and strategy priorities. In the feedback loop mechanism, maintenance personnel can annotate decision results, such as strategy effectiveness and execution difficulty, for model iterative optimization.

[0041] In this embodiment of the invention, online incremental learning can also be introduced, and an incremental training framework based on experience replay can be designed to adapt the model to new scene data. Simultaneously, a meta-learning strategy is introduced to quickly transfer to similar device types. Furthermore, feature weights can be dynamically adjusted according to the device's operating stage (e.g., debugging period, stable period), for example, focusing on vibration signals during the debugging period and temperature trends during the stable period. Specifically, during model iteration, the health assessment model is updated using an incremental learning framework, with new data synchronized weekly; attribution analysis is performed on inefficient strategies (e.g., misjudgment causes, resource conflicts), and the reward function weights are adjusted. Scene classification rules are dynamically expanded based on newly discovered fault modes; and cross-regional device data is shared through a federated learning mechanism to improve the model's generalization ability.

[0042] On the other hand, the present invention also provides a power equipment health monitoring system that integrates spatiotemporal features and semantic information. Specifically, the power equipment health monitoring system may include a data acquisition module and a controller. Specifically, the data acquisition module is connected to the power system and is used to acquire real-time data of each power device in the power system. The controller is connected to the data acquisition module and is used to execute any of the power equipment health monitoring methods described above.

[0043] Specifically, in hardware deployment, high-precision sensors (such as fiber Bragg grating temperature sensors and vibration accelerometers) can be installed at key locations on the device; edge computing nodes (such as NVIDIA Jetson) can be configured to achieve real-time data preprocessing. In software configuration, time-series databases (such as InfluxDB) can be deployed to store historical data, and Kafka message queues can be built to support high-concurrency data transmission.

[0044] During data annotation and model training, fault labels can be manually labeled in historical data to construct a supervised learning dataset. A transfer learning strategy is employed, using a CNN model pre-trained on ImageNet to fine-tune the infrared image classification task. When optimizing the agent's policy, various fault scenarios are simulated in a simulation environment (such as PSCAD / EMTDC) to train the agent's coping strategies. A / B testing is used to compare the impact of different reward functions on the decision-making performance.

[0045] Furthermore, edge computing nodes can be used to stream sensor data, updating device health indices every second. When the health index exceeds a threshold, a tiered warning system, such as green-yellow-red, is triggered. Maintenance commands can also be issued to execution terminals via industrial IoT platforms (such as OPC UA); execution results (such as maintenance time and fault elimination rate) are collected and fed back to the decision-making model.

[0046] In another aspect, the present invention also provides a computer-readable storage medium storing instructions for being read by a machine to cause the machine to perform any of the above-described power equipment health monitoring methods.

[0047] Through the above technical solutions, the power equipment health monitoring method and system that integrates spatiotemporal features and semantic information provided by the present invention extracts and fuses features from historical multi-source data of power equipment, and constructs and trains a health status assessment model based on the fused features. Thus, it can quickly and accurately assess the health status of the current power equipment based on the real-time data of the current power equipment, thereby effectively improving the reliability and safety of power equipment operation.

[0048] This invention improves fault detection accuracy to over 98% through multimodal feature fusion, achieving higher precision. Simultaneously, it enables precise maintenance, reducing unnecessary downtime by over 30%, resulting in greater efficiency. It supports flexible deployment across various types of power equipment and complex environments, broadening its applicability. Furthermore, it provides a standardized and reusable intelligent decision-making paradigm for power equipment operation and maintenance, making it more intelligent and reliable.

[0049] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0050] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0052] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0053] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0054] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0055] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0056] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0057] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for power equipment health monitoring by fusing spatio-temporal features and semantic information, characterized in that, The method comprises: acquiring historical multi-source data of a power equipment and preprocessing the historical multi-source data, wherein the historical multi-source data comprises time series data, spatial data and semantic data; extracting time series features, spatial features and semantic features of the preprocessed historical multi-source data; performing feature fusion on the time series features, the spatial features and the semantic features; constructing and training a health state evaluation model according to the fused features; acquiring real-time data of the current power equipment; acquiring a health state of the current power equipment according to the real-time data and the health state evaluation model.

2. The power equipment health monitoring method of claim 1, wherein, The preprocessing of the historical multi-source data comprises: performing data cleaning on the historical multi-source data; performing standardization processing on the cleaned historical multi-source data.

3. The power equipment health monitoring method of claim 1, wherein, The extraction of the time series features, the spatial features and the semantic features of the preprocessed historical multi-source data comprises: performing time series feature extraction on the time series data by using a long short-term memory network.

4. The power equipment health monitoring method of claim 3, wherein, The extraction of the time series features, the spatial features and the semantic features of the preprocessed historical multi-source data further comprises: performing spatial feature extraction on the spatial data by using a convolutional neural network and / or feature encoding.

5. The power equipment health monitoring method of claim 4, wherein, The extraction of the time series features, the spatial features and the semantic features of the preprocessed historical multi-source data further comprises: performing semantic feature extraction on the semantic data by using a BERT model and / or an LDA algorithm.

6. The power equipment health monitoring method of claim 1, wherein, The feature fusion of the time series features, the spatial features and the semantic features comprises: acquiring feature weights of each feature; acquiring fused features according to formula (1), ，（1） wherein, is a fusion feature, is a feature weight of the th feature, is a number of features, and , is the th feature, is an integer number.

7. The power equipment health monitoring method of claim 6, wherein, The acquisition of the feature weights of each feature comprises: acquiring the feature weights of each feature according to formula (2), ，（2） wherein is the first parameter weight, is an integer number.

8. The power equipment health monitoring method of claim 6, wherein, The construction and training of the health state evaluation model according to the fused features comprises: acquiring a device health index according to formula (3), ，（3） wherein, is a device health index, is a current said fusion feature, is a health feature; performing prediction on a remaining life of the power equipment by using a survival analysis model and the current device health index.

9. A power equipment health monitoring system fusing spatio-temporal features and semantic information, characterized in that, The method comprises: a data acquisition module connected with a power system, used for acquiring real-time data of each power equipment in the power system; a controller connected with the data acquisition module, used for executing the power equipment health monitoring method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores instructions for being read by a machine to make the machine execute the power equipment health monitoring method according to any one of claims 1-8.

Images