Wind power generation system fault diagnosis method and device based on multi-source data fusion and deep learning

Abstract

The invention discloses a wind power generation system fault diagnosis method and device based on multi-source data fusion and deep learning. The method comprises the following steps: collecting vibration, electrical, environment and operation state data of the wind generating set through a sensor array; performing preprocessing and depth feature extraction on the data, and fusing the data into a global feature vector; the global feature vector is input into a hybrid deep learning model for fault diagnosis, the model integrates CNN and LSTM, and a diagnosis threshold can be dynamically adjusted according to the wind speed; when a sensor fault is diagnosed, a fault reconstruction module based on a T-S fuzzy sliding mode observer is started, and fault information is accurately estimated; and feeding the reconstructed fault-free signal into a model prediction fault-tolerant controller to realize active fault-tolerant control of the system and maintain the stability of the output power. The problems that an existing fault diagnosis method is single in view angle, depends on an accurate model, and is disjointed in diagnosis and control are solved, and comprehensive, accurate and self-adaptive state monitoring and fault processing of the wind power generation system are achieved.

Description

Technical Field

[0001] This invention belongs to the field of wind power equipment condition monitoring and fault diagnosis technology, specifically involving a method and device for fault diagnosis of wind power systems based on multi-source data fusion and deep learning. Background Technology

[0002] As a major clean energy source, wind power units are usually installed in areas with harsh environments and poor accessibility, making early fault detection and accurate diagnosis crucial to avoid catastrophic downtime and economic losses.

[0003] Existing fault diagnosis methods are mostly siloed, each targeting specific components or data types, and have significant limitations: while sensitive to mechanical faults, they struggle to detect early signs of electrical faults; they are highly model-dependent, with performance heavily reliant on accurate system mathematical models, posing a significant challenge for nonlinear, strongly coupled wind power systems; and they neglect sensor faults, which, as data sources, can lead to misdiagnosis by the entire diagnostic system. Current methods lack effective identification and data reconstruction capabilities for sensor faults and do not adequately consider the fusion of multi-source information.

[0004] Therefore, there is an urgent need in this field for a systematic solution that can break down data silos, comprehensively utilize multidimensional information, and seamlessly integrate diagnostic results into the control process. Summary of the Invention

[0005] The primary objective of this invention is to overcome the shortcomings of the prior art and provide a method and apparatus for fault diagnosis of wind power generation systems that can achieve early warning, precise positioning, strong self-adaptation, and fault-tolerant control capabilities through multi-source data fusion and deep learning.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A fault diagnosis method for wind power generation systems based on multi-source data fusion and deep learning includes the following steps: Step S1: Collect multi-source heterogeneous operation data of the wind turbine generator set, including vibration signals, electrical parameters, ambient wind speed and unit operating status parameters; Step S2: Preprocess and extract features from the multi-source heterogeneous operating data, and fuse the extracted features to form a global feature vector; Step S3: Input the global feature vector into the pre-trained hybrid deep learning diagnostic model and output the probability distribution of the fault type. The fault determination threshold of the hybrid deep learning diagnostic model can be dynamically adjusted according to the environmental wind speed. Step S4: When the hybrid deep learning diagnostic model outputs an indication of sensor failure, the faulty sensor signal is reconstructed using a sliding mode observer based on TS fuzzy theory to obtain the estimated true physical quantity. Step S5: Input the reconstructed real physical quantity into the model prediction fault-tolerant controller to generate control commands to compensate for the impact of faults and maintain stable system output power.

[0007] A further improvement of the present invention is that, in step S2, a short-time Fourier transform is performed on the vibration signal, electrical parameters, ambient wind speed and unit operating status parameters to generate a time-frequency spectrum, which is then used as the input to the convolutional neural network branch in the hybrid deep learning diagnostic model.

[0008] A further improvement of the present invention is that the hybrid deep learning diagnostic model includes a convolutional neural network branch and a long short-term memory network branch set in parallel, and the outputs of the convolutional neural network branch and the long short-term memory network branch are concatenated before the fully connected layer.

[0009] A further improvement of the present invention is that the hybrid deep learning diagnostic model introduces the environmental wind speed as a working condition label, so that the fault judgment threshold has a preset nonlinear functional relationship with the wind speed.

[0010] A further improvement of the present invention is that the sliding mode observer based on TS fuzzy theory maintains the sliding mode motion through an equivalent output control method and directly acquires fault information to achieve fault reconstruction.

[0011] A fault diagnosis device for wind power generation systems based on multi-source data fusion and deep learning includes: The multi-dimensional data acquisition module is used to collect multi-source heterogeneous operating data of wind turbine generators, including vibration signals, electrical parameters, ambient wind speed and unit operating status parameters; The data processing and feature fusion module is used to preprocess and extract features from the multi-source heterogeneous operating data, and fuse the extracted features to form a global feature vector; The intelligent diagnostic module is used to input the global feature vector into a pre-trained hybrid deep learning diagnostic model and output the probability distribution of fault types. The fault judgment threshold of the hybrid deep learning diagnostic model can be dynamically adjusted according to the environmental wind speed. The fault reconstruction module is used to reconstruct the fault sensor signal using a sliding mode observer based on TS fuzzy theory when the hybrid deep learning diagnostic model outputs an indication of sensor fault, so as to obtain the estimated real physical quantity. The fault-tolerant control module is used to input the reconstructed real physical quantity into the model prediction fault-tolerant controller, generate control commands to compensate for the impact of faults, and maintain the stability of the system output power. The human-computer interaction module is used to display diagnostic results and system status.

[0012] A further improvement of the present invention is that, in the data processing and feature fusion module, short-time Fourier transform is performed on the vibration signal, electrical parameters, ambient wind speed and unit operating status parameters to generate a time-spectrum diagram, which is then used as the input to the convolutional neural network branch in the hybrid deep learning diagnostic model.

[0013] A further improvement of the present invention is that the hybrid deep learning diagnostic model includes a convolutional neural network branch and a long short-term memory network branch set in parallel, and the outputs of the convolutional neural network branch and the long short-term memory network branch are concatenated before the fully connected layer.

[0014] An electronic device includes: a processor and a memory coupled to the processor, the memory storing a computer program, the computer program being executed by the processor to implement the steps of the multi-source data fusion and deep learning wind power generation system fault diagnosis method.

[0015] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the multi-source data fusion and deep learning method for fault diagnosis of wind power generation systems.

[0016] Compared with the prior art, the present invention has at least the following beneficial technical effects: 1. Comprehensive health monitoring: For the first time in wind turbine fault diagnosis, it systematically integrates multi-dimensional data such as mechanical, electrical, environmental, and control status to achieve a comprehensive perception of the overall system status, far exceeding the diagnostic capabilities of a single data source.

[0017] 2. High accuracy and early warning: Hybrid deep learning models (CNN+LSTM) can automatically mine deep and subtle fault precursor features from massive amounts of data, enabling early detection and accurate classification of faults.

[0018] 3. Strong adaptive capability: By introducing operating condition parameters, the fault diagnosis system can intelligently adapt to different operating conditions such as wind speed changes, significantly reducing false alarms and missed alarms.

[0019] 4. Closed-loop fault-tolerant control: It creatively integrates fault diagnosis, signal reconstruction and model prediction fault-tolerant control to form a complete "perception-diagnosis-decision-control" closed loop, which significantly improves the automation and intelligence level of the system and ensures the stability of the power grid.

[0020] 5. High reliability: Through cross-validation of multi-source information and sensor fault reconstruction technology, it can effectively identify and eliminate interference from sensor faults, providing high-confidence diagnostic results.

[0021] 6. Significant economic benefits: Predictive maintenance avoids unplanned downtime and major overhauls, significantly reducing operation and maintenance costs. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 : Overall architecture block diagram of the system described in this invention; Figure 2 : A schematic diagram of the structure of a hybrid deep learning diagnostic model; Figure 3 Schematic diagram of sensor fault reconstruction based on TS fuzzy sliding mode observer; Figure 4 Schematic diagram of the model prediction fault-tolerant controller; Figure 5 Power diagram of a system employing model predictive fault-tolerant control; Figure 6 The structural block diagram of the wind power generation system fault diagnosis device based on multi-source data fusion and deep learning of the present invention. Detailed Implementation

[0024] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

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

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

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

[0028] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0029] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] Example 1 The fault diagnosis method for wind power generation systems based on multi-source data fusion and deep learning provided by this invention includes the following steps: Step S1: Collect multi-source heterogeneous operation data of the wind turbine generator set, including vibration signals, electrical parameters, ambient wind speed and unit operating status parameters; Step S2: Preprocess and extract features from the multi-source heterogeneous operating data, and fuse the extracted features to form a global feature vector; Step S3: Input the global feature vector into the pre-trained hybrid deep learning diagnostic model and output the probability distribution of the fault type. The fault determination threshold of the hybrid deep learning diagnostic model can be dynamically adjusted according to the environmental wind speed. Step S4: When the hybrid deep learning diagnostic model outputs an indication of sensor failure, the faulty sensor signal is reconstructed using a sliding mode observer based on TS fuzzy theory to obtain the estimated true physical quantity. Step S5: Input the reconstructed real physical quantity into the model prediction fault-tolerant controller to generate control commands to compensate for the impact of faults and maintain stable system output power.

[0031] In this embodiment, in step S2, short-time Fourier transform is performed on the vibration signal, electrical parameters, ambient wind speed, and unit operating status parameters to generate a time-frequency spectrum, which is then used as the input to the convolutional neural network branch in the hybrid deep learning diagnostic model.

[0032] In this embodiment, the hybrid deep learning diagnostic model includes a convolutional neural network branch and a long short-term memory network branch set in parallel, and the outputs of the convolutional neural network branch and the long short-term memory network branch are concatenated before the fully connected layer.

[0033] In this embodiment, the hybrid deep learning diagnostic model introduces the environmental wind speed as a working condition label, so that the fault determination threshold has a preset nonlinear functional relationship with the wind speed.

[0034] In this embodiment, the sliding mode observer based on TS fuzzy theory maintains the sliding mode motion through an equivalent output control method and directly acquires fault information to achieve fault reconstruction.

[0035] Example 2 Figure 1 This invention presents the overall architecture of a wind power system fault diagnosis method based on multi-source data fusion and deep learning, the core of which lies in constructing a hierarchical, closed-loop intelligent system. This system is an intelligent wind power system fault diagnosis and operation and maintenance platform integrating multi-source data fusion, a hybrid deep learning model, and fault-tolerant control. Its architecture combines data-driven and model-driven approaches to construct a closed-loop management system covering the entire process from data acquisition to fault diagnosis and fault-tolerant control. The system first acquires vibration signals, electrical parameters, and environmental data of the wind power system in real time through a multi-source data acquisition layer, providing a foundation for subsequent processing. The data preprocessing layer performs time-frequency analysis, wavelet denoising, and feature extraction on the raw data, ultimately generating standardized feature vectors. The hybrid deep learning model layer works collaboratively with CNN and LSTM branches, concatenating CNN and LSTM features and achieving high-precision classification of five fault types through a fully connected layer. The sensor fault reconstruction layer, based on a TS fuzzy sliding mode observer, estimates faults through residual calculation and sliding mode surface integration, generating correction signals to compensate for sensor offset faults. The fault-tolerant control layer employs a model predictive control strategy, optimizing control commands (pitch angle β, inverter frequency f) based on the reconstructed signals to suppress power fluctuations within 1 second of a fault occurrence, while simultaneously adjusting unit operating parameters through closed-loop feedback. The human-machine interface and decision support layer provides interfaces for fault type visualization, maintenance suggestion generation, power fluctuation monitoring, and remote diagnostics, supporting SCADA system integration and cloud-based analysis. Through innovative designs such as multi-source data fusion, hybrid deep learning, real-time fault-tolerant control, and improved fault reconstruction accuracy, the system achieves intelligent operation and maintenance of wind power systems, providing theoretical support and engineering practice solutions for the sustainable development of the wind power industry.

[0036] According to the present invention, a fault diagnosis method for wind power generation systems based on multi-source data fusion and deep learning is proposed, comprising the following steps: Step S1: Collect multi-source heterogeneous operation data of the wind turbine generator set, including at least vibration signals, electrical parameters, ambient wind speed and unit operating status parameters; Step S2: Preprocess and extract features from the multi-source heterogeneous operating data, and fuse the extracted features to form a global feature vector; Step S3: Input the global feature vector into the pre-trained hybrid deep learning diagnostic model and output the probability distribution of the fault type. The fault determination threshold of the hybrid deep learning diagnostic model can be dynamically adjusted according to the environmental wind speed. Step S4: When the hybrid deep learning diagnostic model outputs an indication of sensor failure, the faulty sensor signal is reconstructed using a sliding mode observer based on TS fuzzy theory to obtain the estimated true physical quantity. Step S5: Input the reconstructed real physical quantity into the model prediction fault-tolerant controller to generate control commands to compensate for the impact of faults and maintain stable system output power; Step S6: Provide interfaces for fault type visualization, maintenance suggestion generation, power fluctuation monitoring, and remote diagnostics.

[0037] In step S1 above, the multi-source heterogeneous operation data acquisition of the wind turbine generator set first requires the deployment of sensors, specifically: For mechanical vibration data, vibration acceleration sensors are deployed on key mechanical components such as gearboxes, main shafts, and generator bearings to collect high-frequency vibration waveforms; For electrical parameters, data such as three-phase voltage, current, active power, reactive power, and frequency are collected from the inverter and grid interface. For environmental data, parameters such as wind speed, wind direction, ambient temperature, and humidity are collected; Data on generator unit operating status, such as generator speed, pitch angle, and yaw angle, are collected.

[0038] This layer of hardware includes piezoelectric sensors, strain gauges, voltage and current transformers, anemometers, temperature sensors, etc.

[0039] In step S2 above, the preprocessing and feature extraction of the multi-source heterogeneous operating data, and the fusion of the extracted features to form a global feature vector, specifically involves: a1: Data cleaning and alignment, performing timestamp alignment, outlier removal, and missing value filling for sensor data from different sampling rates.

[0040] a2: Denoising, using wavelet thresholding to denoise the vibration signal (db4 wavelet, soft threshold).

[0041] a3: Advanced feature extraction, performing time-domain, frequency-domain, and time-frequency-domain extraction on vibration signals; a4: Perform harmonic analysis on electrical parameters, extract symmetrical components, and calculate power factor, etc. a5: Environmental data is used as a key operating condition label for subsequent adaptive adjustments to the model.

[0042] a6: Data standardization, which normalizes all feature data to the same scale.

[0043] Figure 2 This paper presents a schematic diagram of a hybrid deep learning diagnostic model, consisting of a multimodal data input layer, a feature extraction sub-network, a feature fusion layer, and a classification output layer. It achieves high-precision fault diagnosis of wind power generation systems through the collaborative design of convolutional neural networks and long short-term memory networks. The model uses the time-frequency spectrum of vibration signals and time-series data of electrical parameters as dual input sources. First, a CNN branch extracts spatial features from the time-frequency spectrum, while an LSTM branch performs temporal modeling of the electrical parameter sequence. A 50-hidden-unit LSTM layer extracts temporal-dependent features, outputting a 50-dimensional feature vector. The two feature vectors are concatenated in the fusion layer to form a 114-dimensional joint feature vector, which is then nonlinearly mapped through a fully connected layer containing 100 neurons. Finally, a Softmax classifier outputs five fault probabilities. By combining the CNN's sensitivity to capturing local frequency domain features of vibration signals with the LSTM's ability to model the temporal dynamics of electrical parameters, the model constructs a multi-dimensional discrimination boundary in the feature space, achieving a fault classification accuracy of 98.32%. Its lightweight design and multi-source data collaborative analysis mechanism significantly improve diagnostic robustness under complex operating conditions. The specific model structure is as follows: a1: Input layer: Data feature-level fusion, which concatenates the feature vectors extracted from vibration, electrical, and state data to form a global high-dimensional feature vector, which comprehensively represents the system state and includes environmental condition labels.

[0044] The CNN branch is mainly used to process spatially local correlated data such as the time-spectrum graph of vibration signals, and automatically learn the local sensitive features of faults.

[0045] The LSTM branch is mainly used to process time series data such as electrical parameters and power, capturing their long-term dependencies and dynamic evolution patterns.

[0046] a2: Feature concatenation and fully connected layer: Concatenates the high-level abstract features extracted from the CNN branch and the LSTM branch.

[0047] a3: Adaptive Output Layer: The model's fault determination threshold is not fixed but dynamically adjusted according to operating condition labels such as wind speed, significantly reducing false alarms under different wind conditions. The model output is the probability distribution of various fault types (such as bearing wear, blade cracks, inverter IGBT faults, sensor offset, etc.).

[0048] a4: Model training, based on Matlab to generate a dataset containing normal operating conditions, bearing wear, gearbox failure, etc. a5: Loss function: Cross-entropy loss + wind speed weighted regularization term; a6: Optimizer: Adam (learning rate 1e-4, momentum 0.9).

[0049] Figure 3 This diagram illustrates the sensor fault reconstruction principle based on a TS fuzzy sliding mode observer. Through multi-module collaboration, it achieves real-time detection and compensation of sensor faults in a wind power generation system. The core process begins with a wind power generation system model, with control signals as input and sensor signals as output. The TS fuzzy model is used to locally linearize the nonlinear system, generating a system estimate. The sliding mode observer, designed based on the TS fuzzy framework, estimates the output and calculates residuals to reflect the deviation between the sensor signal and the observed values. The residual evaluation module determines fault occurrence by setting a threshold, triggering fault detection. The fault reconstruction module maintains sliding mode motion using the sliding surface and equivalent control law, directly calculating the fault estimate, where is the fault distribution matrix. Finally, the fault-tolerant control module compensates the observer output with the reconstructed fault signal, generating a corrected sensor signal and adjusting control commands to maintain stable system operation. The entire system leverages the nonlinear approximation capability of the TS fuzzy model, the robustness of the sliding mode observer, and the fault reconstruction mechanism of equivalent control to achieve high-precision sensor fault detection and dynamic compensation, ensuring the reliability and safety of the wind power generation system under fault scenarios.

[0050] In step S4 above, when the diagnostic model output indicates a sensor fault, the faulty sensor signal is reconstructed using a sliding mode observer based on TS fuzzy theory to obtain an estimated true physical quantity, specifically: System Modeling: Based on the dynamic equations of the wind turbine generator, a TS fuzzy model is constructed. The dynamic state equations of the wind turbine generator are as follows:

[0051] in, For state vectors, The input vector is the electromagnetic torque reference value. This is the output vector. The rotational speed of the high-speed shaft. This represents the electromagnetic torque of the motor.

[0052] Observer Design: A sliding mode observer is a nonlinear observer that leverages its inherent robustness to overcome uncertainties or nonlinearities in the system model. Furthermore, T-S fuzzy theory possesses advantages such as simple structure and strong approximation ability, facilitating the establishment of a global T-S fuzzy model of the system to handle nonlinear systems. Therefore, a sliding mode fault observer for a fuzzy T-S system is designed. When a sensor fails... When the sensor malfunctions, the residual will asymptotically converge to zero; however, when the sensor fails... Different responses are made to different faults. This can be achieved by analyzing the residuals. Logical partitioning is used to determine whether a sensor failure has occurred. The residual is calculated using the following formula: Where y is the sensor output, y hat For observer estimates. Through the sliding surface Maintaining sliding mode motion allows for direct reconstruction of fault information and equivalent control.

[0053] Algorithm steps: a1. Considering the presence of sensor faults in the system, the system state equation can be defined as follows:

[0054] in f ( t (D) indicates a sensor malfunction. f This is the system's fault distribution matrix.

[0055] a2. Design the sliding surface: ; a3. Equivalent control law: , where K is the gain matrix; a4. Fault Refactoring: f_hat = D^T*S.

[0056] Figure 4 This paper presents a schematic diagram of a model predictive fault-tolerant controller (MMC), and designs a fault-tolerant control strategy for sensor faults in the transmission system of a wind turbine generator. Based on received fault information and fault diagnosis results, the MMC compensates for the output of the faulty sensor, ensuring that the controller's input signal is equal to or close to the input when there is no fault, thereby reducing the impact of the fault on the system. In step S5 above, the reconstructed real physical quantity is input to the model prediction fault-tolerant controller to generate control commands to compensate for the impact of faults and maintain stable system output power, specifically as follows: Model predictive fault-tolerant controllers optimize the control input by minimizing a value function to achieve the control objective. Considering the system's ability to follow the desired value and the smooth control of increments, the objective function of model predictive control is determined based on the objective function proposed by Maciejowski.

[0057] Control strategy: Input the reconstructed sensor signal and combine it with the current wind speed to predict the power output for the next 10 steps.

[0058] Objective function:

[0059] in It is the predicted output, i.e., the generator speed w. g , r This represents the output reference value, i.e., the generator speed w. g,r ; u Represents control input; Δ u Represents the change in input; the prediction starting point is generally set to 1, meaning it tracks the reference value of the output from the initial time; the prediction domain is N. p Control domain N u Estimation error weights Q Input rate of change weight R Δ The input weight R is a parameter to be determined. Only by selecting appropriate parameters can the system achieve better control performance and meet the control requirements. By determining a suitable objective function, the control problem can be transformed into an optimization control problem.

[0060] Compensation mechanism: Adjust the pitch angle and inverter output to maintain stable power.

[0061] Example 3 like Figure 6 As shown, the multi-source data fusion and deep learning-based wind power generation system fault diagnosis device provided by the present invention includes: The multi-dimensional data acquisition module is used to collect multi-source heterogeneous operating data of wind turbine generators, including vibration signals, electrical parameters, ambient wind speed and unit operating status parameters; The data processing and feature fusion module is used to preprocess and extract features from the multi-source heterogeneous operating data, and fuse the extracted features to form a global feature vector; The intelligent diagnostic module is used to input the global feature vector into a pre-trained hybrid deep learning diagnostic model and output the probability distribution of fault types. The fault judgment threshold of the hybrid deep learning diagnostic model can be dynamically adjusted according to the environmental wind speed. The fault reconstruction module is used to reconstruct the fault sensor signal using a sliding mode observer based on TS fuzzy theory when the hybrid deep learning diagnostic model outputs an indication of sensor fault, so as to obtain the estimated real physical quantity. The fault-tolerant control module is used to input the reconstructed real physical quantity into the model prediction fault-tolerant controller, generate control commands to compensate for the impact of faults, and maintain the stability of the system output power. The human-computer interaction module is used to display diagnostic results and system status.

[0062] In this embodiment, in the data processing and feature fusion module, short-time Fourier transform is performed on the vibration signal, electrical parameters, ambient wind speed and unit operating status parameters to generate a time-spectrum graph, which is then used as the input to the convolutional neural network branch in the hybrid deep learning diagnostic model.

[0063] In this embodiment, the hybrid deep learning diagnostic model includes a convolutional neural network branch and a long short-term memory network branch set in parallel, and the outputs of the convolutional neural network branch and the long short-term memory network branch are concatenated before the fully connected layer.

[0064] Example 4 The present invention provides an electronic device comprising: a processor and a memory coupled to the processor, the memory storing a computer program, wherein the computer program, when executed by the processor, implements the steps of the wind power generation system fault diagnosis method based on multi-source data fusion and deep learning.

[0065] The electronic device may also include one or more of a multimedia component, an input / output (I / O) interface, and a communication component.

[0066] The processor controls the overall operation of the electronic device to complete all or part of the steps in the storage medium sharing method. The memory stores various types of data to support the operation of the electronic device. This data may include, for example, instructions for any application or method operating on the electronic device, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia components may include a screen and audio components. The screen may be, for example, a touchscreen, and the audio components are used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory or transmitted via a communication component. The audio component also includes at least one speaker for outputting audio signals. The I / O interface provides an interface between the processor and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical. The communication component is used for wired or wireless communication between the electronic device and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof; therefore, the corresponding communication component may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0067] In an exemplary embodiment, the electronic device may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing a storage medium sharing method.

[0068] Example 5 The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the wind power generation system fault diagnosis method based on multi-source data fusion and deep learning.

[0069] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0074] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A wind power system fault diagnosis method based on multi-source data fusion and deep learning, characterized in that, The method comprises the following steps: Step S1: collecting multi-source heterogeneous operation data of a wind turbine generator set, including vibration signals, electrical parameters, environmental wind speed and unit operation state parameters; Step S2: pre-processing and feature extraction are performed on the multi-source heterogeneous operation data, and the extracted features are fused to form a global feature vector; Step S3: the global feature vector is input into a pre-trained hybrid deep learning diagnosis model, and a probability distribution of fault types is output, wherein a fault determination threshold of the hybrid deep learning diagnosis model can be dynamically adjusted according to the environmental wind speed; Step S4: when the hybrid deep learning diagnosis model outputs an indication of a sensor fault, a sliding mode observer based on T-S fuzzy theory is used to reconstruct the fault sensor signal to obtain an estimated real physical quantity; Step S5: the reconstructed real physical quantity is input into a model predictive fault-tolerant controller to generate a control instruction to compensate for the fault impact and maintain stable system output power.

2. The wind power system fault diagnosis method of multi-source data fusion and deep learning according to claim 1, characterized in that, In step S2, short-time Fourier transform is performed on the vibration signals, electrical parameters, environmental wind speed and unit operation state parameters to generate time-frequency spectrograms, which are used as inputs of a convolutional neural network branch in the hybrid deep learning diagnosis model. 3.The wind power system fault diagnosis method of multi-source data fusion and deep learning according to claim 1, characterized in that, The hybrid deep learning diagnosis model comprises a convolutional neural network branch and a long short-term memory network branch arranged in parallel, and the outputs of the convolutional neural network branch and the long short-term memory network branch are spliced before a full connection layer.

4. The wind power system fault diagnosis method of multi-source data fusion and deep learning according to claim 1, characterized in that, The hybrid deep learning diagnosis model introduces the environmental wind speed as a working condition label, so that the fault determination threshold has a preset nonlinear function relationship with the wind speed.

5. The wind power system fault diagnosis method of multi-source data fusion and deep learning according to claim 1, characterized in that, The sliding mode observer based on T-S fuzzy theory maintains sliding mode motion through equivalent output control method and directly obtains fault information to realize fault reconstruction.

6. The wind power system fault diagnosis device based on multi-source data fusion and deep learning, characterized in that, It comprises: A multi-dimensional data acquisition module for acquiring multi-source heterogeneous operation data of a wind turbine generator set, including vibration signals, electrical parameters, environmental wind speed and unit operation state parameters; A data processing and feature fusion module for pre-processing and feature extraction of the multi-source heterogeneous operation data, and fusing the extracted features to form a global feature vector; An intelligent diagnosis module for inputting the global feature vector into a pre-trained hybrid deep learning diagnosis model to output a probability distribution of fault types, wherein a fault determination threshold of the hybrid deep learning diagnosis model can be dynamically adjusted according to the environmental wind speed; A fault reconstruction module for reconstructing a fault sensor signal using a sliding mode observer based on T-S fuzzy theory when the hybrid deep learning diagnosis model outputs an indication of a sensor fault to obtain an estimated real physical quantity; A fault-tolerant control module for inputting the reconstructed real physical quantity into a model predictive fault-tolerant controller to generate a control instruction to compensate for the fault impact and maintain stable system output power; A human-computer interaction module for displaying diagnosis results and system status.

7. The wind power system fault diagnosis device of multi-source data fusion and deep learning according to claim 6, characterized in that, In the data processing and feature fusion module, the vibration signal, the electrical parameter, the environmental wind speed and the unit operation state parameter are subjected to short-time Fourier transform to generate a time-frequency spectrogram, and the time-frequency spectrogram is taken as an input of a convolutional neural network branch in the hybrid deep learning diagnosis model.

8. The wind power system fault diagnosis device of multi-source data fusion and deep learning according to claim 7, characterized in that, The hybrid deep learning diagnosis model comprises a convolutional neural network branch and a long short-term memory network branch arranged in parallel, and outputs of the convolutional neural network branch and the long short-term memory network branch are spliced before a full connection layer.

9. An electronic device, comprising: Comprise: A processor and a memory coupled to the processor, the memory storing a computer program, the computer program being executed by the processor to implement the steps of the multi-source data fusion and deep learning wind power system fault diagnosis method of any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program implements the steps of the multi-source data fusion and deep learning wind power system fault diagnosis method of any one of claims 1-5 when executed by a processor.

Images