Fault diagnosis method and system for wind-water linkage equipment

Abstract

The invention belongs to the technical field of fault diagnosis, and discloses a fault diagnosis method and system for wind-water linkage equipment. The method comprises the following steps: constructing a large wind-water linkage equipment fault diagnosis model; performing optimization training on the wind-water linkage equipment fault diagnosis large model; updating the large fault diagnosis model of the wind-water linkage equipment; collecting real-time multi-mode monitoring data; real-time multi-mode monitoring data is written into an OPC UA instance, and semantic mapping is carried out; extracting the structural features of the real-time multi-modal monitoring data graph by using the updated large fault diagnosis model of the wind-water linkage equipment, and performing fault diagnosis; generating an optimization control strategy by using the updated large fault diagnosis model of the wind-water linkage equipment; and continuously training the updated large fault diagnosis model of the wind-water linkage equipment. The problems that in the prior art, diagnosis accuracy is insufficient, real-time performance is poor, data processing capacity is limited, model updating is difficult, and intelligent optimization control is lacked are solved.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology, specifically relating to a fault diagnosis method and system for wind and water linkage equipment. Background Technology

[0002] Air-water linkage technology is a novel control method for subway station air conditioning systems. By linking the air and water systems, it optimizes air conditioning operation, improving energy efficiency and comfort. This addresses the problem of traditional air conditioning systems where airflow and water flow control are often independent, leading to low energy efficiency and poor comfort. Air-water linkage equipment includes: refrigeration units, chilled water pumps, cooling pumps, cooling towers, combined air conditioners, exhaust fans, air handling units, fresh air units, ductwork, water collectors, and distributors. During operation, various factors can cause equipment failures in air-water linkage equipment. The stable and reliable operation of this equipment is crucial for ensuring the comfort of subway station air conditioning systems. Therefore, achieving real-time and accurate fault diagnosis of air-water linkage equipment and timely repair when faults occur has become an important development direction in this field.

[0003] In the field of fault diagnosis and control of wind and water linkage equipment, although existing technologies have achieved equipment monitoring and management to a certain extent, there are still many shortcomings, including: 1) Insufficient diagnostic accuracy: Traditional fault diagnosis methods mainly rely on human experience and simple threshold judgments, which makes it difficult to accurately identify complex fault modes, resulting in a high rate of misdiagnosis and missed diagnosis; 2) Poor real-time performance: Existing technologies often adopt a centralized data processing method, which requires data to be transmitted to a central server for processing, resulting in large delays and failing to meet the needs of real-time fault diagnosis and rapid response. 3) Limited data processing capabilities: With the increase in the number of devices and the diversification of monitoring data, the data processing capabilities of existing technologies are insufficient to meet the real-time processing needs of large-scale, multimodal data; 4) Difficulty in model updates: Existing fault diagnosis models often struggle to adapt to changes in equipment operating conditions, resulting in long update cycles and an inability to promptly reflect new fault modes and equipment characteristics. 5) Lack of intelligent optimization control: Existing technologies mainly focus on fault diagnosis, lacking intelligent optimization control strategies based on diagnostic results, and thus cannot achieve autonomous adjustment and preventive maintenance of equipment. Summary of the Invention

[0004] To address the problems of insufficient diagnostic accuracy, poor real-time performance, limited data processing capabilities, difficulty in model updates, and lack of intelligent optimization control in existing technologies, this invention aims to provide a method and system for diagnosing faults in wind and water linkage equipment.

[0005] The technical solution adopted in this invention is as follows: A method for diagnosing faults in a wind-water linkage device includes the following steps: In the cloud data center, artificial intelligence algorithms are used to build a large-scale model for diagnosing faults in feng shui linkage equipment. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale model for diagnosing faults in feng shui linkage equipment. The large-scale model for diagnosing faults in feng shui linkage equipment is then deployed to all edge computing gateways. The edge computing gateway optimizes and trains a large-scale fault diagnosis model for the wind and water linkage equipment based on several historical multimodal monitoring data collected from the wind and water linkage equipment. It then extracts the historical parameters and historical experience of the trained large-scale fault diagnosis model and uploads them to the cloud data center. The cloud data center integrates all historical large model parameters according to the alliance learning mechanism, updates the large model for feng shui linkage equipment fault diagnosis, obtains the updated large model for feng shui linkage equipment fault diagnosis, and synchronizes the updated large model for feng shui linkage equipment fault diagnosis to all edge computing gateways. The monitoring data acquisition device collects real-time multimodal monitoring data of the corresponding wind and water linkage equipment, and uses the OPCUA protocol to upload the real-time multimodal monitoring data to the edge computing gateway within the communication range; The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, obtains the written OPC UA instance, and performs semantic mapping on the written OPC UA instance to obtain the semantically mapped OPC UA instance. The edge computing gateway uses the updated large-scale fault diagnosis model for wind and water linkage equipment to extract the real-time multimodal monitoring data graph structure features of the OPCUA instance after semantic mapping, and performs fault diagnosis based on the real-time multimodal monitoring data graph structure features to obtain real-time fault diagnosis results. The edge computing gateway uses the updated large-scale fault diagnosis model of the wind and water linkage equipment. Based on the structural characteristics of the real-time multimodal monitoring data and the real-time fault diagnosis results, it generates an optimized control strategy, obtains the real-time optimized control strategy, and sends the real-time optimized control strategy to the corresponding wind and water linkage equipment. The cloud data center continuously trains the updated fault diagnosis model of the feng shui linkage equipment according to the continuous learning mechanism, and then synchronizes the continuously trained fault diagnosis model of the feng shui linkage equipment to all edge computing gateways.

[0006] Furthermore, the large-scale fault diagnosis model for feng shui linkage equipment includes a named entity and entity relationship extraction model, a feng shui linkage equipment knowledge graph, a multimodal monitoring data feature extraction model, a feng shui linkage equipment fault diagnosis model, and a feng shui linkage equipment optimization control model.

[0007] Furthermore, in the cloud data center, an artificial intelligence algorithm is used to construct a large-scale fault diagnosis model for feng shui linkage equipment. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale fault diagnosis model for feng shui linkage equipment. The large-scale fault diagnosis model for feng shui linkage equipment is then deployed to all edge computing gateways, including the following steps: The cloud data center collects several samples of multimodal monitoring data and several samples of knowledge data of feng shui linkage equipment, and performs preprocessing to obtain several preprocessed samples of multimodal monitoring data and several preprocessed samples of knowledge data of feng shui linkage equipment. Using natural language processing algorithms, a named entity and entity relationship extraction model is constructed. Using the named entity and entity relationship extraction model, several named entities and several knowledge entity relationships are extracted from the knowledge data of each preprocessed Feng Shui linkage device. Based on the several named entities and several knowledge entity relationships, a knowledge graph of Feng Shui linkage device is constructed. The named entity and entity relationship extraction model is constructed based on the BERT-BiLSTM-CRF-SVM algorithm; Based on the data structure of the multimodal monitoring data, an OPC UA information model of the monitoring data acquisition device is constructed, and based on the OPC UA information model, the multimodal monitoring data of several samples is converted into a graph-structured multimodal monitoring data information graph of several samples. Based on the knowledge graph of the Feng Shui linkage equipment, semantic mapping is performed on the multimodal monitoring data information graphs of several samples to obtain several semantically mapped multimodal monitoring data information graphs of samples. Based on the information graphs of multimodal monitoring data of several semantically mapped samples, a feature extraction model for multimodal monitoring data is constructed using multimodal feature extraction and deep learning algorithms, and structural features of multimodal monitoring data graphs of several samples are obtained. The multimodal monitoring data feature extraction model is built based on the FPN-LSTM-GCN algorithm; Based on the structural features of multimodal monitoring data from several samples, a fault diagnosis model for wind and water linkage equipment was constructed using deep learning and swarm intelligence optimization algorithms, and fault diagnosis results for several samples were obtained. The fault diagnosis model for wind and water linkage equipment is constructed based on the RF-MLP-ASGA algorithm; Based on the structural features of several sample multimodal monitoring data graphs and the corresponding sample fault diagnosis results, an enhanced reinforcement learning algorithm is used to construct an optimized control model for wind and water linkage equipment. The optimized control model for wind and water linkage equipment is constructed based on the ML-MOGRPO algorithm; By integrating the named entity and entity relationship extraction model, the knowledge graph of feng shui linkage equipment, the multimodal monitoring data feature extraction model, the feng shui linkage equipment fault diagnosis model, and the feng shui linkage equipment optimization control model, a large-scale feng shui linkage equipment fault diagnosis model is obtained. Based on the experience playback mechanism of the continuous learning mechanism, a large model experience playback pool is set up for the large model of fault diagnosis of wind and water linkage equipment. Based on the elastic weight mechanism of the continuous learning mechanism, the original loss function of the large model of fault diagnosis of wind and water linkage equipment is adjusted to obtain the comprehensive loss function. Extract the original large model parameters of the large model for fault diagnosis of feng shui linkage equipment, send the OPC UA information model and the original large model parameters to all edge computing gateways, and reconstruct the large model based on the original large model parameters at the edge computing gateways to complete the deployment of the large model for fault diagnosis of feng shui linkage equipment. Based on the OPC UA information model, deploy an OPC UA server on the edge computing gateway and build an OPC UA instance in the address space of the OPC UA server.

[0008] Furthermore, the edge computing gateway optimizes and trains a large-scale fault diagnosis model for the wind-water linkage equipment based on several historical multimodal monitoring data collected from the wind-water linkage equipment. It then extracts the historical parameters and historical experience of the trained large-scale fault diagnosis model and uploads them to the cloud data center, including the following steps: The edge computing gateway receives several historical multimodal monitoring data from the wind and water linkage equipment sent by the corresponding monitoring data acquisition device, and performs preprocessing to obtain several preprocessed historical multimodal monitoring data. Based on several preprocessed historical multimodal monitoring data, the large-scale fault diagnosis model for wind and water linkage equipment was optimized and trained. Using a comprehensive loss function, the first historical loss value is generated for each optimization training process, and the historical experience of the large model for diagnosing faults in the wind and water linkage equipment is extracted for each optimization training process. If the number of training iterations reaches the threshold or the first historical loss value is lower than the loss value threshold, a large-scale model for fault diagnosis of the wind and water linkage equipment after training is obtained, and the corresponding historical large-scale model parameters are extracted. Upload the historical large model parameters and some historical experience obtained from the current edge computing gateway to the cloud data center.

[0009] Furthermore, the cloud data center, based on the consortium learning mechanism, integrates all historical large-scale model parameters to update the large-scale model for fault diagnosis of feng shui linkage equipment, obtaining an updated large-scale model for fault diagnosis of feng shui linkage equipment. This updated model is then synchronized to all edge computing gateways, including the following steps: The cloud data center receives historical large model parameters and some historical experience from all edge computing gateways, and integrates all historical large model parameters according to the consortium learning mechanism to obtain the optimal large model parameters. Based on the optimal large model parameters, the large model for fault diagnosis of wind and water linkage equipment is updated to obtain the updated large model for fault diagnosis of wind and water linkage equipment. Several historical experiences are stored in the large model experience playback pool. Extract the updated parameters of the large-scale model for fault diagnosis of the Feng Shui linkage equipment, send the updated parameters to all edge computing gateways, and update the large-scale model based on the updated parameters at the edge computing gateways to complete the synchronization of the updated large-scale model for fault diagnosis of the Feng Shui linkage equipment.

[0010] Furthermore, the edge computing gateway writes real-time multimodal monitoring data into the corresponding OPC UA instance to obtain the written OPC UA instance, and performs semantic mapping on the written OPC UA instance to obtain the semantically mapped OPC UA instance, including the following steps: The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, resulting in a written OPC UA instance. Using the named entity and entity relationship extraction model, extract several data named entities and several data entity relationships from the OPC UA instance after writing; Based on several named entities and relationships of knowledge entities in the knowledge graph of the Fengshui linkage device, semantic mapping is performed on several named entities and relationships of data entities of the written OPC UA instance to obtain the semantically mapped OPC UA instance.

[0011] Furthermore, the edge computing gateway uses the updated large-scale fault diagnosis model for the wind and water linkage equipment to extract the real-time multimodal monitoring data graph structure features of the semantically mapped OPC UA instance. Based on these features, fault diagnosis is performed to obtain real-time fault diagnosis results, including the following steps: The edge computing gateway extracts the real-time multimodal monitoring data information map of the OPC UA instance after semantic mapping, and inputs the real-time multimodal monitoring data information map into the updated large-scale model for fault diagnosis of wind and water linkage equipment. The updated multimodal monitoring data feature extraction model of the updated wind and water linkage equipment fault diagnosis big model is used to extract the real-time multimodal monitoring data graph structure features of OPC UA instance after semantic mapping; Based on the structural characteristics of the real-time multimodal monitoring data graph, the updated fault diagnosis model of the wind and water linkage equipment is used to perform fault diagnosis and obtain real-time fault diagnosis results.

[0012] Furthermore, the edge computing gateway uses the updated fault diagnosis model for the wind and water linkage equipment. Based on the structural characteristics of the real-time multimodal monitoring data graph and the real-time fault diagnosis results, it generates an optimized control strategy, obtains a real-time optimized control strategy, and sends the real-time optimized control strategy to the corresponding wind and water linkage equipment. This includes the following steps: The edge computing gateway adjusts the state space of the updated wind and water linkage equipment optimized control model based on the graph structure characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, thus obtaining the adjusted state space. Several first historical experiences are randomly selected from the large model experience replay pool, and the action space of the updated feng shui linkage equipment optimization control model is adjusted based on these first historical experiences to obtain the adjusted action space. Based on the adjusted state space and the adjusted action space, the updated feng shui linkage equipment optimization control model is used to generate an optimized control strategy, obtain a real-time optimized control strategy, and send the real-time optimized control strategy to the corresponding feng shui linkage equipment. The updated fault diagnosis model for wind and water linkage equipment generates real-time experience during the processing of real-time multimodal monitoring data. The real-time experience, real-time multimodal monitoring data, real-time fault diagnosis results, and real-time optimized control strategies are then uploaded to the cloud data center.

[0013] Furthermore, in the cloud data center, based on a continuous learning mechanism, the updated fault diagnosis model for the feng shui linkage equipment is continuously trained to obtain a continuously trained fault diagnosis model for the feng shui linkage equipment. This continuously trained fault diagnosis model is then synchronized to all edge computing gateways, including the following steps: The cloud data center provides distributed storage for real-time multimodal monitoring data, real-time fault diagnosis results, and real-time optimization control strategies for each wind and water linkage device. Several second historical experiences are randomly selected from the large model experience replay pool, and these second historical experiences are mixed with real-time experiences to obtain several mixed experiences. Based on several hybrid experiences, the updated large-scale model for fault diagnosis of wind and water linkage equipment is continuously trained, and a comprehensive loss function is used to generate the second historical loss value in each continuous training process. If the number of training iterations reaches the threshold or the second historical loss value is lower than the loss value threshold, a large model for fault diagnosis of the wind and water linkage equipment after continuous training is obtained, and the parameters of the large model after continuous training are extracted. After continuous training, the parameters of the large model are sent to all edge computing gateways. At the edge computing gateways, the large model is updated based on the parameters after continuous training, thus completing the synchronization of the large model for fault diagnosis of wind and water linkage equipment after continuous training. A fault diagnosis system for feng shui linkage equipment is provided to implement a fault diagnosis method for feng shui linkage equipment. The system includes a cloud data center, several edge computing gateways, several monitoring data acquisition devices, and several feng shui linkage devices. The cloud data center is connected to several edge computing gateways, and both the cloud data center and each edge computing gateway are equipped with a large-scale fault diagnosis model for feng shui linkage equipment. Each edge computing gateway is communicatively connected to its corresponding monitoring data acquisition device and feng shui linkage device, and each monitoring data acquisition device is communicatively connected to its corresponding feng shui linkage device.

[0014] The beneficial effects of this invention are as follows: This invention provides a method and system for fault diagnosis of feng shui linkage equipment. By constructing a powerful large-scale fault diagnosis model for feng shui linkage equipment and continuously optimizing and training it, the accuracy of fault diagnosis is significantly improved, and the occurrence of misdiagnosis and missed diagnosis is reduced. By utilizing an edge computing gateway for local data processing and model training, data transmission latency is greatly reduced, enabling real-time fault diagnosis and rapid response. The combination of distributed architecture and edge computing effectively enhances the system's ability to process large-scale, multimodal data, adapting to the monitoring needs of complex equipment groups. The introduction of a continuous learning mechanism and a large-scale model parameter synchronization strategy enables the fault diagnosis model to be updated in a timely manner to adapt to changes in equipment operating status. Based on the fault diagnosis results, intelligent optimization control strategies are provided to achieve autonomous adjustment and preventive maintenance of the equipment, thereby improving equipment operating efficiency.

[0015] Other beneficial effects of the present invention will be further explained in the specific embodiments. Attached Figure Description

[0016] Figure 1 This is a flowchart of the fault diagnosis method for wind and water linkage equipment in this invention.

[0017] Figure 2 This is a structural block diagram of the fault diagnosis system for the wind and water linkage equipment in this invention. Detailed Implementation

[0018] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1: like Figure 1 As shown, this embodiment provides a method for diagnosing faults in a wind-water linkage device, including the following steps: S1: Cloud data center, using artificial intelligence algorithms to build a large-scale model for feng shui linkage equipment fault diagnosis. Based on the continuous learning mechanism, a large-scale model experience playback pool and a comprehensive loss function are set for the large-scale model for feng shui linkage equipment fault diagnosis, and the large-scale model for feng shui linkage equipment fault diagnosis is deployed to all edge computing gateways. The large-scale model for fault diagnosis of Feng Shui linkage equipment includes a named entity and entity relationship extraction model, a knowledge graph of Feng Shui linkage equipment, a multimodal monitoring data feature extraction model, a fault diagnosis model of Feng Shui linkage equipment, and an optimization control model of Feng Shui linkage equipment. In the cloud data center, an artificial intelligence algorithm is used to build a large-scale fault diagnosis model for feng shui linkage equipment. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale fault diagnosis model for feng shui linkage equipment. The large-scale fault diagnosis model for feng shui linkage equipment is then deployed to all edge computing gateways, including the following steps: S1-1: Cloud data center, collects several sample multimodal monitoring data and several feng shui linkage equipment knowledge data, and performs preprocessing to obtain several preprocessed sample multimodal monitoring data and several preprocessed feng shui linkage equipment knowledge data; S1-2: Using natural language processing algorithms, construct a named entity and entity relationship extraction model. Using the named entity and entity relationship extraction model, extract several named entities and several knowledge entity relationships from the knowledge data of each preprocessed Feng Shui linkage device. Based on the several named entities and several knowledge entity relationships, construct a knowledge graph of Feng Shui linkage device. The named entity and entity relation extraction model is based on an algorithm derived from Bidirectional Encoder Representations from Transformers (BERT), Bidirectional Long Short-Term Memory (BiLSTM), Conditional Random Field (CRF), and Support Vector Machine (SVM). The model includes a word embedding module based on BERT, a semantic feature extraction module based on BiLSTM, a named entity extraction module based on CRF, and an entity relation extraction module based on SVM, which are connected in sequence. The word embedding module converts each word in the input knowledge into a vector representation in a high-dimensional space. Leveraging BERT's pre-training capabilities, it captures the semantic changes of words within their context. The word vectors generated by BERT are rich in semantic information, helping subsequent modules better understand the knowledge data. Through a bidirectional LSTM network, it learns the long-term dependencies and contextual information of the input sequence. BiLSTM can capture complex relationships between words, improving the model's semantic understanding and providing more accurate input features for the named entity extraction module, thus improving entity recognition accuracy. The named entity extraction module uses CRF to label the sequence and identify named entities in the text. CRF can utilize contextual information for constraints, reducing incorrect labeling, improving entity recognition accuracy, and generating structured named entity results for easier subsequent processing and analysis. The entity relation extraction module uses SVM to classify the extracted entity pairs into relations, handling various complex relation types and enhancing the model's relation extraction capabilities. S1-3: Based on the data structure of the multimodal monitoring data, construct an information model of the Open Platform Communications Unified Architecture (OPC UA) for the monitoring data acquisition device, and based on the OPC UA information model, convert several samples of multimodal monitoring data into a graph-structured information graph of several samples of multimodal monitoring data, including the following steps: S1-3-1: Based on the data structure of the multimodal monitoring data of the sample, confirm the monitoring indicators of the monitoring data acquisition device; the multimodal monitoring data includes sequence monitoring data (monitoring parameters such as temperature, humidity, current, voltage, and power) and image monitoring data (including monitoring images of the wind and water linkage equipment, etc.). S1-3-2: Based on the monitoring indicators of each monitoring data acquisition device, construct the corresponding sensor nodes; the nodes include object nodes, variable nodes, and method nodes; object nodes: represent objects in the real world, such as sequence data acquisition sensors or image data acquisition sensors; variable nodes: represent data values, such as temperature, pressure, etc.; method nodes: represent functions or operations that can be called. S1-3-3: Set corresponding node attributes and node relationships for each node, and construct the corresponding OPC UA information model based on all nodes, node attributes, and node relationships; node attributes are descriptive attributes that each node has, and these attributes define the characteristics of the node; S1-3-4: Based on the OPC UA information model, convert several sample multimodal monitoring data into a graph structure of several sample multimodal monitoring data information graph; S1-4: Based on the knowledge graph of the wind and water linkage equipment, perform semantic mapping on the multimodal monitoring data information graph of several samples to obtain several semantically mapped multimodal monitoring data information graphs of samples. S1-5: Based on the multimodal monitoring data information graphs of several semantically mapped samples, a multimodal monitoring data feature extraction model is constructed using multimodal feature extraction and deep learning algorithms, and the structural features of the multimodal monitoring data graphs of several samples are obtained. The multimodal monitoring data feature extraction model is constructed based on the Feature Pyramid Networks (FPN) - Long Short-Term Memory (LSTM) - Graph Convolutional Network (GCN) algorithm. The multimodal monitoring data feature extraction model includes an image data feature extraction module based on the FPN algorithm, a sequence data feature extraction module based on the LSTM algorithm, and a graph structure feature construction module based on the GCN algorithm. The image data feature extraction module and the sequence data feature extraction module are both connected to the graph structure feature construction module. The image data feature extraction module includes a basic network architecture built on the Convolutional Neural Networks (CNN) algorithm. It extracts feature maps from feature maps at different scales and fuses feature maps from different levels extracted by the CNN using a feature pyramid network structure and skip connections. Through upsampling and lateral connections, a feature pyramid is constructed, transferring high-level semantic information to lower levels and enhancing the semantic expressive power of low-level features. This effectively combines low-level detail features with high-level semantic features. The sequence data feature extraction module extracts long-term dependency features from time-series data. LSTM, through its special gating mechanism, can effectively capture and preserve long-term dependencies in sequence data. The image data feature extraction module extracts image features from image nodes as node features, while the sequence data feature extraction module extracts sequence data features from sequence nodes (including attribute nodes, data nodes, method nodes, and other non-image nodes) as node features. The GCN network performs feature propagation on the multimodal monitoring data information graph through convolution-like operations, integrating the node features of each node in the multimodal monitoring data information graph and extracting edge features between nodes to form the graph structure features of the multimodal monitoring data information graph. S1-6: Based on the structural features of multimodal monitoring data from several samples, a fault diagnosis model for wind and water linkage equipment is constructed using deep learning and swarm intelligence optimization algorithms, and fault diagnosis results for several samples are obtained. The fault diagnosis model for Feng Shui linkage equipment is constructed based on the Random Forest (RF)-Multilayer Perceptron (MLP)-Accurate Snow Geese Algorithm (ASGA) algorithm. The Feng Shui linkage equipment fault diagnosis model includes a key feature screening module based on the RF algorithm, a linkage equipment fault diagnosis module based on the MLP algorithm, and a fault diagnosis result optimization module based on the ASGA algorithm, which are connected in sequence. The key feature selection module filters out the most important feature subset for fault diagnosis from the extracted multimodal monitoring data graph structure features. RF, through the combination of multiple decision trees, can evaluate the importance of features and select the most discriminative features, reducing feature dimensionality, computational complexity, and improving model efficiency. The selected key features are more representative, which helps improve the accuracy and generalization ability of subsequent fault diagnosis predictions. The MLP in the linkage equipment fault diagnosis module further processes and analyzes the extracted key features, learns the complex mapping relationship between input features and output labels, and utilizes the powerful nonlinear modeling capability of MLP to improve the accuracy of linkage equipment fault diagnosis predictions and output probability distributions. The fault diagnosis result optimization module optimizes the prediction probability through the ASGA algorithm, improving the accuracy of the prediction probability and reducing prediction errors. S1-7: Based on the structural features of several sample multimodal monitoring data graphs and the corresponding sample fault diagnosis results, an enhanced reinforcement learning algorithm is used to construct an optimized control model for wind and water linkage equipment. The Fengshui linkage equipment optimization control model is constructed based on the meta-learning (ML) - multi-objective group relative policy optimization (MOGRPO) algorithm, and the Fengshui linkage equipment optimization control model includes a meta-learning module based on the ML algorithm and an optimization control strategy generation module based on the MOGRPO algorithm connected in sequence. The meta-learning module optimizes the initial network parameters of the policy network in the control policy generation module, enabling these parameters to quickly adapt to new and unseen inputs, thus improving the model's generalization ability. Even with unseen inputs, the policy network can be updated based on previous learning experience, enhancing the adaptability of the optimized control policy generation model. The objective function set of the optimized control policy generation module can handle multiple conflicting optimization objective functions, such as minimizing the impact of optimized control, minimizing the cost of optimized control, and maximizing the efficiency of optimized control response, generating optimized control policies that balance these objectives. The agent learns historical optimized control policies through an experience replay pool, continuously optimizing its own policy generation capabilities. The agent controls the policy network based on the learned experience to generate more effective optimized control policies. The design of the experience replay pool and the agent allows the model to continuously learn and optimize, improving the quality of policy generation. Due to the use of a group exploration approach, the optimized control policy generation module can avoid getting trapped in local optima to some extent. The policy network outputs the probability distribution of actions in a given state. The optimized control policy generation module directly updates the policy network through gradients, eliminating the need for the Critic model in traditional reinforcement learning, making the algorithm structure more concise. Based on the structural features of several sample multimodal monitoring data graphs and the corresponding sample fault diagnosis results, an enhanced reinforcement learning algorithm is used to construct an optimized control model for the wind and water linkage equipment, including the following steps: S1-7-1: Using the ML-MOGRPO algorithm, construct the initial optimal control policy generation model; the initial optimal control policy generation model includes the initial meta-learning module and the initial optimal control policy generation module; The introduced meta-learning mechanism enables the model to continuously learn and optimize, further improving the adaptability and effectiveness of the optimization control strategy. By comprehensively considering multiple objectives such as cost, impact and response efficiency, the generated optimization control strategy is more comprehensive and reasonable. S1-7-2: The optimization objective of the optimization control is used as the scenario for meta-learning. Based on the structural features of the multimodal monitoring data of several samples and the corresponding sample fault diagnosis results, the initial meta-learning module is trained to obtain the final meta-learning module. S1-7-3: Use the final meta-learning module to initialize the policy network of the initial optimization control policy generation module under different scenarios to obtain the initialized policy network; S1-7-4: Based on the different scenarios of the meta-learning module, i.e. the optimization objective of the optimization control policy, set the objective function set for the initial optimization control policy generation module with the initialized policy network, connect the comprehensive experience replay pool, and set the action space and state space for the agent of the initial optimization control policy generation module. S1-7-5: The optimization control strategy generation problem is used as a simulation environment, and an optimized control strategy generation module is obtained based on the initialized policy network and the intelligent agent with action space and state space. S1-7-6: Traverse all objective functions in the objective function set, and based on several historical fault prediction results, optimize and train the optimized control strategy generation module to obtain the final optimized control strategy generation module, and generate several historical optimized control strategy generation experiences. S1-7-7: Integrate the final meta-learning module and the final optimized control strategy generation module to obtain the final optimized control strategy generation model, and store several historical optimized control strategy generation experiences in the comprehensive experience replay pool; S1-8: Integrate the named entity and entity relationship extraction model, the knowledge graph of feng shui linkage equipment, the multimodal monitoring data feature extraction model, the feng shui linkage equipment fault diagnosis model, and the feng shui linkage equipment optimization control model to obtain a large-scale feng shui linkage equipment fault diagnosis model. S1-9: Based on the experience playback mechanism of the continuous learning mechanism, a large model experience playback pool is set up for the large model of fault diagnosis of wind and water linkage equipment. Based on the elastic weight mechanism of the continuous learning mechanism, the original loss function of the large model of fault diagnosis of wind and water linkage equipment is adjusted to obtain the comprehensive loss function. S1-10: Extract the original large model parameters of the large model for fault diagnosis of the Fengshui linkage equipment, send the OPC UA information model and the original large model parameters to all edge computing gateways, and reconstruct the large model based on the original large model parameters at the edge computing gateways to complete the deployment of the large model for fault diagnosis of the Fengshui linkage equipment. S1-11: Based on the OPC UA information model, deploy an OPC UA server on the edge computing gateway and build an OPC UA instance in the address space of the OPC UA server; S2: Edge computing gateway, based on several historical multimodal monitoring data collected from the wind and water linkage equipment, optimizes and trains a large-scale fault diagnosis model for the wind and water linkage equipment, extracts the historical parameters and some historical experience of the trained large-scale fault diagnosis model, and uploads them to the cloud data center, including the following steps: S2-1: Edge computing gateway, which receives several historical multimodal monitoring data of the wind and water linkage device sent by the corresponding monitoring data acquisition device, and performs preprocessing to obtain several preprocessed historical multimodal monitoring data. S2-2: Based on several preprocessed historical multimodal monitoring data, optimize and train the large-scale model for fault diagnosis of wind and water linkage equipment; S2-3: Use the comprehensive loss function to generate the first historical loss value in each optimization training process, and extract the historical experience of the large model for diagnosing the faults of the wind and water linkage equipment in each optimization training process. S2-4: If the number of optimization training times reaches the threshold or the first historical loss value is lower than the loss value threshold, then the large model for fault diagnosis of the wind and water linkage equipment after training is obtained, and the corresponding historical large model parameters are extracted. S2-5: Upload the historical large model parameters and some historical experience obtained from the current edge computing gateway to the cloud data center; S3: Cloud Data Center. Based on the consortium learning mechanism, it integrates all historical large model parameters to update the large model for fault diagnosis of feng shui linkage equipment, obtaining the updated large model for fault diagnosis of feng shui linkage equipment. This updated large model is then synchronized to all edge computing gateways, including the following steps: S3-1: Cloud Data Center, receives historical large model parameters and some historical experience from all edge computing gateways, and integrates all historical large model parameters according to the consortium learning mechanism to obtain the optimal large model parameters; S3-2: Update the large model for fault diagnosis of wind and water linkage equipment according to the optimal large model parameters to obtain the updated large model for fault diagnosis of wind and water linkage equipment, and store some historical experience in the large model experience playback pool. S3-3: Extract the updated large model parameters of the updated Feng Shui linkage equipment fault diagnosis large model, send the updated large model parameters to all edge computing gateways, and update the large model at the edge computing gateways according to the updated large model parameters to complete the synchronization of the updated Feng Shui linkage equipment fault diagnosis large model. S4: Monitoring data acquisition device, which collects real-time multimodal monitoring data of the corresponding wind and water linkage equipment, and uploads the real-time multimodal monitoring data to the edge computing gateway within the communication range using the OPC UA protocol; S5: The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, obtaining the written OPC UA instance, and performs semantic mapping on the written OPC UA instance to obtain the semantically mapped OPC UA instance. This includes the following steps: S5-1: Edge computing gateway, writes real-time multimodal monitoring data to the corresponding OPC UA instance, and obtains the written OPC UA instance; S5-2: Use the named entity and entity relationship extraction model to extract several data named entities and several data entity relationships from the OPC UA instance after writing; S5-3: Based on several knowledge named entities and several knowledge entity relationships in the knowledge graph of the Fengshui linkage device, perform semantic mapping on several data named entities and several data entity relationships of the written OPC UA instance to obtain the semantically mapped OPC UA instance. S6: Edge computing gateway, using the updated large-scale fault diagnosis model for wind and water linkage equipment, extracts the structural features of the real-time multimodal monitoring data graph of the OPC UA instance after semantic mapping, and performs fault diagnosis based on the structural features of the real-time multimodal monitoring data graph to obtain real-time fault diagnosis results, including the following steps: S6-1: Edge computing gateway, extracts the real-time multimodal monitoring data information map of the OPC UA instance after semantic mapping, and inputs the real-time multimodal monitoring data information map into the updated large model for fault diagnosis of wind and water linkage equipment; S6-2: Using the updated multimodal monitoring data feature extraction model of the updated large-scale fault diagnosis model for wind and water linkage equipment, extract the real-time multimodal monitoring data graph structure features of OPC UA instances after semantic mapping, including the following steps: S6-2-1: Image data feature extraction module of the updated multimodal monitoring data feature extraction model of the updated wind and water linkage equipment fault diagnosis big model, extracting several real-time image data features of OPC UA instance after semantic mapping; S6-2-2: The sequence data feature extraction module uses the updated multimodal monitoring data feature extraction model to extract several real-time sequence data features of the OPC UA instance after semantic mapping; S6-2-3: Using the graph structure feature construction module of the updated multimodal monitoring data feature extraction model, construct the real-time multimodal monitoring data graph structure features of the OPC UA instance after semantic mapping based on several real-time image data features and several real-time sequence data features; S6-3: Based on the structural characteristics of the real-time multimodal monitoring data graph, the updated fault diagnosis model of the wind-water linkage equipment is used to perform fault diagnosis and obtain real-time fault diagnosis results, including the following steps: S6-3-1: The key feature screening module of the updated wind and water linkage equipment fault diagnosis model uses the updated wind and water linkage equipment fault diagnosis big model to screen the key features of the real-time multimodal monitoring data graph structure features and obtain several real-time key features. S6-3-2: The linkage equipment fault diagnosis module using the updated feng shui linkage equipment fault diagnosis model performs linkage equipment fault diagnosis based on several real-time key features and obtains the real-time fault diagnosis probability distribution. S6-3-3: Using the fault diagnosis result optimization module of the updated wind-water linkage equipment fault diagnosis model, the real-time fault diagnosis probability distribution is optimized to obtain the real-time fault prediction result, including the following steps: S6-3-3-1: Encode the real-time probability optimization parameters of the real-time fault diagnosis probability distribution into an individual vector of the initial solution; S6-3-3-2: Enter the initialization phase. Use the Circle chaotic mapping sequence to initialize and generate several initial solutions, resulting in an initial ASGA population composed of several initial ASGA individuals (initial solutions). The formula is:

[0020] In the formula, The initial ASGA individuals generated for the Circle chaotic mapping sequence, i.e., the initial solutions; These are randomly generated ASGA individuals; i For ASGA individual indicators; For the remainder function; Compared to a randomly distributed population, the improved ASGA population, which uses the Circle chaotic mapping sequence to generate the initial population, has a more uniform initial position distribution, expands the search range of the ASGA population in space, increases the diversity of population positions, and to some extent improves the algorithm's tendency to get trapped in local optima, thereby improving the algorithm's optimization efficiency. S6-3-3-3: With minimizing the prediction error as the optimization objective, the fitness function of the ASGA algorithm is set according to the optimization objective, and the ASGA population parameters and maximum number of iterations are also set. The formula is:

[0021] In the formula, The fitness function; This represents the prediction error value. S6-3-3-4: Use the fitness function to obtain the initial fitness value of each initial ASGA individual in the initial ASGA population, and take the initial ASGA individual with the lowest fitness value as the leader goose. S6-3-3-5: Entering the exploration phase, a leader goose rotation mechanism, a call guidance mechanism, and a dynamic reverse mechanism are introduced to iteratively update the initial ASGA population, resulting in an updated ASGA population, and retaining the best individuals. The leader goose rotation mechanism selects a new leader goose in each iteration based on the fitness values ​​of individual ASGA individuals. This mechanism can prevent the leader goose from getting trapped in local optima too early and enhance the global search capability of the algorithm. The formula is:

[0022] In the formula, As the leading goose in an update; For the first The initial ASGA individual with the third-to-last fitness value in the initial ASGA population for the first iteration number; For the first The initial ASGA individual with the fifth-to-last fitness value in the initial ASGA population at the initial iteration number; This represents the current iteration number; The optimal individual; It is the first weighting factor; This is a function for generating random numbers; The call guidance mechanism adjusts the individual position update using a sound wave propagation attenuation model based on the distance between the ASGA individual and the leader goose. ASGA individuals that are closer to the leader goose have a greater influence on their position update and can quickly move closer to the optimal solution. ASGA individuals that are farther away have a less influence on their position update and can maintain a certain level of exploration ability. This mechanism can avoid excessive aggregation or dispersion of the group and improve the local search accuracy of the algorithm. The formula is:

[0023] In the formula, For each updated ASGA individual; For the first The initial ASGA individuals for the number of iterations; The initial sound intensity received by the ASGA individual; For sound intensity parameters; The initial sound intensity; The lowest acceptable sound intensity; The convergence factor; The initial ASGA individual that is furthest away; The parameter is random. Let Brownian motion function be used. These are Brownian motion parameters; For XOR processing;

[0024] In the formula, is the convergence factor; tanh(.) is the hyperbolic tangent function; This represents the current iteration number; This represents the maximum number of iterations. a max , a minThese are the maximum and minimum values ​​of the convergence factor, respectively; λ For the deceleration rate parameter, For decreasing period parameters, λ =-2 π , = π ; The dynamic reverse mechanism dynamically reverses the initial ASGA individuals, increasing the diversity of exploration directions and avoiding getting trapped in local optima; The formula is:

[0025] In the formula, For a single updated reverse ASGA individual; γ The coefficient of inertia is decreasing; L max , L min These are the maximum and minimum values ​​in the vector space, respectively. The leader goose from the first update, several ASGA individuals from the first update, and several reverse ASGA individuals from the first update are integrated to obtain the ASGA population from the first update, and the ASGA individual with the lowest fitness value is retained as the best individual. S6-3-3-6: Entering the development stage, anomaly boundary strategy and Gaussian mutation mechanism are introduced to perform a second update on the ASGA population updated once, resulting in a second-updated ASGA population, and the best individual is retained. The abnormal boundary strategy calculates the difference between the fitness value of each ASGA individual updated once and the average fitness value of the population. For ASGA individuals with fitness values ​​much higher than the population average, their position update method will be adjusted, such as using Gaussian mutation mechanism, larger step size or smaller step size. This mechanism can help individuals avoid getting trapped in local optima and improve the convergence speed and accuracy of the algorithm. The formula is:

[0026] In the formula, This is an ASGA individual that has undergone a second update; For each updated ASGA individual; The fitness function; This represents the average fitness value of the population. The ASGA individual with the highest fitness value; These are the second and third weighting factors; These are parameters for the Gaussian mutation mechanism; S6-3-3-7: If the number of iterations is greater than or equal to the iteration number threshold or the fitness value of the best individual is less than the fitness threshold, then the best individual will be output as the optimal solution. S6-3-3-8: Decode the individual vector of the optimal solution to obtain the optimal real-time probability optimization parameters, and optimize the real-time fault diagnosis probability distribution based on the optimal real-time probability optimization parameters to obtain the optimized real-time fault diagnosis probability distribution. S6-3-3-9: Take the real-time fault prediction label with the highest probability in the optimized real-time fault diagnosis probability distribution as the real-time fault prediction result. S7: Edge computing gateway, using the updated large-scale fault diagnosis model for wind and water linkage equipment, generates optimized control strategies based on the structural characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, obtains real-time optimized control strategies, and sends the real-time optimized control strategies to the corresponding wind and water linkage equipment, including the following steps: S7-1: Edge computing gateway, based on the structural characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, adjusts the state space of the updated wind-water linkage equipment fault diagnosis big model and the updated wind-water linkage equipment optimized control model to obtain the adjusted state space, including the following steps: S7-1-1: Edge computing gateway, based on the graph structure characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, uses the meta-learning module of the updated wind and water linkage equipment optimized control model of the updated wind and water linkage equipment fault diagnosis big model to adjust the initial network parameters of the strategy network of the optimized control strategy generation module to obtain the adjusted strategy network; S7-1-2: Based on the structural characteristics of the real-time multimodal monitoring data graph and the real-time fault diagnosis results, adjust the state space of the optimized control strategy generation module of the updated wind-water linkage equipment optimized control model to obtain the adjusted state space. S7-2: Randomly select several first historical experiences from the large model experience replay pool, and adjust the action space of the updated feng shui linkage equipment optimization control model based on these first historical experiences to obtain the adjusted action space. S7-3: Based on the adjusted state space and adjusted action space, use the updated feng shui linkage equipment optimization control model to generate an optimized control strategy, obtain a real-time optimized control strategy, and send the real-time optimized control strategy to the corresponding feng shui linkage equipment, including the following steps: S7-3-1: Select a real-time optimization objective function from the set of objective functions, and based on the real-time optimization objective function, use an agent to control the adjusted policy network to generate the probability distribution of all possible optimized control actions in the adjusted action space corresponding to each real-time state in the adjusted state space. S7-3-2: Take the most probable optimized control action in the adjusted action space as the corresponding real-time state's optimized control action; S7-3-3: Integrate and optimize the execution of control actions for all real-time states in the adjusted state space to obtain a real-time optimized control strategy, and send the real-time optimized control strategy to the corresponding wind and water linkage equipment; S7-4: Collect the real-time experience generated by the updated large model for fault diagnosis of wind and water linkage equipment during the processing of real-time multimodal monitoring data, and upload the real-time experience, real-time multimodal monitoring data, real-time fault diagnosis results and real-time optimization control strategies to the cloud data center; S8: Cloud Data Center, based on a continuous learning mechanism, continuously trains the updated fault diagnosis model of the feng shui linkage equipment to obtain a continuously trained fault diagnosis model of the feng shui linkage equipment, and synchronizes the continuously trained fault diagnosis model of the feng shui linkage equipment to all edge computing gateways, including the following steps: S8-1: Cloud Data Center, which provides distributed storage for real-time multimodal monitoring data, real-time fault diagnosis results, and real-time optimization control strategies for each wind and water linkage device; S8-2: Randomly extract several second historical experiences from the large model experience replay pool, and mix these second historical experiences with real-time experiences to obtain several mixed experiences; S8-3: Based on several hybrid experiences, the updated large-scale model for fault diagnosis of wind and water linkage equipment is continuously trained, and a comprehensive loss function is used to generate the second historical loss value in each continuous training process. S8-4: If the number of continuous training sessions reaches the threshold or the second historical loss value is lower than the loss value threshold, then the large model for fault diagnosis of the wind and water linkage equipment after continuous training is obtained, and the parameters of the large model after continuous training are extracted. S8-5: Send the parameters of the large model after continuous training to all edge computing gateways, and update the large model based on the parameters of the large model after continuous training at the edge computing gateways, thus completing the synchronization of the large model for fault diagnosis of wind and water linkage equipment after continuous training. Example 2: like Figure 2 As shown, this embodiment provides a fault diagnosis system for feng shui linkage equipment, which is used to implement a fault diagnosis method for feng shui linkage equipment. The system includes a cloud data center, several edge computing gateways, several monitoring data acquisition devices, and several feng shui linkage devices. The cloud data center is connected to several edge computing gateways, and both the cloud data center and each edge computing gateway are equipped with a large-scale fault diagnosis model for feng shui linkage equipment. Each edge computing gateway is communicatively connected to the corresponding monitoring data acquisition device and the feng shui linkage device, and each monitoring data acquisition device is communicatively connected to the corresponding feng shui linkage device.

[0027] The cloud data center is used to construct a large-scale fault diagnosis model for feng shui linkage equipment using artificial intelligence algorithms. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale model, and the model is deployed to all edge computing gateways. Based on a consortium learning mechanism, all historical large-scale model parameters are integrated to update the model, resulting in an updated model, which is then synchronized to all edge computing gateways. Finally, based on the continuous learning mechanism, the updated model is continuously trained to obtain a continuously trained model, which is also synchronized to all edge computing gateways. The edge computing gateway is used to optimize and train a large-scale fault diagnosis model for feng shui linkage equipment based on several historical multimodal monitoring data collected from the feng shui linkage equipment. It extracts historical model parameters and historical experience from the trained model and uploads them to the cloud data center. Real-time multimodal monitoring data is written to the corresponding OPC UA instance, resulting in a written OPC UA instance. Semantic mapping is then performed on the written OPC UA instance to obtain a semantically mapped OPC UA instance. Using the updated feng shui linkage equipment fault diagnosis model, an optimized control strategy is generated based on the graph structure characteristics of the real-time multimodal monitoring data and the real-time fault diagnosis results. This real-time optimized control strategy is then sent to the corresponding feng shui linkage equipment. The monitoring data acquisition device is used to collect real-time multimodal monitoring data of the corresponding wind and water linkage equipment, and uses the OPC UA protocol to upload the real-time multimodal monitoring data to the edge computing gateway within the communication range.

[0028] This invention provides a method and system for fault diagnosis of feng shui linkage equipment. By constructing a powerful large-scale fault diagnosis model for feng shui linkage equipment and continuously optimizing and training it, the accuracy of fault diagnosis is significantly improved, and the occurrence of misdiagnosis and missed diagnosis is reduced. By utilizing an edge computing gateway for local data processing and model training, data transmission latency is greatly reduced, enabling real-time fault diagnosis and rapid response. The combination of distributed architecture and edge computing effectively enhances the system's ability to process large-scale, multimodal data, adapting to the monitoring needs of complex equipment groups. The introduction of a continuous learning mechanism and a large-scale model parameter synchronization strategy enables the fault diagnosis model to be updated in a timely manner to adapt to changes in equipment operating status. Based on the fault diagnosis results, intelligent optimization control strategies are provided to achieve autonomous adjustment and preventive maintenance of the equipment, thereby improving equipment operating efficiency.

[0029] This invention is not limited to the optional embodiments described above, and anyone can derive other various forms of products based on the inspiration of this invention. The specific embodiments described above should not be construed as limiting the scope of protection of this invention; the scope of protection of this invention should be determined by the claims, and the specification can be used to interpret the claims.

Claims

1. A method for diagnosing faults in a wind-water linkage device, characterized in that: Includes the following steps: In the cloud data center, artificial intelligence algorithms are used to build a large-scale model for diagnosing faults in feng shui linkage equipment. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale model for diagnosing faults in feng shui linkage equipment. The large-scale model for diagnosing faults in feng shui linkage equipment is then deployed to all edge computing gateways. The edge computing gateway optimizes and trains a large-scale fault diagnosis model for the wind and water linkage equipment based on several historical multimodal monitoring data collected from the wind and water linkage equipment. It then extracts the historical parameters and historical experience of the trained large-scale fault diagnosis model and uploads them to the cloud data center. The cloud data center integrates all historical large model parameters according to the alliance learning mechanism, updates the large model for feng shui linkage equipment fault diagnosis, obtains the updated large model for feng shui linkage equipment fault diagnosis, and synchronizes the updated large model for feng shui linkage equipment fault diagnosis to all edge computing gateways. The monitoring data acquisition device collects real-time multimodal monitoring data of the corresponding wind and water linkage equipment, and uploads the real-time multimodal monitoring data to the edge computing gateway within the communication range using the OPC UA protocol; The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, obtains the written OPC UA instance, and performs semantic mapping on the written OPC UA instance to obtain the semantically mapped OPC UA instance. The edge computing gateway uses the updated large-scale fault diagnosis model for wind and water linkage equipment to extract the real-time multimodal monitoring data graph structure features of the OPC UA instance after semantic mapping, and performs fault diagnosis based on the real-time multimodal monitoring data graph structure features to obtain real-time fault diagnosis results. The edge computing gateway uses the updated large-scale fault diagnosis model of the wind and water linkage equipment. Based on the structural characteristics of the real-time multimodal monitoring data and the real-time fault diagnosis results, it generates an optimized control strategy, obtains the real-time optimized control strategy, and sends the real-time optimized control strategy to the corresponding wind and water linkage equipment. The cloud data center continuously trains the updated fault diagnosis model of the feng shui linkage equipment according to the continuous learning mechanism, and then synchronizes the continuously trained fault diagnosis model of the feng shui linkage equipment to all edge computing gateways.

2. The method for fault diagnosis of a wind-water linkage device according to claim 1, characterized in that: The aforementioned large-scale model for fault diagnosis of feng shui linkage equipment includes a named entity and entity relationship extraction model, a feng shui linkage equipment knowledge graph, a multimodal monitoring data feature extraction model, a feng shui linkage equipment fault diagnosis model, and a feng shui linkage equipment optimization control model.

3. The method for diagnosing faults in a wind-water linkage device according to claim 2, characterized in that: In the cloud data center, an artificial intelligence algorithm is used to build a large-scale fault diagnosis model for feng shui linkage equipment. Based on a continuous learning mechanism, an experience replay pool and a comprehensive loss function are set up for the large-scale fault diagnosis model for feng shui linkage equipment. The large-scale fault diagnosis model for feng shui linkage equipment is then deployed to all edge computing gateways, including the following steps: The cloud data center collects several samples of multimodal monitoring data and several samples of knowledge data of feng shui linkage equipment, and performs preprocessing to obtain several preprocessed samples of multimodal monitoring data and several preprocessed samples of knowledge data of feng shui linkage equipment. Using natural language processing algorithms, a named entity and entity relationship extraction model is constructed. Using the named entity and entity relationship extraction model, several named entities and several knowledge entity relationships are extracted from the knowledge data of each preprocessed Feng Shui linkage device. Based on the several named entities and several knowledge entity relationships, a knowledge graph of Feng Shui linkage device is constructed. The named entity and entity relationship extraction model is constructed based on the BERT-BiLSTM-CRF-SVM algorithm; Based on the data structure of the multimodal monitoring data, an OPC UA information model of the monitoring data acquisition device is constructed, and based on the OPC UA information model, the multimodal monitoring data of several samples is converted into a graph-structured multimodal monitoring data information graph of several samples. Based on the knowledge graph of the Feng Shui linkage equipment, semantic mapping is performed on the multimodal monitoring data information graphs of several samples to obtain several semantically mapped multimodal monitoring data information graphs of samples. Based on the information graphs of multimodal monitoring data of several semantically mapped samples, a feature extraction model for multimodal monitoring data is constructed using multimodal feature extraction and deep learning algorithms, and structural features of multimodal monitoring data graphs of several samples are obtained. The multimodal monitoring data feature extraction model is constructed based on the FPN-LSTM-GCN algorithm; Based on the structural features of multimodal monitoring data from several samples, a fault diagnosis model for wind and water linkage equipment was constructed using deep learning and swarm intelligence optimization algorithms, and fault diagnosis results for several samples were obtained. The aforementioned fault diagnosis model for the wind and water linkage equipment is constructed based on the RF-MLP-ASGA algorithm; Based on the structural features of several sample multimodal monitoring data graphs and the corresponding sample fault diagnosis results, an enhanced reinforcement learning algorithm is used to construct an optimized control model for wind and water linkage equipment. The aforementioned wind and water linkage equipment optimization control model is constructed based on the ML-MOGRPO algorithm; By integrating the named entity and entity relationship extraction model, the knowledge graph of feng shui linkage equipment, the multimodal monitoring data feature extraction model, the feng shui linkage equipment fault diagnosis model, and the feng shui linkage equipment optimization control model, a large-scale feng shui linkage equipment fault diagnosis model is obtained. Based on the experience playback mechanism of the continuous learning mechanism, a large model experience playback pool is set up for the large model of fault diagnosis of wind and water linkage equipment. Based on the elastic weight mechanism of the continuous learning mechanism, the original loss function of the large model of fault diagnosis of wind and water linkage equipment is adjusted to obtain the comprehensive loss function. Extract the original large model parameters of the large model for fault diagnosis of feng shui linkage equipment, send the OPC UA information model and the original large model parameters to all edge computing gateways, and reconstruct the large model based on the original large model parameters at the edge computing gateways to complete the deployment of the large model for fault diagnosis of feng shui linkage equipment. Based on the OPC UA information model, deploy an OPC UA server on the edge computing gateway and build an OPC UA instance in the address space of the OPC UA server.

4. The method for diagnosing faults in a wind-water linkage device according to claim 3, characterized in that: The edge computing gateway optimizes and trains a large-scale fault diagnosis model for the wind-water linkage equipment based on historical multimodal monitoring data collected from the equipment. It then extracts the historical parameters and historical experience of the trained model and uploads them to the cloud data center. The process includes the following steps: The edge computing gateway receives several historical multimodal monitoring data from the wind and water linkage equipment sent by the corresponding monitoring data acquisition device, and performs preprocessing to obtain several preprocessed historical multimodal monitoring data. Based on several preprocessed historical multimodal monitoring data, the large-scale fault diagnosis model for wind and water linkage equipment was optimized and trained. Using a comprehensive loss function, the first historical loss value is generated for each optimization training process, and the historical experience of the large model for diagnosing faults in the wind and water linkage equipment is extracted for each optimization training process. If the number of training iterations reaches the threshold or the first historical loss value is lower than the loss value threshold, a large-scale model for fault diagnosis of the wind and water linkage equipment after training is obtained, and the corresponding historical large-scale model parameters are extracted. Upload the historical large model parameters and some historical experience obtained from the current edge computing gateway to the cloud data center.

5. The method for fault diagnosis of a wind-water linkage device according to claim 4, characterized in that: In the cloud data center, based on the consortium learning mechanism, all historical large-scale model parameters are integrated to update the large-scale model for fault diagnosis of feng shui linkage equipment, resulting in an updated large-scale model for fault diagnosis of feng shui linkage equipment. This updated model is then synchronized to all edge computing gateways, including the following steps: The cloud data center receives historical large model parameters and some historical experience from all edge computing gateways, and integrates all historical large model parameters according to the consortium learning mechanism to obtain the optimal large model parameters. Based on the optimal large model parameters, the large model for fault diagnosis of wind and water linkage equipment is updated to obtain the updated large model for fault diagnosis of wind and water linkage equipment. Several historical experiences are stored in the large model experience playback pool. Extract the updated parameters of the large-scale model for fault diagnosis of the Feng Shui linkage equipment, send the updated parameters to all edge computing gateways, and update the large-scale model based on the updated parameters at the edge computing gateways to complete the synchronization of the updated large-scale model for fault diagnosis of the Feng Shui linkage equipment.

6. The method for fault diagnosis of a wind-water linkage device according to claim 5, characterized in that: The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, obtaining the written OPC UA instance, and performs semantic mapping on the written OPC UA instance to obtain the semantically mapped OPC UA instance. This process includes the following steps: The edge computing gateway writes real-time multimodal monitoring data to the corresponding OPC UA instance, resulting in the written OPC UA instance. Using the named entity and entity relationship extraction model, extract several data named entities and several data entity relationships from the OPC UA instance after writing; Based on several knowledge named entities and several knowledge entity relationships in the knowledge graph of the Fengshui linkage device, semantic mapping is performed on several data named entities and several data entity relationships of the written OPC UA instance to obtain the semantically mapped OPC UA instance.

7. A method for diagnosing faults in a wind-water linkage device according to claim 6, characterized in that: The edge computing gateway uses the updated large-scale fault diagnosis model for wind and water linkage equipment to extract the structural features of the real-time multimodal monitoring data graph of the semantically mapped OPC UA instance. Based on these structural features, fault diagnosis is performed to obtain real-time fault diagnosis results, including the following steps: The edge computing gateway extracts the real-time multimodal monitoring data information map of the OPC UA instance after semantic mapping, and inputs the real-time multimodal monitoring data information map into the updated large-scale model for fault diagnosis of wind and water linkage equipment. The updated multimodal monitoring data feature extraction model of the updated wind and water linkage equipment fault diagnosis big model is used to extract the real-time multimodal monitoring data graph structure features of OPC UA instance after semantic mapping; Based on the structural characteristics of the real-time multimodal monitoring data graph, the updated fault diagnosis model of the wind and water linkage equipment is used to perform fault diagnosis and obtain real-time fault diagnosis results.

8. The method for fault diagnosis of a wind-water linkage device according to claim 7, characterized in that: The edge computing gateway uses the updated fault diagnosis model for wind and water linkage equipment. Based on the structural characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, it generates optimized control strategies, obtains real-time optimized control strategies, and sends these strategies to the corresponding wind and water linkage equipment. The process includes the following steps: The edge computing gateway adjusts the state space of the updated wind and water linkage equipment optimized control model based on the graph structure characteristics of real-time multimodal monitoring data and real-time fault diagnosis results, thus obtaining the adjusted state space. Several first historical experiences are randomly selected from the large model experience replay pool, and the action space of the updated feng shui linkage equipment optimization control model is adjusted based on these first historical experiences to obtain the adjusted action space. Based on the adjusted state space and the adjusted action space, the updated feng shui linkage equipment optimization control model is used to generate an optimized control strategy, obtain a real-time optimized control strategy, and send the real-time optimized control strategy to the corresponding feng shui linkage equipment. The updated fault diagnosis model for wind and water linkage equipment generates real-time experience during the processing of real-time multimodal monitoring data. The real-time experience, real-time multimodal monitoring data, real-time fault diagnosis results, and real-time optimized control strategies are then uploaded to the cloud data center.

9. A method for diagnosing faults in a wind-water linkage device according to claim 8, characterized in that: In the cloud data center, based on a continuous learning mechanism, the updated fault diagnosis model for feng shui linkage equipment is continuously trained to obtain a continuously trained fault diagnosis model for feng shui linkage equipment. This continuously trained fault diagnosis model is then synchronized to all edge computing gateways, including the following steps: The cloud data center provides distributed storage for real-time multimodal monitoring data, real-time fault diagnosis results, and real-time optimization control strategies for each wind and water linkage device. Several second historical experiences are randomly selected from the large model experience replay pool, and these second historical experiences are mixed with real-time experiences to obtain several mixed experiences. Based on several hybrid experiences, the updated large-scale model for fault diagnosis of wind and water linkage equipment is continuously trained, and a comprehensive loss function is used to generate the second historical loss value in each continuous training process. If the number of training iterations reaches the threshold or the second historical loss value is lower than the loss value threshold, a large model for fault diagnosis of the wind and water linkage equipment after continuous training is obtained, and the parameters of the large model after continuous training are extracted. After continuous training, the parameters of the large model are sent to all edge computing gateways. At the edge computing gateways, the large model is updated based on the parameters after continuous training, thus completing the synchronization of the large model for fault diagnosis of wind and water linkage equipment after continuous training.

10. A fault diagnosis system for feng shui linkage equipment, used to implement the fault diagnosis method for feng shui linkage equipment as described in any one of claims 1-9, characterized in that: The system includes a cloud data center, several edge computing gateways, several monitoring data acquisition devices, and several feng shui linkage devices. The cloud data center is connected to several edge computing gateways, and both the cloud data center and each edge computing gateway are equipped with a large-scale fault diagnosis model for the feng shui linkage devices. Each edge computing gateway is communicatively connected to the corresponding monitoring data acquisition device and the feng shui linkage device, and each monitoring data acquisition device is communicatively connected to the corresponding feng shui linkage device.

Images