Wind power plant equipment fault intelligent diagnosis method and system

Abstract

The invention discloses a wind power plant equipment fault intelligent diagnosis method and system. The method comprises the following steps: acquiring multi-dimensional sensor data of wind power plant equipment; carrying out preliminary screening on the data by utilizing an edge analysis algorithm, and determining data needing to be further diagnosed; the screened data are input into a pre-constructed fault diagnosis model, the data are fused through a multi-modal deep learning model in the pre-constructed fault diagnosis model, and then a fault occurrence area is determined through transfer learning, domain adaptation and a multi-agent reinforcement learning framework; meanwhile, the causal inference engine and the knowledge graph enhanced neural symbol inference module are responsible for deep processing and fault diagnosis. And finally, inputting the determined fault area and the diagnosis result into a dynamic system health index calculation module to generate a final fault diagnosis result. According to the method, the diagnosis efficiency and accuracy of the equipment fault of the wind power plant are remarkably improved, and a powerful guarantee is provided for stable operation of the wind power plant.

Description

Technical Field

[0001] The invention belongs to the technical field of wind power generation, and relates to an intelligent diagnosis method and system for wind farm equipment faults. Background Art

[0002] In the field of renewable energy, wind power generation is becoming increasingly important as a clean and renewable energy form. With the rapid development of wind power technology and the continuous expansion of wind farms, how to ensure the efficient and stable operation of wind farms has become an urgent problem to be solved. The complexity of wind farm equipment, especially wind turbines, and the diversity of operating environments have put forward higher requirements for fault diagnosis technology.

[0003] Traditionally, wind farm equipment fault diagnosis mainly relies on manual inspections and simple threshold judgment methods based on a single type of sensor data. This method is not only inefficient, but also difficult to cope with the increasingly complex failure modes and changing operating environments of modern wind power equipment. The limitation of single sensor data is that it can only provide status information of a certain aspect of the equipment, but cannot fully reflect the overall health of the equipment. At the same time, simple threshold judgment methods often cannot accurately capture the early signs of equipment failure, resulting in delayed fault detection, increased downtime and maintenance costs. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problems in the prior art that traditional fault diagnosis methods have single data, low efficiency, and cannot accurately capture equipment failures in real time, and are difficult to meet the needs of modern wind farms, and to provide a wind farm equipment fault intelligent diagnosis method and system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A first aspect of the present invention provides a method for intelligent diagnosis of wind farm equipment faults, comprising the following steps: Collect multi-dimensional sensor data from wind farm equipment; Performing a preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; Input the multi-dimensional sensor data that needs further diagnosis into the pre-built fault diagnosis model to obtain the fault diagnosis results; Among them, the pre-built fault diagnosis model includes a multimodal deep learning model, a transfer learning and domain adaptation module, a multi-agent reinforcement learning framework, a causal reasoning engine, a knowledge graph enhanced neural symbolic reasoning module and a dynamic system health index calculation module; the multi-dimensional sensor data that needs further diagnosis is fused through the multimodal deep learning model, and the area is determined through transfer learning and domain adaptation and multi-agent reinforcement learning framework, and the diagnosis fault is processed through the causal reasoning engine and the knowledge graph enhanced neural symbolic reasoning module; the determined area and the diagnosed fault are input into the dynamic system health index calculation module to obtain the fault diagnosis result.

[0006] Further, the multi-dimensional sensor data is preliminarily analyzed using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis, specifically: Preprocess multi-dimensional sensor data; An edge analysis algorithm is used to detect abnormal values ​​of the preprocessed multi-dimensional sensor data; the abnormal values ​​are the multi-dimensional sensor data that need further diagnosis.

[0007] Furthermore, the multimodal deep learning model includes a spatiotemporal attention mechanism and a multi-scale causal convolutional network.

[0008] Furthermore, the transfer learning and domain adaptation module adopts a dynamic weight alignment method and an adversarial domain adaptation method.

[0009] Furthermore, the multi-agent reinforcement learning framework adopts a multi-agent soft actor-critic algorithm.

[0010] Furthermore, the causal reasoning engine adopts an improved PC algorithm based on a Bayesian network and a method of integrating prior probabilities based on expert knowledge.

[0011] Furthermore, the knowledge graph enhanced neural symbolic reasoning module includes a knowledge graph embedding module and a neural symbolic reasoning module; the knowledge graph embedding module adopts the TransE model; and the neural symbolic reasoning module adopts a method combining symbolic rules with neural networks.

[0012] A second aspect of the present invention provides a wind farm equipment fault intelligent diagnosis system, comprising: The data collection layer is used to collect multi-dimensional sensor data of wind farm equipment; The edge computing layer is used to perform preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; The cloud analysis layer is used to input the multi-dimensional sensor data that needs further diagnosis into the pre-built fault diagnosis model to obtain the fault diagnosis results.

[0013] Furthermore, the wind farm equipment fault intelligent diagnosis system also includes a network transmission layer, which is used to realize data transmission between the edge computing layer and the cloud analysis layer.

[0014] Furthermore, the wind farm equipment fault intelligent diagnosis system also includes an application interface layer, which receives data from the cloud analysis layer and is used to provide diagnosis results and an interactive interface to users.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses an intelligent fault diagnosis method for wind farm equipment. By collecting multi-dimensional sensor data of wind farm equipment, the method can comprehensively capture the operating status information of the equipment, including but not limited to key parameters such as vibration, temperature, current, and voltage, thereby providing a rich and accurate data basis for subsequent fault diagnosis. An edge analysis algorithm is used to perform preliminary analysis on massive sensor data to quickly identify data points with abnormalities or potential faults, thereby effectively reducing the burden of data transmission and processing and improving the response speed and accuracy of the diagnostic system. A multimodal deep learning model is used to deeply fuse multi-dimensional data from different sensors, thereby mining the potential correlations and features between the data and improving the accuracy and robustness of fault diagnosis. At the same time, combined with the transfer learning and domain adaptation modules, it can effectively utilize historical data and cross-domain knowledge, quickly adapt to the characteristics of different wind farms and equipment, and further improve the generalization ability of the diagnostic model; the introduction of the multi-agent reinforcement learning framework simulates the collaborative decision-making process of multiple agents in a complex environment, can more accurately locate the area where the fault occurs, and improve the efficiency and accuracy of fault location by continuously optimizing strategies; combined with the causal reasoning engine and the neural symbolic reasoning module enhanced by the knowledge graph, it can not only accurately classify and identify faults, but also reveal the root cause and propagation path of the fault, providing a scientific basis for fault prevention and maintenance. The introduction of the knowledge graph further enriches the knowledge base of the diagnostic system and improves the intelligence level and self-learning ability of the system; through the dynamic system health index calculation module, the health status of wind farm equipment is evaluated and predicted in real time, providing intuitive health indicators and early warning information for operation and maintenance personnel, which helps to discover potential faults in advance and take corresponding maintenance measures, reducing the impact of equipment failures on wind farm operations.

[0016] Furthermore, the edge analysis algorithm performs preliminary screening at the source of the data, and only sensor data that is marked as abnormal will be sent to the central server or cloud platform for further diagnosis. This approach significantly reduces the amount of data that needs to be transmitted, reduces the consumption of network bandwidth, and also speeds up the response to faults. Since the edge analysis algorithm runs directly on the edge devices where the data is generated (such as fan control cabinets, sensor nodes, etc.), it can process the sensor data and feedback the results in the first place. This instant feedback mechanism enables the system to respond to changes in equipment status more quickly, improving the timeliness of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 This is a block diagram of the intelligent diagnosis method for wind farm equipment faults of the present invention; Figure 2 This is a block diagram of the wind farm equipment fault intelligent diagnosis system of the present invention; Figure 3 This is a system architecture diagram of an embodiment of a wind farm equipment fault intelligent diagnosis system of the present invention.

[0019] Among them: 201-data collection layer; 202-edge computing layer; 203-cloud analysis layer. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention described and marked in the drawings here can be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0023] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0024] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0025] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0026] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 1 The present invention provides a method for intelligent diagnosis of wind farm equipment faults, comprising the following steps: S1, collects multi-dimensional sensor data of wind farm equipment; High-precision sensors are installed at key parts of each wind turbine (such as generators, gearboxes, bearings, etc.). These sensors can monitor and collect multi-dimensional data such as temperature, vibration, speed, current, etc. in real time. The collected data is transmitted to the data center or edge computing node of the wind farm via wired or wireless means.

[0027] S2, performing preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; Edge analysis algorithms are deployed on edge computing nodes to pre-process the received multi-dimensional sensor data, including data cleaning, denoising, normalization and other operations.

[0028] Edge analysis algorithms use statistical or machine learning-based methods (such as threshold detection, isolation forest, etc.) to analyze preprocessed data in real time to detect whether there are outliers or abnormal patterns.

[0029] The detected abnormal data (i.e., multi-dimensional sensor data that requires further diagnosis) is marked and prepared to be sent to the central server for in-depth analysis.

[0030] S3, input the multi-dimensional sensor data that needs further diagnosis into a pre-built fault diagnosis model to obtain a fault diagnosis result; the pre-built fault diagnosis model includes a multimodal deep learning model, a transfer learning and domain adaptation module, a multi-agent reinforcement learning framework, a causal reasoning engine, a knowledge graph-enhanced neural symbolic reasoning module and a dynamic system health index calculation module.

[0031] On the central server, a comprehensive fault diagnosis model is pre-built and trained, which integrates a multimodal deep learning model, a transfer learning and domain adaptation module, a multi-agent reinforcement learning framework, a causal reasoning engine, a knowledge graph-enhanced neural symbolic reasoning module, and a dynamic system health index calculation module.

[0032] The abnormal data transmitted from the edge computing node is input into the model. The model first uses a multimodal deep learning model to extract and fuse the data features to capture the correlation and complementary information between different sensors.

[0033] Next, the transfer learning and domain adaptation module adjusts the model's parameters and decision boundaries based on historical fault data and current equipment status to adapt to changes caused by different operating conditions and equipment aging.

[0034] The multi-agent reinforcement learning framework simulates the collaboration and competition among multiple agents, optimizes the fault diagnosis strategy through continuous trial and error and learning, and improves the accuracy and efficiency of diagnosis.

[0035] The causal reasoning engine and the knowledge graph-enhanced neural symbolic reasoning module combine domain knowledge and expert experience to conduct in-depth analysis and reasoning on the causes of faults and construct fault chains and fault trees.

[0036] Finally, the dynamic system health index calculation module calculates the health index of the equipment based on the diagnostic results of the model and the real-time status of the equipment, providing an intuitive basis for equipment health assessment for operation and maintenance personnel.

[0037] An embodiment of the present invention provides a method for intelligent diagnosis of wind farm equipment faults, comprising the following steps: Collect multi-dimensional sensor data from wind farm equipment; install high-precision sensors at key locations of wind farm equipment to collect multi-dimensional data in real time, and transmit the data to edge computing nodes via wired or wireless means.

[0038] The multi-dimensional sensor data is preliminarily analyzed using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; after the edge computing node receives the data, it first preprocesses the data, including denoising, compression, and normalization, and then applies an edge analysis algorithm (such as statistical-based anomaly detection, machine learning models, etc.) to perform a preliminary analysis on the data to identify abnormal data; the abnormal data is marked and transmitted to the central server.

[0039] The multi-dimensional sensor data that needs to be further diagnosed is input into a pre-built fault diagnosis model to obtain a fault diagnosis result; the pre-built fault diagnosis model includes: Multimodal deep learning model: The spatiotemporal attention mechanism is used to capture the temporal and spatial features in the data, and the multi-scale causal convolutional network is combined to extract fault features of different scales.

[0040] Transfer learning and domain adaptation module: Dynamic weight alignment method and adversarial domain adaptation method are used to enable the model to adapt to changes in data distribution caused by different working conditions and equipment aging.

[0041] Multi-agent reinforcement learning framework: Through the multi-agent soft actor-critic algorithm, the collaboration and competition between multiple agents are simulated to optimize the fault diagnosis strategy.

[0042] Causal reasoning engine: Based on the Bayesian network, it adopts the improved PC algorithm, combines the prior probability of expert knowledge, builds fault chains and fault trees, and deeply analyzes the causes of faults.

[0043] Knowledge graph-enhanced neural symbolic reasoning module: Knowledge graph embedding module: The TransE model is used to embed entities and relations in the knowledge graph into a low-dimensional vector space.

[0044] Neural symbolic reasoning module: combines symbolic rules and neural networks to enhance the model's reasoning ability by leveraging structured information in the knowledge graph.

[0045] Dynamic system health index calculation module: Calculates the health index of the equipment based on diagnostic results and real-time data, providing intuitive equipment health assessment for operation and maintenance personnel.

[0046] See also Figure 2 An embodiment of the present invention provides a wind farm equipment fault intelligent diagnosis system, comprising: The data collection layer is used to collect multi-dimensional sensor data of wind farm equipment; The edge computing layer is used to perform preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; The cloud analysis layer is used to input the multi-dimensional sensor data that needs further diagnosis into the pre-built fault diagnosis model to obtain the fault diagnosis results.

[0047] See also Figure 3 An embodiment of the present invention provides a wind farm equipment fault intelligent diagnosis system, comprising: Data collection layer, edge computing layer, network transmission layer, cloud analysis layer and application interface layer.

[0048] Data collection layer: used to collect multi-dimensional sensor data of wind farm equipment. The data collection layer is the perception basis of the system and is responsible for comprehensively and accurately collecting the operating data of wind power equipment. It mainly includes the following components: Multi-dimensional sensor network, including: Nano-MEMS sensor array: Arranged in key parts such as blades and hubs to collect data on microstructure changes. Highly sensitive triaxial accelerometers and strain sensors are used, with a size of approximately 50μm x 50μm x 10μm and a sensitivity of 0.1μg / √Hz.

[0049] Fiber Bragg grating sensor network: deployed along the wind turbine tower and blades to achieve distributed strain and temperature monitoring. Using FBG sensors with a wavelength range of 1510nm-1590nm, the temperature sensitivity is about 10pm / ℃ and the strain sensitivity is about 1.2pm / με.

[0050] Acoustic sensor array: Installed around the cabin to capture abnormal sound signals. It uses a MEMS microphone array with a frequency response range of 20Hz-20kHz and a signal-to-noise ratio greater than 65dB.

[0051] High-precision vibration sensor: Installed in gearboxes, generators and other parts to monitor vibration status. Use piezoelectric accelerometer with a frequency range of 0.5Hz-10kHz and a sensitivity of 100mV / g.

[0052] Oil monitoring sensor: used to analyze the gearbox oil status in real time. It uses a capacitive particle counter and can detect particles in the range of 4μm-100μm.

[0053] Adaptive data acquisition system for collecting data from multi-dimensional sensor networks. Dynamic sampling rate adjustment: Automatically adjust the sampling frequency according to the device status and environmental conditions, ranging from 1kHz to 100kHz. Data compression: Adopts a real-time data compression algorithm based on compressed sensing, with a compression ratio of up to 10:1. Local cache: Each acquisition node is equipped with a 128GB high-speed SSD, which supports 72 hours of cyclic storage of raw data.

[0054] The data preprocessing unit improves data quality by performing signal amplification, filtering, and other processing. Sensor calibration is automatically performed once a week to ensure data accuracy. Simple statistical methods are used for preliminary anomaly detection to mark potential problem data.

[0055] The edge computing layer is used to pre-process and preliminarily analyze the multi-dimensional sensor data; the edge computing layer performs preliminary processing and analysis at the source of the data, reducing the burden on the cloud and improving the real-time performance of the system. It mainly includes: Edge computing hardware platform, the edge computing hardware platform processor uses Intel Xeon D-2183IT (16 cores, 32 threads); GPU is NVIDIA Tesla T4 (16GB GDDR6) for AI reasoning acceleration; FPGA is Xilinx Artix-7 XC7A200T for programmable hardware acceleration. Memory is 128GB DDR4-3200. Storage is 2TB NVMe SSD.

[0056] Edge intelligence software stack, including real-time operating system using RT-Linux with PREEMPT_RT patch. Containerized environment using lightweight Kubernetes based on K3s. Time series database: InfluxDB Edge is deployed. Stream processing engine uses Apache Flink. Edge AI framework uses NVIDIA Triton Inference Server.

[0057] The edge analysis algorithm consists of the following steps: a) Signal processing: Adaptive filtering: Adaptive filter based on the least mean square error (LMS) algorithm.

[0058] Time-frequency analysis: Short-time Fourier transform (STFT) and wavelet packet decomposition (WPD).

[0059] b) Feature extraction: Calculate statistical features, spectral features and time-frequency features.

[0060] c) Lightweight Machine Learning Model: Anomaly detection: Local outlier factor (LOF) and isolation forest algorithm are used.

[0061] Fault classification: Using quantized decision tree and support vector machine models.

[0062] The network transport layer is used to realize data transmission between the edge computing layer and the cloud analysis layer; the network transport layer ensures reliable, secure and efficient data transmission between the edge computing layer and the cloud analysis layer. The main features include: The primary communication link uses 5G NR technology, using the 3.5GHz (n78) frequency band, 100MHz bandwidth, and 64T64R MIMO configuration. The backup link uses low-orbit satellite communications (such as SpaceX Starlink) with an average latency of 20-40ms.

[0063] Network security mechanisms include: Data encryption uses TLS 1.3 protocol and supports post-quantum cryptography (PQC) algorithm. Identity authentication is based on two-way authentication of X.509 certificates. Suricata IDPS is deployed for intrusion detection to monitor network traffic in real time. Software-defined network (SDN) technology is used to implement flexible routing strategies. Machine learning algorithms are used to predict network congestion and dynamically adjust routes. Adaptive data compression dynamically adjusts the compression level of the Zstandard compression algorithm according to network bandwidth. Local cache mechanism: Redis is used as a high-speed cache to temporarily store data when the network is unstable.

[0064] The cloud analysis layer is used to conduct in-depth analysis and fault diagnosis of the transmitted data; the cloud analysis layer is the core of the system, responsible for complex data analysis and decision support. It mainly includes the following modules: (1) Multimodal deep learning model. The multimodal deep learning model mainly includes: a) Spatiotemporal attention mechanism: used to capture the spatiotemporal dependencies between wind turbine components. The core formula is as follows:

[0065] in, is the final spatiotemporal fusion feature, is an adaptive parameter, and They are temporal and spatial attention weights, and are the temporal and spatial feature sequences respectively.

[0066] b) Multi-scale Causal Convolutional Network: Use causal convolutional layers with different dilation rates to capture failure modes at different time scales.

[0067] (2) Transfer learning and domain adaptation module: Dynamic Weight Alignment: Minimize the difference between source and target domain model weights.

[0068] Adversarial domain adaptation: training a domain discriminator to reduce the difference in feature distribution between the source and target domains.

[0069] (3) Multi-agent reinforcement learning framework: The Multi-Agent Soft Actor-Critic (MASAC) algorithm is used to model each wind turbine as an agent to optimize the diagnosis and control strategy of the entire wind farm.

[0070] (4) Causal Inference Engine: Based on Bayesian network, the interpretability of fault diagnosis is improved. The improved PC algorithm is used for structure learning, and expert knowledge is introduced as prior.

[0071] (5) Knowledge graph-enhanced neural symbolic reasoning module: Combining the knowledge graph embedding of the TransE model and the symbolic reasoning of the neural network. The core formula is as follows:

[0072] in, is the output result, is the activation function, is the weight matrix, is the entity embedding, represents the symbolic reasoning result based on the knowledge graph $KG$, is the bias term.

[0073] (6) Dynamic system health index calculation module: Provides comprehensive evaluation indicators for the health status of the fan. The calculation formula is as follows:

[0074] in, It's time The dynamic system health index, is the number of health indicators, It is The dynamic weight of the indicators, is the normalization function, It is Health indicators in time The value of .

[0075] The application interface layer is used to provide users with diagnostic results and interactive interfaces; the application interface layer provides a friendly and intuitive interactive interface for different types of users. It mainly includes: (1) Web application: Front-end framework: React 18 with Next.js. State management: ReduxToolkit. UI component library: Material-UI. Data visualization: D3.js and ECharts.

[0076] (2) Mobile application: Development framework: Flutter 3.0. State management: Riverpod. Local storage: Hive database.

[0077] (3) API Gateway: The core is based on the Envoy service mesh. Authentication uses OAuth 2.0 + OpenIDConnect. Current limiting implements adaptive current limiting based on the token bucket algorithm. The document uses the OpenAPI 3.0 specification and is displayed through the Swagger UI.

[0078] (4) Mixed reality interactive system: AR device uses Microsoft HoloLens 2. Development framework uses Unity+ Mixed Reality Toolkit (MRTK). 3D modeling: Autodesk Inventor is used to create a high-precision wind turbine model. Real-time data stream: Real-time sensor data is transmitted through the WebSocket protocol. Interaction method: Voice commands: Natural language understanding based on Azure Speech Services.

[0079] Gesture control: Use the built-in gesture recognition provided by MRTK.

[0080] Virtual touch: 3D UI interaction based on ray casting.

[0081] The interaction and connection relationship between the various layers of the system in this embodiment are as follows: the communication method between the data acquisition layer and the edge computing layer adopts the EtherCAT real-time industrial Ethernet protocol; the sensor data is transmitted to the edge computing node in real time; the edge computing layer can send control instructions such as sampling rate adjustment to the data acquisition layer. The edge computing layer and the network transmission layer use the MQTT protocol for data transmission and implement TLS encryption and two-way authentication, and perform adaptive data compression before transmission. The main link between the network transmission layer and the cloud analysis layer uses the 5G network to transmit data; the backup link automatically switches to satellite communication when the main link fails; the communication mode adopts the publish-subscribe mode to ensure the real-time and reliability of the data. The cloud analysis layer and the application interface layer provide RESTful API and GraphQL interface; the data format uses JSON for data exchange; for data that needs to be updated in real time, the WebSocket protocol is used. The overall interaction process is that the collected data is analyzed by the data acquisition layer and the edge computing layer, and then transmitted from the network transmission layer to the cloud analysis layer for further detection and analysis, and then transmitted to the application interface layer for display to the user. New models trained in the cloud are regularly sent to the edge computing layer.

[0082] The system workflow of this embodiment is as follows: First, the data acquisition layer continuously collects wind turbine operation data through a multi-dimensional sensor network. Then the edge computing layer preprocesses and preliminarily analyzes the raw data. The processed data is sent to the cloud through the network transmission layer. The cloud analysis layer performs in-depth analysis and fault diagnosis. The cloud analysis layer is the core part of the system. It uses a variety of advanced technologies such as multimodal deep learning models, transfer learning and domain adaptation modules, multi-agent reinforcement learning frameworks, causal reasoning engines, knowledge graph-enhanced neural symbolic reasoning modules, and dynamic system health index calculation modules to perform in-depth analysis and fault diagnosis on data from the edge computing layer. The multimodal deep learning model is able to process multi-source heterogeneous data from different sensors and different types, and extract deep feature representations. The transfer learning and domain adaptation module realizes knowledge transfer across wind farms and across devices; by utilizing the fault diagnosis knowledge of existing wind farms or equipment, it can quickly adapt to the diagnosis needs of new wind farms or new equipment, reducing dependence on new data and training time. The multi-agent reinforcement learning framework optimizes the diagnostic strategy of the entire wind farm; simulates the collaborative work of multiple agents in the wind farm, finds the optimal diagnostic strategy through continuous trial and error and optimization, and improves the operation and maintenance efficiency and economic benefits of the entire wind farm. The causal reasoning engine provides explainable diagnostic results; by providing explainable diagnostic results through hierarchical association of fault causes, it helps operation and maintenance personnel understand the root cause and propagation path of the fault, and provides strong support for subsequent fault repair and prevention. The knowledge graph-enhanced neural symbolic reasoning module integrates domain knowledge and data-driven methods; by building a knowledge graph in the field of wind turbines, the reasoning ability and interpretability of the system are enhanced. The dynamic system health index calculation module dynamically evaluates the overall health status of the wind turbine based on real-time and historical data, providing timely and accurate reference for operation and maintenance decisions. The diagnostic results are presented to the user through the application interface layer. The diagnostic results obtained after processing by the cloud analysis layer are presented to the user in an intuitive and friendly way through the application interface layer. These results may include fault type, fault location, fault severity, maintenance recommendations, and health status assessment reports. Users can access this information through the web interface, mobile application, or API interface, and make further operations or decisions as needed. At the same time, the system also provides functions such as historical data query and trend analysis to help users fully understand the operating status and performance changes of the fan.

[0083] The cloud analysis layer system of the present invention adopts an innovative multimodal deep learning model to effectively integrate heterogeneous data from different sensors. The multimodal deep learning model of the present invention introduces a spatiotemporal attention mechanism and a multi-scale causal convolutional network; the spatiotemporal attention mechanism can adaptively capture the complex spatiotemporal dependencies between the various components of the wind turbine. The specific implementation is as follows: Temporal Attention: A Bidirectional Long Short-Term Memory (BiLSTM) network is used to calculate the attention weights in the temporal dimension.

[0084] Spatial Attention: Graph Attention Network (GAT) is used to capture the spatial dependencies between components.

[0085] Fusion strategy: Dynamically regulating the importance of temporal and spatial attention through a gating mechanism.

[0086] Multi-scale causal convolutional networks can handle failure modes at different time scales, including: Multi-scale receptive field: Use dilated convolution with different dilation rates to achieve multi-scale receptive field.

[0087] Causal convolution: ensure the temporal causality of the model and avoid information leakage.

[0088] Residual Connections: Use residual connections and layer normalization to improve training stability.

[0089] In order to solve the problem of knowledge transfer between different wind farms and different types of wind turbines, the system transfers learning and domain adaptation technology, specifically: adopting a dynamic weight alignment method, designing a learnable mapping matrix, and minimizing the difference in model weights between the source domain and the target domain. Regularization technology prevents overfitting, such as L2 regularization and dropout. Adopting an adversarial domain adaptation method: introducing a domain discriminator, and reducing the difference in feature distribution between the source domain and the target domain through adversarial training. Using a gradient reversal layer (GradientReversal Layer) to implement adversarial training.

[0090] In order to further solve the accuracy of the fault source domain of the wind farm, the present invention adopts a multi-agent reinforcement learning framework; the multi-agent reinforcement learning framework of the present invention adopts the multi-agent soft actor-critic (MASAC) algorithm to model the entire wind farm as a multi-agent system: its state space includes local observations and state information of neighboring wind turbines. The action space includes continuous actions such as diagnostic threshold adjustment and feature selection. Reward function: comprehensively considers diagnostic accuracy, false alarm rate, missed alarm rate and maintenance cost. Policy network: a 3-layer fully connected network is used, and the hidden layer dimension is [512, 256]. Value network: a 3-layer fully connected network is also used, and the hidden layer dimension is [512, 256]. Optimization algorithm: use Adam optimizer, the learning rate is set to 1e-4, and the batch size is 256.

[0091] In order to improve the interpretability of fault diagnosis, a causal reasoning engine is used to infer the causal relationship between one fault and the other. The causal reasoning engine based on the Bayesian network of the present invention improves the interpretability of fault diagnosis. The specific method is as follows: Structural learning: An improved PC algorithm is used to introduce expert knowledge as a priori.

[0092] Parameter learning: The conditional probability table is estimated using the expectation-maximization (EM) algorithm.

[0093] Inference algorithm: Combine variable elimination and importance sampling to improve inference efficiency.

[0094] Causal Effect Estimation: Calculate the Average Causal Effect (ACE) using do-calculus.

[0095] In order to further improve the interpretability of fault diagnosis, the present invention adopts a knowledge graph-enhanced neural symbolic reasoning module, which combines the advantages of symbolic reasoning and neural networks, specifically: Knowledge graph construction: Use OWL 2 Web Ontology Language to define domain ontology.

[0096] Graph storage: Neo4j graph database is used to store RDF triples.

[0097] Knowledge Graph Embedding: Use the TransE model to map entities and relations into a low-dimensional vector space.

[0098] Neural Symbolic Reasoning: Designing end-to-end differentiable reasoning networks combining symbolic rules and neural networks.

[0099] The Dynamic System Health Index (DSHI) of the present invention provides a comprehensive assessment of the health status of the wind turbine: Multidimensional health indicators include vibration, temperature, oil status, etc. These indicators can fully reflect the operating status of the system. The min-max normalization method is used to unify indicator data of different dimensions and magnitudes to the same scale, which is convenient for subsequent data processing and comparison. The importance of each indicator is dynamically adjusted using the attention mechanism. This method can automatically adjust the weight of each health indicator in the comprehensive evaluation according to the current state and operating environment of the system, thereby improving the accuracy and robustness of the evaluation. Time series analysis methods such as LSTM (Long Short-Term Memory Network) are used to capture the time evolution trend of the system health status. The LSTM network can handle long-term dependencies and has good modeling capabilities for health data with time series characteristics.

[0100] In the process of implementing the diagnostic system of the present invention, the following performance indicators are achieved through innovative technical solutions: Fault prediction lead time: The average 7-14 days of fault prediction lead time provides operation and maintenance personnel with sufficient time to conduct fault troubleshooting and maintenance preparation; Diagnostic accuracy: For known fault modes, the diagnostic accuracy exceeds 95%, ensuring the accuracy and reliability of fault identification; False alarm rate: The false alarm rate is less than 1%; Missing reporting rate: The missing reporting rate is less than 0.5%; System response time: The edge response time is less than 100ms, and the cloud complex analysis response time is less than 1s, ensuring that the system can quickly respond to various operation and maintenance needs; Data compression rate: The data compression rate exceeds 90%, effectively reducing the cost of data transmission and storage; System availability: The system availability reaches 99.999% (five 9s), ensuring the continuous and stable operation of the system; Security: Complies with industrial control system security standards such as IEC 62443, ensuring the information security and data security of the system.

[0101] The wind farm equipment fault intelligent diagnosis system of the present invention is deployed by using an automated deployment script based on Ansible; GitOps workflow is used in combination with ArgoCD to achieve continuous deployment.

[0102] In order to ensure the stable operation of the intelligent wind power operation and maintenance system and timely discover and deal with potential problems, the present invention uses Prometheus + Grafana to build a full-stack monitoring system and set multi-level alarm thresholds. Deploy the Prometheus service in the system to collect monitoring data such as system operation status and performance indicators; configure the monitoring targets of Prometheus, including key system components, service interfaces, etc.; install and configure Grafana as a visualization platform for monitoring data; create a dashboard to display key system indicators and operation status, such as CPU usage, memory usage, response time, etc. Define alarm rules in Prometheus, set multi-level alarm thresholds, such as warning, severe, etc. Configure alarm notification channels, including SMS, email, and mobile application push.

[0103] In order to maintain the advancement and stability of the intelligent wind power operation and maintenance system, the present invention adopts blue-green deployment and canary release strategies to update the system.

[0104] Blue-Green Deployment: Prepare a new environment (blue environment) in the production environment that is completely isolated from the current running system.

[0105] Deploy the new version to the blue environment and perform comprehensive functional and performance testing.

[0106] After the test passes, traffic is switched to the blue environment to achieve zero downtime updates.

[0107] Canary Release: Select a small number of users or devices in the blue environment as canary users (or devices).

[0108] Gradually switch canary users to the new version, observe its operation and collect feedback.

[0109] Adjust the new version or optimize the system configuration based on feedback to ensure that the new version is stable and reliable before full promotion.

[0110] To ensure high availability and data security of the intelligent wind power operation and maintenance system, the present invention implements a remote multi-active architecture and regularly conducts disaster recovery drills: Implement a multi-site active-active architecture to ensure high availability of the system: deploy system replicas in data centers in different geographical locations to form a multi-site active-active architecture; configure a data synchronization mechanism to ensure data consistency and real-time performance between data centers; set up load balancing and failover strategies to ensure automatic switching to other data centers when a single data center fails.

[0111] Conduct disaster recovery drills regularly to verify the effectiveness of the backup and recovery process: Develop a detailed disaster recovery plan, including the recovery process, required resources, expected time, etc.; conduct disaster recovery drills regularly to simulate real disaster scenarios and verify the effectiveness of the backup and recovery process; record problems and improvement points during the drills and continuously optimize the disaster recovery plan.

[0112] Backup strategy optimization: formulate appropriate backup strategies based on system characteristics and business needs; regularly back up key system data and store it in a safe and reliable location; verify the integrity and recoverability of backup data to ensure that the system status can be quickly restored when needed.

[0113] The wind farm equipment fault intelligent diagnosis system of the present invention is designed with a flexible architecture, which reserves expansion space for future technological development and business needs: interfaces are reserved for integration with other renewable energy systems such as photovoltaics and energy storage. Blockchain interfaces are reserved: to prepare for future energy trading and carbon credit management. The potential of future quantum computing is considered in the algorithm design. An interface is reserved for the continuous enhancement of edge AI capabilities.

[0114] In summary, the wind farm equipment fault intelligent diagnosis system provided by the present invention effectively solves the problems existing in the prior art such as insufficient diagnostic accuracy, poor interpretability, and difficulty in adapting to different wind farm environments through the comprehensive application of advanced technologies such as multimodal data fusion, edge computing, deep learning, transfer learning, reinforcement learning, and causal reasoning. The modular design and advanced learning algorithms of the system make it highly adaptable and extensible, and it can continuously absorb new data and knowledge and continuously improve its diagnostic capabilities. Through the application of this system, the operating efficiency of wind farms can be significantly improved, maintenance costs can be reduced, and strong support can be provided for the sustainable development of the wind power industry.

[0115] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A wind farm equipment fault intelligent diagnosis method, characterized in that: The following steps are involved: Collect multi-dimensional sensor data from wind farm equipment; Performing a preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; Input the multi-dimensional sensor data that needs further diagnosis into the pre-built fault diagnosis model to obtain the fault diagnosis results; Among them, the pre-built fault diagnosis model includes a multimodal deep learning model, a transfer learning and domain adaptation module, a multi-agent reinforcement learning framework, a causal reasoning engine, a knowledge graph enhanced neural symbolic reasoning module and a dynamic system health index calculation module; the multi-dimensional sensor data that needs further diagnosis is fused through the multimodal deep learning model, and the area is determined through transfer learning and domain adaptation and multi-agent reinforcement learning framework, and the diagnosis fault is processed through the causal reasoning engine and the knowledge graph enhanced neural symbolic reasoning module; the determined area and the diagnosed fault are input into the dynamic system health index calculation module to obtain the fault diagnosis result.

2. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The edge analysis algorithm is used to perform a preliminary analysis on the multi-dimensional sensor data to determine the multi-dimensional sensor data that needs further diagnosis, specifically: Preprocess multi-dimensional sensor data; An edge analysis algorithm is used to detect abnormal values ​​of the preprocessed multi-dimensional sensor data; the abnormal values ​​are the multi-dimensional sensor data that need further diagnosis.

3. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The multimodal deep learning model includes a spatiotemporal attention mechanism and a multi-scale causal convolutional network.

4. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The transfer learning and domain adaptation module adopts a dynamic weight alignment method and an adversarial domain adaptation method.

5. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The multi-agent reinforcement learning framework adopts a multi-agent soft actor-critic algorithm.

6. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The causal reasoning engine adopts an improved PC algorithm based on a Bayesian network and a method of integrating prior probabilities based on expert knowledge.

7. The intelligent diagnosis method for wind farm equipment faults according to claim 1 is characterized in that: The knowledge graph enhanced neural symbolic reasoning module includes a knowledge graph embedding module and a neural symbolic reasoning module; the knowledge graph embedding module adopts the TransE model; the neural symbolic reasoning module adopts a method combining symbolic rules with neural networks.

8. An intelligent fault diagnosis system for wind farm equipment, characterized in that: include: The data collection layer is used to collect multi-dimensional sensor data of wind farm equipment; The edge computing layer is used to perform preliminary analysis on the multi-dimensional sensor data using an edge analysis algorithm to determine the multi-dimensional sensor data that needs further diagnosis; The cloud analysis layer is used to input the multi-dimensional sensor data that needs further diagnosis into the pre-built fault diagnosis model to obtain the fault diagnosis results.

9. The wind farm equipment fault intelligent diagnosis system according to claim 8, characterized in that: The wind farm equipment fault intelligent diagnosis system also includes a network transmission layer, which is used to realize data transmission between the edge computing layer and the cloud analysis layer.

10. The wind farm equipment fault intelligent diagnosis system according to claim 9, characterized in that: The wind farm equipment fault intelligent diagnosis system also includes an application interface layer, which receives data from the cloud analysis layer and is used to provide diagnosis results and an interactive interface to users.

Images