Multi-modal deep learning edge computing platform for power equipment

Abstract

The invention belongs to the technical field of power equipment, and particularly relates to a multi-modal deep learning edge computing platform for power equipment. The edge computing platform comprises a data acquisition module, a data fusion module, a dynamic model updating module, an edge computing optimization module, an intelligent alarm and prediction module, an upstream and downstream data link integration module, an edge device resource management module, a security enhancement module and a simulation and emulation module. And the data fusion module performs splicing, weighted fusion or PCA mathematical transformation on the features of different modalities in the feature extraction stage to form a comprehensive feature vector, and performs fusion according to model output results of different modalities in the model decision-making layer, including use of a voting mechanism, weighted scoring or logistic regression. According to the invention, by integrating a multi-modal deep learning technology and an edge computing architecture, comprehensive performance improvement and cost saving are brought to power equipment monitoring and management.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment, and specifically relates to a multimodal deep learning edge computing platform for power equipment. Background Art

[0002] Power equipment is a vital component of the power system, covering the entire process from power generation, transmission to use. Power generation equipment is the starting point of these devices, including generators, wind turbines, solar panels, etc. Generators convert mechanical energy or other forms of energy into electrical energy, while wind turbines use wind to drive generators to generate electricity, and solar panels convert sunlight into electrical energy through the photovoltaic effect. These devices start the conversion between the production and supply of electricity.

[0003] After electricity is generated, it needs to be processed by substation equipment. The main functions of substation equipment include voltage adjustment and circuit protection. Common substation equipment includes transformers, circuit breakers and disconnectors. Transformers are used to increase or decrease voltage to meet the power demand at different stages. Circuit breakers can automatically cut off faulty circuits to prevent system damage, while disconnectors are used to disconnect circuits during maintenance or overhaul.

[0004] Electricity transmission over long distances relies on transmission equipment, including high-voltage transmission lines and cables. High-voltage transmission lines can transmit electricity from power plants to load centers with minimal energy loss. The focus of this stage is to reduce power loss and ensure the stability of power transmission.

[0005] When electricity reaches the distribution network, the distribution equipment is responsible for distributing it to various users or terminal devices. Distribution boxes and distribution boards are common distribution equipment that distribute electricity to different areas or users and ensure the stability of power supply in each branch. Finally, electricity is delivered to specific electrical equipment, including motors, electric lights, electric heaters, and air conditioners. These electrical equipment converts electricity into various forms of energy to meet the needs of people's daily life and industrial production.

[0006] The existing computing platforms have poor edge computing efficiency for power equipment, large data transmission delays, and are not convenient for dynamically allocating computing, storage, and network resources according to system load and task priority, which is not conducive to reducing the operating costs of power equipment. To this end, we propose a multimodal deep learning edge computing platform for power equipment to solve the above problems. Summary of the invention

[0007] The purpose of this invention is to provide a multimodal deep learning edge computing platform for power equipment, which can bring comprehensive performance improvement and cost savings to power equipment monitoring and management by integrating multimodal deep learning technology and edge computing architecture.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A multimodal deep learning edge computing platform for electric power equipment, the edge computing platform comprising a data acquisition module, a data fusion module, a dynamic model update module, an edge computing optimization module, an intelligent alarm and prediction module, an upstream and downstream data link integration module, an edge device resource management module, a security enhancement module, and a simulation and emulation module.

[0010] In a preferred embodiment, the data fusion module concatenates, weightedly fuses or uses PCA mathematical transformation on features of different modalities in the feature extraction stage to form a comprehensive feature vector. In the model decision layer, the model output results of different modalities are fused, including using a voting mechanism, weighted scoring or logistic regression, using a network time synchronization protocol to ensure that different sensors and devices have the same time reference, and adopting a timestamp-related data caching strategy.

[0011] In a preferred solution, the dynamic model update module periodically runs an incremental training strategy, continuously adds newly collected data to the existing data set, performs fine-tuning and updates the original model, and dynamically adjusts the learning rate according to performance indicators, including the rate of decrease of the loss function or the prediction accuracy. Multiple models are run on the edge device, and when the platform detects changes in data features or device status, it automatically switches to the model that best suits the current situation during runtime.

[0012] In a preferred solution, the edge computing optimization module reduces the model complexity, reduces the amount of calculation and improves the operating efficiency by removing less important neurons or connections in the network, optimizes the network structure through deep separable convolution and low-rank decomposition, reduces model parameters, and quantizes the floating-point weights of the model to lower bits of precision. The storage requirements of the model on the edge device are greatly reduced without sacrificing too much performance, and the storage occupancy and transmission bandwidth of the model are reduced by Huffman encoding and deep compression of the model parameters.

[0013] In a preferred embodiment, the intelligent alarm and prediction module establishes an adaptive threshold model based on historical data, regularly adjusts the threshold of anomaly detection, uses a deep learning algorithm in combination with multimodal data to learn the characteristics of each modality, and uses a time series prediction model to predict potential failure points of equipment based on historical data. When the prediction model captures a potential failure, it automatically generates warning information, notifies maintenance personnel, and records a warning log.

[0014] In a preferred solution, the upstream and downstream data link integration module uses standardized interface protocols, including OPC UA and MQTT, to exchange data with the power control system, perform real-time control and monitoring, adjust the parameters of the control system based on the analysis and prediction results of the edge computing platform, including switch status and load distribution, and conduct data docking and integration with the asset management and maintenance management systems, integrating the data and functions of different systems into a unified user interface to provide a comprehensive system view.

[0015] In a preferred solution, the edge device resource management module dynamically adjusts the provision of computing resources, storage resources and network bandwidth according to the system load and task priority, monitors the usage of various resources of the edge devices in real time and adjusts the resource allocation strategy in a timely manner, distributes computing tasks among multiple edge devices, monitors the load of each edge device in real time, and dynamically adjusts the task allocation strategy as needed.

[0016] In a preferred embodiment, the security enhancement module uses the TLS / SSL encryption protocol during data transmission to encrypt data stored on the device, protect the security of static data, and adopts multi-factor authentication technology to limit user access rights to different functions and data based on the user's role and needs.

[0017] In a preferred embodiment, the simulation and emulation module is used to test the performance of the system in different operating scenarios and fault conditions, simulate different operating scenarios and fault conditions, test the system's responsiveness and fault handling capabilities, use virtual reality technology to simulate human operation and system interaction, test new algorithms and system functions, and verify and test the model in a simulation environment to check its predictive ability and performance.

[0018] In a preferred embodiment, the data acquisition module obtains various types of data from the power equipment, including sensor data, image data, and audio data. The data includes temperature, pressure, current, and voltage sensor information, as well as images or videos captured by surveillance cameras.

[0019] The technical effects achieved by the present invention are:

[0020] The platform can integrate multimodal data from different sensors and form more comprehensive and accurate feature vectors through data fusion technology, thereby improving the accuracy of fault detection and prediction. The edge computing architecture enables data processing and analysis to be performed near the device, reducing data transmission delays and improving the system's real-time response capabilities, which is crucial for power scenarios that require rapid response.

[0021] The platform supports dynamic updating and optimization of models, can adapt to changing data features and device status, and keep the model's prediction performance always in the best state. The edge device resource management module can dynamically allocate computing, storage, and network resources according to system load and task priority to ensure efficient use of resources and avoid waste.

[0022] By adopting TLS / SSL encryption protocol and multi-factor authentication technology, the platform provides strong security protection during data transmission and storage, effectively protecting sensitive data from being leaked or abused. The upstream and downstream data link integration module enables the data and functions of different systems to be integrated on a unified user interface, providing operators with a comprehensive system view and simplifying the operation and management process.

[0023] Through deep learning algorithms and time series prediction models, the platform can accurately predict potential failure points of equipment and generate early warning information in a timely manner, which helps reduce unexpected downtime and improve equipment operating efficiency and maintenance effects. The modular design of the platform enables each functional module to be independently upgraded and expanded, facilitating customized development and deployment based on actual needs. By optimizing resource allocation, improving prediction accuracy, and reducing fault response time, the platform helps reduce the operating costs of power equipment, including maintenance costs, energy consumption, and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of a multimodal deep learning edge computing platform for power equipment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0026] See also Figure 1 As shown, the present invention provides a multimodal deep learning edge computing platform for power equipment, and the edge computing platform includes a data acquisition module, a data fusion module, a dynamic model update module, an edge computing optimization module, an intelligent alarm and prediction module, an upstream and downstream data link integration module, an edge device resource management module, a security enhancement module, and a simulation and emulation module.

[0027] The data acquisition module obtains various types of data from the power equipment, including sensor data, image data, and audio data. The data includes temperature, pressure, current, and voltage sensor information, as well as images or videos taken by surveillance cameras.

[0028] In the feature extraction stage, the data fusion module concatenates, weightedly fuses or uses PCA mathematical transformation to form a comprehensive feature vector for features of different modalities. In the model decision layer, it fuses the model output results of different modalities, including the use of voting mechanism, weighted scoring or logistic regression, and uses network time synchronization protocol to ensure that different sensors and devices have the same time reference, and adopts timestamp-related data caching strategy.

[0029] The dynamic model update module regularly runs an incremental training strategy, continuously adding newly collected data to the existing data set, fine-tuning and updating the original model, and dynamically adjusting the learning rate based on performance indicators, including the rate of decrease of the loss function or the prediction accuracy. Multiple models run on edge devices, and when the platform detects changes in data features or device status, it automatically switches to the model that best suits the current situation during runtime.

[0030] The edge computing optimization module reduces model complexity, reduces computational effort, and improves operational efficiency by removing less important neurons or connections in the network. It optimizes the network structure through deep separable convolution and low-rank decomposition, reduces model parameters, and quantizes the model's floating-point weights to lower bits of precision. It significantly reduces the model's storage requirements on edge devices without sacrificing too much performance. It reduces the model's storage occupancy and transmission bandwidth by performing Huffman coding and deep compression on the model parameters.

[0031] The intelligent alarm and prediction module establishes an adaptive threshold model based on historical data, regularly adjusts the threshold of anomaly detection, uses deep learning algorithms in combination with multimodal data to learn the characteristics of each modality, and uses time series prediction models to predict potential failure points of equipment based on historical data. When the prediction model captures potential failures, it automatically generates warning information, notifies maintenance personnel, and records warning logs.

[0032] The upstream and downstream data link integration module uses standardized interface protocols, including OPC UA and MQTT, to exchange data with the power control system, perform real-time control and monitoring, and adjust the control system parameters, including switch status and load distribution, based on the analysis and prediction results of the edge computing platform. It also conducts data docking and integration with the asset management and maintenance management systems, integrating the data and functions of different systems into a unified user interface to provide a comprehensive system view.

[0033] The edge device resource management module dynamically adjusts the provision of computing resources, storage resources, and network bandwidth according to system load and task priority, monitors the usage of various resources of edge devices in real time and adjusts the resource allocation strategy in a timely manner, distributes computing tasks among multiple edge devices, monitors the load of each edge device in real time, and dynamically adjusts the task allocation strategy as needed.

[0034] The security enhancement module uses TLS / SSL encryption protocol during data transmission to encrypt data stored on the device, protect the security of static data, and adopts multi-factor authentication technology to limit user access to different functions and data based on user roles and needs.

[0035] The simulation and emulation module is used to test the performance of the system in different operating scenarios and fault conditions, simulate different operating scenarios and fault conditions, test the system's responsiveness and fault handling capabilities, use virtual reality technology to simulate the interaction between human operations and the system, test new algorithms and system functions, and verify and test the model in a simulation environment to check its predictive ability and performance.

[0036] The platform can integrate multimodal data from different sensors and form more comprehensive and accurate feature vectors through data fusion technology, thereby improving the accuracy of fault detection and prediction. The edge computing architecture enables data processing and analysis to be performed near the device, reducing data transmission delays and improving the system's real-time response capabilities, which is crucial for power scenarios that require rapid response.

[0037] The platform supports dynamic updating and optimization of models, can adapt to changing data features and device status, and keep the model's prediction performance always in the best state. The edge device resource management module can dynamically allocate computing, storage, and network resources according to system load and task priority to ensure efficient use of resources and avoid waste.

[0038] By adopting TLS / SSL encryption protocol and multi-factor authentication technology, the platform provides strong security protection during data transmission and storage, effectively protecting sensitive data from being leaked or abused. The upstream and downstream data link integration module enables the data and functions of different systems to be integrated on a unified user interface, providing operators with a comprehensive system view and simplifying the operation and management process.

[0039] Through deep learning algorithms and time series prediction models, the platform can accurately predict potential failure points of equipment and generate early warning information in a timely manner, which helps reduce unexpected downtime and improve equipment operating efficiency and maintenance effects. The modular design of the platform enables each functional module to be independently upgraded and expanded, facilitating customized development and deployment based on actual needs. By optimizing resource allocation, improving prediction accuracy, and reducing fault response time, the platform helps reduce the operating costs of power equipment, including maintenance costs, energy consumption, and labor costs.

[0040] In the present invention, the data acquisition module obtains various types of data from the power equipment, including sensor data, image data and audio data. Then, the data fusion module extracts and fuses the features of the data of different modes, generates a comprehensive feature vector, and integrates the model output results of different modes through a voting mechanism or weighted scoring at the model decision layer. At the same time, the platform uses the network time synchronization protocol to ensure the consistency of data time, and adopts a timestamp-related data caching strategy;

[0041] On this basis, the dynamic model update module regularly performs incremental training and model fine-tuning, dynamically adjusts the learning rate according to performance indicators, and automatically switches to the most suitable model when data features or device status change. The edge computing optimization module optimizes computing resources and improves operating efficiency by simplifying models, quantizing parameters, and compressing storage.

[0042] The intelligent alarm and prediction module uses historical data to establish adaptive threshold models and time series prediction models to detect potential faults and generate warning information to notify maintenance personnel. The upstream and downstream data link integration module realizes data exchange and integration with the power control system and other management systems through standardized interface protocols, providing a unified system view;

[0043] In addition, the edge device resource management module dynamically adjusts computing, storage and network resources according to system load and task priority, and allocates tasks in real time. The security enhancement module uses encryption technology and multi-factor authentication during data transmission and storage to ensure the security of data and systems. Finally, the simulation and emulation module is used to test the performance of the system in different operating scenarios and fault conditions, and to verify the effectiveness of algorithms and system functions.

[0044] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0045] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A multi-modal deep learning edge computing platform for power equipment, characterized by: The edge computing platform includes a data acquisition module, a data fusion module, a dynamic model update module, an edge computing optimization module, an intelligent alarm and prediction module, an upstream and downstream data link integration module, an edge device resource management module, a security enhancement module, and a simulation and emulation module.

2. The multimodal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The data fusion module concatenates, weightedly fuses or uses PCA mathematical transformation on the features of different modalities in the feature extraction stage to form a comprehensive feature vector. In the model decision layer, the model output results of different modalities are fused, including using a voting mechanism, weighted scoring or logistic regression, using a network time synchronization protocol to ensure that different sensors and devices have the same time reference, and adopting a timestamp-related data caching strategy.

3. The multimodal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The dynamic model update module regularly runs an incremental training strategy, continuously adds newly collected data to the existing data set, and fine-tunes and updates the original model. It dynamically adjusts the learning rate according to performance indicators, including the rate of decrease of the loss function or the prediction accuracy, and runs multiple models on the edge device. When the platform detects changes in data features or device status, it automatically switches to the model that best suits the current situation during runtime.

4. The multimodal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The edge computing optimization module reduces model complexity, reduces computational effort and improves operational efficiency by removing less important neurons or connections in the network. It optimizes the network structure through deep separable convolution and low-rank decomposition, reduces model parameters, and quantizes the model's floating-point weights to lower bits of precision. It greatly reduces the model's storage requirements on edge devices without sacrificing too much performance. It reduces the model's storage occupancy and transmission bandwidth by performing Huffman coding and deep compression on the model parameters.

5. The multi-modal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The intelligent alarm and prediction module establishes an adaptive threshold model based on historical data, regularly adjusts the threshold of anomaly detection, uses a deep learning algorithm in combination with multimodal data to learn the characteristics of each modality, and uses a time series prediction model to predict potential failure points of equipment based on historical data. When the prediction model captures a potential failure, it automatically generates warning information, notifies maintenance personnel, and records a warning log.

6. The multimodal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The upstream and downstream data link integration module uses standardized interface protocols, including OPC UA and MQTT, to exchange data with the power control system, perform real-time control and monitoring, and adjust the parameters of the control system, including switch status and load distribution, based on the analysis and prediction results of the edge computing platform. It also conducts data docking and integration with the asset management and maintenance management systems, integrates the data and functions of different systems into a unified user interface, and provides a comprehensive system view.

7. The multi-modal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The edge device resource management module dynamically adjusts the provision of computing resources, storage resources and network bandwidth according to the system load and task priority, monitors the usage of various resources of the edge devices in real time and adjusts the resource allocation strategy in time, distributes computing tasks among multiple edge devices, monitors the load of each edge device in real time, and dynamically adjusts the task allocation strategy as needed.

8. The multi-modal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The security enhancement module uses the TLS / SSL encryption protocol during data transmission to encrypt data stored on the device, protect the security of static data, and adopts multi-factor authentication technology to limit user access rights to different functions and data based on the user's role and needs.

9. The multimodal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The simulation and emulation module is used to test the performance of the system in different operating scenarios and fault conditions, simulate different operating scenarios and fault conditions, test the system's responsiveness and fault handling capabilities, use virtual reality technology to simulate human operation and system interaction, test new algorithms and system functions, and verify and test the model in a simulation environment to check its predictive ability and performance.

10. The multi-modal deep learning edge computing platform for power equipment according to claim 1, characterized in that: The data acquisition module acquires various types of data from the power equipment, including sensor data, image data, and audio data. The data includes temperature, pressure, current, and voltage sensor information, as well as images or videos captured by monitoring cameras.

Images