Island power distribution network multi-modal data fusion and early warning method for severe environment

Abstract

The invention discloses a severe environment-oriented island power distribution network multi-modal data fusion and early warning method, which comprises the steps of obtaining multi-modal data in an island power distribution network operation process and performing data preprocessing, including data cleaning, de-noising and normalization processing, so as to obtain multi-modal normalized data of the same dimension; extracting equipment image features and equipment electrical operation features from the multi-modal normalized data, performing weighted fusion processing to obtain a fusion feature matrix, and performing equipment state recognition on the fusion feature matrix through an intelligent detection model; and when a fault condition is met, identifying a fault type and positioning a fault position to obtain a fault detection result of the island power distribution network equipment, carrying out fault early warning strategy matching of a corresponding risk level on the island power distribution network equipment according to the fault detection result, and executing the matched fault early warning strategy to carry out fault early warning. The method has the advantages of improving the accuracy and response speed of fault detection and effectively guaranteeing safe and stable operation of the island power distribution network.

Description

Technical Field

[0001] This invention relates to the technical field of data processing for isolated power distribution networks, and in particular to a method for multimodal data fusion and early warning of isolated power distribution networks in harsh environments. Background Technology

[0002] An isolated distribution network refers to a distribution network system that can operate independently after being disconnected from the main power grid. These systems are typically located in remote or harsh environments, such as islands, deserts, and mountains. These areas have complex and variable natural environments, making them susceptible to extreme weather events such as torrential rain, blizzards, strong winds, high temperatures, and low temperatures; geological activities such as earthquakes, landslides, and mudslides; and human-caused damage such as theft and misoperation. These harsh environments pose significant challenges to the equipment condition, operational performance, and security of the distribution network.

[0003] In island regions, power distribution network equipment is constantly exposed to high humidity and salt spray environments, which can easily lead to equipment corrosion and decreased insulation performance. In desert regions, frequent sandstorms make heat dissipation difficult, and sand and dust can block heat dissipation channels, affecting normal equipment operation. In mountainous areas, power distribution networks may face severe weather conditions such as low temperatures, icing, and strong winds, leading to problems such as decreased mechanical performance of equipment, conductor galloping, and insulator flashover. In addition, the special characteristics of isolated island power distribution networks are also reflected in their limited operation and maintenance resources. Once a fault occurs, the troubleshooting and repair time is long, which can easily lead to power supply interruptions, seriously affecting the normal production and life of local residents and businesses.

[0004] Traditional power distribution network monitoring and detection methods primarily rely on single data sources, such as monitoring individual electrical parameters like voltage, current, and power. While these methods can reflect the operational status of the power distribution network to some extent, their reliance on a single data source makes it difficult to comprehensively and accurately reflect the network's true operating condition under harsh environments. For example, monitoring only electrical parameters cannot promptly detect mechanical damage, insulation aging, or environmental corrosion, leading to insufficient accuracy and timeliness in fault detection. Another example is image analysis methods based on visible light inspection using drones equipped with high-resolution cameras that cruise along preset routes to acquire visible light images of power distribution network equipment. However, this method relies solely on visible light images and cannot capture internal anomalies such as abnormal temperatures or partial discharges, exhibiting inherent limitations. Summary of the Invention

[0005] To address the problems of single data source, high false alarm rate, poor adaptability to harsh environments, and poor early warning in the monitoring of isolated distribution network equipment in existing technologies, this application provides a multimodal data fusion and early warning method for isolated distribution networks in harsh environments. This method can improve the accuracy and response speed of fault detection, and also has real-time early warning and diagnostic capabilities, helping operation and maintenance personnel to make quick decisions and effectively ensuring the safe and stable operation of isolated distribution networks.

[0006] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution: A multi-modal data fusion and early warning method for isolated power distribution networks in harsh environments, the method comprising: Acquire multimodal data during the operation of an isolated power distribution network, and perform data preprocessing on the multimodal data, including data cleaning, noise reduction and normalization, to obtain multimodal normalized data with the same dimension; Equipment image features and equipment electrical operation features are extracted from the multimodal normalized data, and the equipment image features and equipment electrical operation features are weighted and fused to obtain a fused feature matrix; The device status is identified by the fused feature matrix through a pre-built intelligent detection model. When the device status meets the fault conditions, the fault type is identified and the fault location is located to obtain the fault detection result of the islanded distribution network equipment. Based on the fault detection results, fault warning strategies of corresponding risk levels are matched for the islanded power distribution network equipment, and the matched fault warning strategies are executed to provide fault warnings.

[0007] In a preferred embodiment, this application can be further configured as follows: The acquisition of multimodal data during the operation of an isolated distribution network, and the preprocessing of the multimodal data, including data cleaning, noise reduction, and normalization, to obtain normalized multimodal data of the same dimension, includes the following normalization processing: According to the preset data format, the cleaned and denoised multimodal data is converted into multimodal normalized data with the same dimension. The expression for the same dimension is as follows: (1) in, This represents multimodal normalized data. Represents multimodal data. These represent the maximum and minimum values ​​of the multimodal data, respectively.

[0008] In a preferred embodiment, this application can be further configured as follows: extracting equipment image features and equipment electrical operation features from the multimodal normalized data, and performing weighted fusion processing on the equipment image features and the equipment electrical operation features to obtain a fused feature matrix, specifically including: Equipment image features of the isolated power distribution network are obtained by extracting equipment image features from the multimodal normalized data using a convolutional neural network. The electrical operation characteristics of the equipment in the islanded power distribution network are obtained by extracting the equipment electrical operation characteristics from the multimodal normalized data using a time series analysis algorithm. Based on the importance of the data source of the multimodal normalized data, weights are assigned to the equipment image features and the equipment electrical operation features; The weighted equipment image features and equipment electrical operation features are fused and stitched together to obtain the fused feature matrix of the islanded distribution network equipment.

[0009] In a preferred embodiment, this application can be further configured such that: the weighting of the device image features and the device electrical operation features based on the importance of the data source of the multimodal normalized data specifically includes: The weight allocation ratio expression is as follows: (2) in, Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data With tags Mutual information, modal features This indicates the image features or electrical operating characteristics of the equipment. This represents the adjustment coefficient that controls the sensitivity to differences between different modes. This represents the total amount of multimodal normalized data. Indicates the first of the same device Multimodal normalized data.

[0010] In a preferred embodiment, this application can be further configured such that: the feature fusion and stitching of the weighted device image features and device electrical operation features to obtain the fused feature matrix of the islanded distribution network equipment specifically includes: The expression for the fused feature matrix is ​​as follows: (3) in, Represents the fused feature matrix. Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data.

[0011] In a preferred embodiment, this application can be further configured as follows: the device status identification is performed on the fused feature matrix using a pre-built intelligent detection model; when the device status meets the fault conditions, the fault type is identified and the fault location is determined; the intelligent detection model in the fault detection result of the isolated distribution network equipment specifically includes: A smart detection model is constructed by training the historical feature vectors of isolated distribution network equipment using a multilayer perceptron and a Bi-LSTM network. The intelligent detection model is tested and verified using cross-validation and hyperparameter optimization, and the model is optimized based on the verification results.

[0012] In a preferred embodiment, this application can be further configured as follows: the device status identification is performed on the fused feature matrix using a pre-built intelligent detection model; when the device status meets the fault conditions, the fault type is identified and the fault location is determined, thereby obtaining the fault detection result of the islanded distribution network equipment. Specifically, this includes: The cross-modal feature nonlinear relationship in the fusion feature matrix is ​​extracted by the multilayer perceptron of the intelligent detection model to construct feature channels between different modes; The time-series patterns of the fusion feature matrix in the fault evolution process are extracted by the Bi-LSTM network of the intelligent detection model, and the time steps of different modal features are analyzed. An attention mechanism is introduced to calculate the correlation weights between different feature channels and their corresponding time steps. The device status is identified based on the correlation weights. When the device status meets the preset fault conditions, the key abnormal feature regions are determined. Identify the fault type of the key abnormal feature area, locate the fault location of the key abnormal feature area, and obtain the fault detection results of the islanded distribution network equipment.

[0013] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: A multi-modal data fusion and early warning system for isolated distribution networks in harsh environments is provided. The system is applied to the aforementioned multi-modal data fusion and early warning method for isolated distribution networks in harsh environments. The system includes: The multimodal data acquisition module is used to collect data on the operating environment and equipment status of isolated power distribution networks through various sensor devices; The data preprocessing module is used to preprocess the collected multimodal data, including data cleaning, noise reduction and normalization. The multimodal data fusion module is used to extract equipment image features and equipment electrical operation features from multimodal normalized data, and form a fusion feature matrix through weighted fusion; The intelligent detection module is used to detect the status, fault type, and fault location of power distribution network equipment based on a fused feature matrix using an intelligent detection model. The fault warning and diagnosis module is used to match warning strategies based on the detection results and execute the corresponding warning strategies, and provide diagnostic suggestions. The visualization and feedback module is used to display multimodal data, detection results, and early warning information through a visual interface.

[0014] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for multimodal data fusion and early warning of islanded power distribution networks in harsh environments.

[0015] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for multimodal data fusion and early warning in islanded power distribution networks for harsh environments.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. The technical solution of this application integrates multiple sensor data such as electrical parameters, environmental information, images, videos, and vibration and sound, and utilizes advanced data preprocessing and deep learning algorithms to achieve comprehensive perception and accurate identification of the status of power distribution equipment. Through multi-level and multi-dimensional data fusion and intelligent analysis, the system not only improves the accuracy and response speed of fault detection, but also has real-time early warning and diagnostic capabilities, helping maintenance personnel to make rapid decisions and effectively ensuring the safe and stable operation of isolated power distribution networks; 2. Comprehensive Detection and Fault Early Warning: By integrating data acquisition from multiple modalities such as electrical parameters, environmental parameters, image and video data, and vibration and sound data, a unified feature representation is constructed to achieve comprehensive monitoring of the distribution network status. This multimodal data fusion method not only covers traditional electrical parameter monitoring, but also introduces monitoring of environmental factors, equipment appearance, and mechanical status, enabling earlier detection of potential fault hazards and early warning. 3. Intelligent Detection Algorithm: The system employs advanced machine learning algorithms (such as MLP) and deep learning models (such as Bi-LSTM) to construct an intelligent detection model. These algorithms can automatically learn complex patterns and features in the data, adapt to complex data distributions in harsh environments, further improve the performance of fault detection, and introduce an attention mechanism to output the contribution of key modalities and uniformly achieve the three classification tasks of equipment status, fault type, and fault location. 4. Improved Operation and Maintenance Efficiency and Economic Benefits: Through visualization and feedback modules, maintenance personnel can monitor the operating status of the power distribution network in real time and understand the health status of equipment in a timely manner. Simultaneously, through early warning and precise fault location, maintenance personnel can respond to faults more quickly, reducing power outage time and maintenance costs. The design and implementation of the distributed architecture, including the deployment of edge computing terminals, the selection of data transmission protocols, and the collection and fusion of multimodal data, ensures the stable operation of the system in harsh environments. 5. Strong system generalization ability: It has strong adaptability and robustness, and can effectively cope with complex data distribution in harsh environments, ensuring the safe and stable operation of isolated power distribution networks. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is a flowchart illustrating the implementation of the multimodal data fusion and early warning method for islanded power distribution networks in harsh environments, as described in this embodiment.

[0019] Figure 2 This is a flowchart illustrating the implementation of step S20 of the islanded power distribution network multimodal data fusion and early warning method in this embodiment.

[0020] Figure 3 This is a flowchart of the intelligent detection model construction process for the multimodal data fusion and early warning method for isolated power distribution networks in this embodiment.

[0021] Figure 4 This is a flowchart illustrating the implementation of step S30 of the multimodal data fusion and early warning method for isolated power distribution networks in this embodiment.

[0022] Figure 5 This is a structural block diagram of the multimodal data fusion and early warning system for isolated power distribution networks in harsh environments, as described in this embodiment.

[0023] Figure 6 This is a schematic diagram of the internal structure of a computer device used to implement multimodal data fusion and early warning methods for isolated power distribution networks. Detailed Implementation

[0024] 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, not all, of the embodiments of the present invention. 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.

[0025] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] In one embodiment, such as Figure 1 As shown, this application discloses a multi-modal data fusion and early warning method for islanded distribution networks in harsh environments, which specifically includes the following steps: S10: Acquire multimodal data during the operation of the isolated distribution network, perform data preprocessing on the multimodal data, including data cleaning, noise reduction and normalization, to obtain multimodal normalized data with the same dimension.

[0029] Specifically, multiple sensors are deployed to collect multimodal data during the operation of the isolated power distribution network. This multimodal data includes electrical parameters, environmental parameters, equipment appearance parameters, and equipment mechanical condition parameters. Data cleaning is performed on the collected multimodal data, including removing invalid and outlier values ​​and merging duplicate data to reduce redundancy. Data denoising removes noise from electrical parameters to improve data accuracy. Normalization converts data from different modes to the same dimension, specifically including: According to the preset data format, the cleaned and denoised multimodal data is converted into multimodal normalized data with the same dimension. The expression for the same dimension is as follows: (1) in, This represents multimodal normalized data. Represents multimodal data. These represent the maximum and minimum values ​​of the multimodal data, respectively.

[0030] S20: Extract equipment image features and equipment electrical operation features from multimodal normalized data, and perform weighted fusion processing on the equipment image features and equipment electrical operation features to obtain a fused feature matrix.

[0031] Specifically, such as Figure 2 As shown, step S20 includes: S201: Extract equipment image features from multimodal normalized data using a convolutional neural network to obtain equipment image features of an isolated power distribution network.

[0032] Specifically, convolutional neural networks are used to analyze the equipment appearance image parameters from multimodal normalized data, and then deep features of the equipment appearance are extracted to obtain the equipment image features of isolated power distribution network equipment.

[0033] S202: Extract the electrical operation characteristics of equipment from multimodal normalized data using time series analysis algorithms to obtain the electrical operation characteristics of equipment in the isolated distribution network.

[0034] Specifically, the electrical parameters in the multimodal normalized data are extracted using time series analysis to obtain the electrical operating characteristics of the isolated power distribution network equipment.

[0035] S203: Based on the importance of the data source of the multimodal normalized data, assign weights to the equipment image features and the equipment electrical operation features.

[0036] Specifically, weights are assigned based on the importance of multimodal normalized data. The weight assignment is automatically optimized by combining mutual information with cross-validation and feature contribution rate to find the optimal ratio. The expression for the weight assignment ratio is as follows: (2) in, Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data With tags Mutual information, modal features This indicates the image features or electrical operating characteristics of the equipment. This represents the adjustment coefficient that controls the sensitivity to differences between different modes. This represents the total amount of multimodal normalized data. Indicates the first of the same device Multimodal normalized data.

[0037] S204: The weighted equipment image features and equipment electrical operation features are fused and stitched together to obtain the fused feature matrix of the islanded power distribution network equipment.

[0038] Specifically, the expression for the fused feature matrix is ​​as follows: (3) in, Represents the fused feature matrix. Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data.

[0039] S30: The device status is identified by the fusion feature matrix through a pre-built intelligent detection model. When the device status meets the fault conditions, the fault type is identified and the fault location is located to obtain the fault detection results of the islanded distribution network equipment.

[0040] Specifically, such as Figure 3 As shown, the intelligent detection model in step S30 specifically includes: S301: A smart detection model is constructed by training the historical feature vectors of isolated distribution network equipment through a multilayer perceptron and a Bi-LSTM network.

[0041] Specifically, based on the fusion features of historical feature vectors of isolated distribution network equipment, a smart detection model is constructed by training the fusion features through a multilayer perceptron and a Bi-LSTM network.

[0042] S302: The intelligent detection model is tested and verified using cross-validation and hyperparameter optimization, and the model is optimized based on the verification results.

[0043] Specifically, during model training, cross-validation and hyperparameter optimization are used to verify the detection results of the intelligent detection model. Based on the verification results, the model is optimized to improve its accuracy and robustness, thereby achieving efficient and stable intelligent fault detection.

[0044] Specifically, such as Figure 4 As shown, step S30 specifically includes: S303: Extract cross-modal feature nonlinear relationships in the fusion feature matrix through the multilayer perceptron of the intelligent detection model, and construct feature channels between different modes.

[0045] Specifically, the static cross-modal feature nonlinear relationship in the fusion feature matrix is ​​extracted by a multilayer perceptron, thereby constructing feature channels between different modalities.

[0046] S304: Extract the temporal pattern of the fusion feature matrix in the fault evolution process through the Bi-LSTM network of the intelligent detection model, and analyze the time step of different modal features.

[0047] Specifically, the Bi-LSTM network captures the temporal patterns in the fault evolution process based on the fused feature matrix, thereby achieving dynamic enhancement of feature representation, and then analyzing the time steps between different modal features based on the temporal patterns.

[0048] S305: An attention mechanism is introduced to calculate the correlation weights between different feature channels and their corresponding time steps. The device status is identified based on the correlation weights. When the device status meets the preset fault conditions, the key abnormal feature area is determined.

[0049] Specifically, an attention mechanism is introduced to calculate the correlation weights between different feature channels and their corresponding time steps, enabling the model to adaptively focus on key abnormal feature regions, such as spectral energy mutations, temperature anomalies, or local image degradation. Based on the correlation weights, the device status is identified, and when the device status meets the preset fault conditions of the key abnormal feature region, the key abnormal feature region is determined.

[0050] S306: Identify the fault type of the key abnormal feature area, locate the fault location of the key abnormal feature area, and obtain the fault detection results of the islanded distribution network equipment.

[0051] Specifically, the fault type is identified in key abnormal feature areas through an expert knowledge base, and the fault location of the corresponding fault type is performed in key abnormal feature fishing areas to obtain the fault detection results of isolated island power distribution network equipment.

[0052] S40: Based on the fault detection results, match fault warning strategies of corresponding risk levels for islanded distribution network equipment, and execute the matched fault warning strategies to issue fault warnings.

[0053] Specifically, based on the fault detection results, the islanded power distribution network equipment is classified into corresponding fault risk levels, and a fault early warning strategy corresponding to the risk level is matched. The matched fault early warning strategy is executed according to the matching results to issue fault early warnings, such as issuing early warning information containing fault type, location and severity, or automatically cutting off power or issuing alarms when a major accident is detected.

[0054] In this embodiment, historical data is also used to suggest current fault handling strategies to assist staff in making quick judgments and decisions.

[0055] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0056] In one embodiment, a multimodal data fusion and early warning system for islanded distribution networks in harsh environments is provided. This system corresponds one-to-one with the multimodal data fusion and early warning method for islanded distribution networks in harsh environments described in the previous embodiment. Figure 5 As shown, this multimodal data fusion and early warning system for isolated power distribution networks in harsh environments includes a multimodal data acquisition module, a data preprocessing module, a multimodal data fusion module, an intelligent detection module, a fault early warning and diagnosis module, and a visualization and feedback module. Detailed descriptions of each functional module are as follows: The multimodal data acquisition module is used to collect data on the operating environment and equipment status of isolated power distribution networks through various sensor devices.

[0057] Specifically, the electrical parameter acquisition unit uses high-precision current transformers and voltage transformers to monitor key indicators such as voltage, current, and power in real time. The environmental parameter acquisition unit collects temperature, humidity, wind speed, and rainfall data through weather stations and environmental sensors to reflect changes in the external environment. The image and video acquisition unit uses high-definition cameras and drones to dynamically capture the appearance and anomalies of power distribution equipment. The vibration and sound acquisition unit uses vibration and acoustic sensors to sense the mechanical state of equipment, comprehensively covering multi-dimensional information from electrical, environmental, and mechanical perspectives.

[0058] The data preprocessing module is used to preprocess the collected multimodal data, including data cleaning, noise reduction, and normalization.

[0059] The multimodal data fusion module is used to extract equipment image features and equipment electrical operation features from multimodal normalized data, and form a fusion feature matrix through weighted fusion.

[0060] The intelligent detection module is used to detect the status, fault type, and fault location of power distribution network equipment based on a fusion feature matrix using an intelligent detection model.

[0061] The fault warning and diagnosis module is used to match warning strategies based on the detection results, execute the corresponding warning strategies, and provide diagnostic suggestions.

[0062] The visualization and feedback module is used to display multimodal data, detection results, and early warning information through a visual interface, facilitating real-time monitoring of the power distribution network's operational status by maintenance personnel. It provides storage for historical data and supports queries based on time, equipment, and other criteria, facilitating subsequent maintenance and analysis.

[0063] The hardware deployment in this embodiment includes: (1) Deployment of data acquisition equipment Install high-precision current transformers and voltage transformers in critical equipment (such as transformers and switchgear). Deploy temperature and humidity sensors, wind speed sensors, and rainfall sensors inside and outside the substation. Install high-definition cameras with night vision capabilities and long-term video storage in critical equipment areas. Install vibration and sound sensors on critical equipment components.

[0064] (2) Deployment of data processing and communication equipment Configure high-performance, low-power servers for data preprocessing, fusion, and intelligent detection. Establish an industrial Ethernet or wireless communication network to ensure stable communication and comprehensive coverage within the substation. Simultaneously deploy firewalls or physical isolation devices to ensure data transmission security.

[0065] (3) Deployment of monitoring and early warning equipment Install a comprehensive monitoring device that integrates environmental monitoring, video processing, and other functions to monitor equipment status and environmental parameters in real time. Configure audible and visual alarms and SMS notification modules, and set temperature warning thresholds, humidity warning thresholds, and abnormal alarm response times.

[0066] (4) Emergency power supply and storage equipment Configure redundant power supplies and uninterruptible power supplies (UPS) to ensure continuous server operation during power outages. The server is equipped with sufficient storage space to support historical data retrieval and analysis, and long-term data storage.

[0067] Specific limitations regarding the multimodal data fusion and early warning system for isolated distribution networks in harsh environments can be found in the limitations of the multimodal data fusion and early warning method for isolated distribution networks in harsh environments described above, and will not be repeated here. Each module in the aforementioned multimodal data fusion and early warning system for isolated distribution networks in harsh environments can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0068] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores relevant data for multimodal data fusion and fault early warning in isolated distribution networks under harsh environments. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a multimodal data fusion and early warning method for isolated distribution networks in harsh environments.

[0069] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being executed by a processor to implement the steps of a multimodal data fusion and early warning method for islanded power distribution networks in harsh environments.

[0070] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application of the technical solution and the constraints involved. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0071] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0072] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0073] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A multi-modal data fusion and early warning method for isolated distribution networks in harsh environments, characterized in that, The method includes: Acquire multimodal data during the operation of an isolated power distribution network, and perform data preprocessing on the multimodal data, including data cleaning, noise reduction and normalization, to obtain multimodal normalized data with the same dimension; Equipment image features and equipment electrical operation features are extracted from the multimodal normalized data, and the equipment image features and equipment electrical operation features are weighted and fused to obtain a fused feature matrix; The device status is identified by the fused feature matrix through a pre-built intelligent detection model. When the device status meets the fault conditions, the fault type is identified and the fault location is located to obtain the fault detection result of the islanded distribution network equipment. Based on the fault detection results, fault warning strategies of corresponding risk levels are matched for the islanded power distribution network equipment, and the matched fault warning strategies are executed to provide fault warnings.

2. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 1, characterized in that, The process of acquiring multimodal data during the operation of the isolated power distribution network, and preprocessing the multimodal data, including data cleaning, noise reduction, and normalization, to obtain normalized multimodal data of the same dimension, includes the following normalization processing: According to the preset data format, the cleaned and denoised multimodal data is converted into multimodal normalized data with the same dimension. The expression for the same dimension is as follows: (1) in, This represents multimodal normalized data. Represents multimodal data. These represent the maximum and minimum values ​​of the multimodal data, respectively.

3. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 1, characterized in that, The step of extracting equipment image features and equipment electrical operation features from the multimodal normalized data, and then performing a weighted fusion process on the equipment image features and the equipment electrical operation features to obtain a fused feature matrix, specifically includes: Equipment image features of the isolated power distribution network are obtained by extracting equipment image features from the multimodal normalized data using a convolutional neural network. The electrical operation characteristics of the equipment in the islanded power distribution network are obtained by extracting the equipment electrical operation characteristics from the multimodal normalized data using a time series analysis algorithm. Based on the importance of the data source of the multimodal normalized data, weights are assigned to the equipment image features and the equipment electrical operation features; The weighted equipment image features and equipment electrical operation features are fused and stitched together to obtain the fused feature matrix of the islanded distribution network equipment.

4. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 3, characterized in that, The step of assigning weights to the device image features and the device electrical operation features based on the importance of the data source of the multimodal normalized data specifically includes: The weight allocation ratio expression is as follows: (2) in, Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data With tags Mutual information, modal features This indicates the image features or electrical operating characteristics of the equipment. This represents the adjustment coefficient that controls the sensitivity to differences between different modes. This represents the total amount of multimodal normalized data. Indicates the first Multimodal normalized data.

5. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 3, characterized in that, The step of fusing and stitching together the weighted equipment image features and equipment electrical operation features to obtain the fused feature matrix of the islanded distribution network equipment specifically includes: The expression for the fused feature matrix is ​​as follows: (3) in, Represents the fused feature matrix. Indicates the first The weight allocation ratio of multimodal normalized data. Indicates the first Modal features of multimodal normalized data.

6. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 1, characterized in that, The process of identifying the equipment status of the fused feature matrix using a pre-built intelligent detection model, identifying the fault type and locating the fault when the equipment status meets the fault conditions, and obtaining the intelligent detection model in the fault detection results of the isolated distribution network equipment specifically includes: A smart detection model is constructed by training the historical feature vectors of isolated distribution network equipment using a multilayer perceptron and a Bi-LSTM network. The intelligent detection model is tested and verified using cross-validation and hyperparameter optimization, and the model is optimized based on the verification results.

7. The method for multimodal data fusion and early warning of isolated distribution networks in harsh environments according to claim 6, characterized in that, The process involves using a pre-built intelligent detection model to identify the device status of the fused feature matrix. When the device status meets the fault conditions, the fault type is identified and the fault location is determined, resulting in the fault detection results for the isolated distribution network equipment. Specifically, this includes: The cross-modal feature nonlinear relationship in the fusion feature matrix is ​​extracted by the multilayer perceptron of the intelligent detection model to construct feature channels between different modes; The time-series patterns of the fusion feature matrix in the fault evolution process are extracted by the Bi-LSTM network of the intelligent detection model, and the time steps of different modal features are analyzed. An attention mechanism is introduced to calculate the correlation weights between different feature channels and their corresponding time steps. The device status is identified based on the correlation weights. When the device status meets the preset fault conditions, the key abnormal feature regions are determined. Identify the fault type of the key abnormal feature area, locate the fault location of the key abnormal feature area, and obtain the fault detection results of the islanded distribution network equipment.

8. A multi-modal data fusion and early warning system for isolated power distribution networks in harsh environments, characterized in that: The system is applied to the multimodal data fusion and early warning method for islanded distribution networks in harsh environments as described in any one of claims 1-7, and the system includes: The multimodal data acquisition module is used to collect data on the operating environment and equipment status of isolated power distribution networks through various sensor devices; The data preprocessing module is used to preprocess the collected multimodal data, including data cleaning, noise reduction and normalization. The multimodal data fusion module is used to extract equipment image features and equipment electrical operation features from multimodal normalized data, and form a fusion feature matrix through weighted fusion; The intelligent detection module is used to detect the status, fault type, and fault location of power distribution network equipment based on a fused feature matrix using an intelligent detection model. The fault warning and diagnosis module is used to match warning strategies based on the detection results and execute the corresponding warning strategies, and provide diagnostic suggestions. The visualization and feedback module is used to display multimodal data, detection results, and early warning information through a visual interface.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multimodal data fusion and early warning method for islanded power distribution networks in harsh environments as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multimodal data fusion and early warning method for islanded power distribution networks in harsh environments as described in any one of claims 1 to 7.

Images