Operation and maintenance method and system for small wind power equipment

Abstract

The invention relates to the technical field of wind power operation and maintenance, and discloses an operation and maintenance method and system for small wind power equipment, and the method comprises the steps: collecting the vibration, acoustics and image data of a blade through a detection node disposed at a fan site; constructing a digital twinborn body of the fan blade; a lightweight AI model deployed on the detection node is utilized to perform real-time analysis on the acquired data to realize preliminary identification of blade abnormality; when abnormity is identified, the associated feature data is uploaded to a cloud platform, the digital twin is driven to perform simulation deduction, and fault diagnosis and life prediction are output; according to the diagnosis and life prediction result, an operation and maintenance decision instruction is automatically generated and issued, multi-modal sensing data and a physical mechanism model are fused, an optimized operation and maintenance decision can be automatically generated, the operation and maintenance efficiency is remarkably improved, and the periodic operation and maintenance cost of the wind power equipment is reduced.

Description

Technical Field

[0001] This invention relates to the field of wind power operation and maintenance technology, and more specifically, to a method and system for the operation and maintenance of small wind power equipment. Background Technology

[0002] With the acceleration of the global energy transition, wind power, as an important component of clean energy, is expanding its application from large-scale centralized wind farms to distributed, miniaturized scenarios. Small wind turbines, due to their flexible installation and wide applicability, demonstrate significant advantages in areas such as power supply in remote regions, communication base stations, and agriculture, forestry, animal husbandry, and fisheries. However, compared to traditional large wind turbines, small wind turbines face more prominent operation and maintenance challenges. Small wind turbines are typically deployed in remote areas such as mountains, islands, and border regions, with dispersed geographical locations and inconvenient transportation, resulting in high costs and long cycles for maintenance personnel to reach the site. Traditional periodic inspection methods are not only inefficient but also difficult to detect sudden faults in a timely manner. Furthermore, small wind turbines have small unit capacity and limited power generation revenue; adopting the high-cost monitoring solutions used in large wind farms would significantly reduce the project's economic viability. However, without effective monitoring, typical faults such as blade cracks, icing, and imbalance can lead to decreased power generation efficiency or, in severe cases, blade breakage and tower collapse, causing even greater economic losses. Summary of the Invention

[0003] This application provides an operation and maintenance method and system for small wind turbines, which effectively solves the technical problems in the prior art, such as high operation and maintenance costs, untimely fault detection, and lack of accurate life prediction, caused by the dispersed distribution and complex environment of small wind turbines. It realizes real-time monitoring of wind turbine blade status, early identification of anomalies, fault diagnosis, and quantitative prediction of remaining life, and finally forms a closed-loop automated operation and maintenance from status perception to decision execution, which significantly improves the reliability and economy of small wind turbine operation.

[0004] To achieve the above objectives, the present invention provides a method for the operation and maintenance of small wind power equipment, comprising:

[0005] Vibration, acoustic, and image data of the blades are collected through detection nodes deployed at the wind turbine site;

[0006] Constructing digital twins of wind turbine blades;

[0007] By utilizing a lightweight AI model deployed on the detection node, the collected data is analyzed in real time to achieve preliminary identification of leaf anomalies;

[0008] When an anomaly is detected, the associated feature data is uploaded to the cloud platform, and the digital twin is driven to perform simulation and deduction, outputting fault diagnosis and lifespan prediction.

[0009] According to the results of the diagnosis and life prediction, automatically generate and issue operation and maintenance decision instructions.

[0010] Further, through the detection node deployed on the fan site, vibration, acoustic and image data of the blade are collected, specifically including:

[0011] Through the vibration acceleration sensor installed at a specific position of the fan hub or the nacelle head, vibration signals representing the dynamic response of the blade structure are collected;

[0012] Through the audio collection array deployed outside the nacelle and pointing to the blade rotation plane, aerodynamic acoustic signals and impact sound signals during blade rotation are collected;

[0013] Through the weather-resistant high-definition camera installed on the tower or nearby tower, periodic image sequences of the blade surface are collected under specific lighting conditions or auxiliary lighting;

[0014] The detection node synchronously collects and encapsulates the one-dimensional convolutional neural network, the time series anomaly detection model and the lightweight visual Transformer model through the integrated multi-protocol data interface.

[0015] Further, a digital twin of the fan blade is constructed, specifically including:

[0016] Based on the three-dimensional geometric model of the fan blade, the material physical properties and the aerodynamic load characteristics, a structural mechanics simulation model is established;

[0017] The structural mechanics simulation model is bidirectionally coupled with the vibration, acoustic and image data of the blade;

[0018] The structural mechanics simulation model is driven by the vibration, acoustic and image data of the blade to dynamically simulate the stress distribution, deformation state and vibration mode of the fan blade under actual working conditions.

[0019] Further, the collected data is analyzed in real time using the lightweight AI model deployed on the detection node to achieve preliminary identification of blade abnormalities, specifically including:

[0020] The collected raw data is preprocessed, and the preprocessing includes time-frequency transformation of vibration data and acoustic data to generate frequency domain features, and normalization and enhancement operations on image data;

[0021] The processed multi-modal feature data is synchronously input into the lightweight AI model;

[0022] Further, the lightweight AI model specifically includes:

[0023] A one-dimensional convolutional neural network analyzes the vibration frequency domain characteristics, identifies specific resonance frequencies related to blade imbalance or structural cracks, and obtains resonance frequency confidence;

[0024] A time series anomaly detection model processes acoustic signals, captures stripped and crack propagation anomaly acoustic signals embedded in aerodynamic noise, and obtains acoustic signal confidence;

[0025] A lightweight visual Transformer model identifies image data, detects crack, icing, and lightning damage features on the blade surface, and obtains surface feature confidence;

[0026] The confidence outputs of the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model are fused by weighted decision, and when the integrated anomaly confidence after fusion exceeds the dynamic adaptive threshold, it is determined that the blade anomaly is preliminarily identified and data upload is triggered.

[0027] Further, the confidence outputs of the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model are fused by weighted decision, specifically including:

[0028] A weighted decision fusion model is established, and appropriate weight coefficients are assigned to the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model based on the quality evaluation results of each sensor data and historical diagnostic accuracy, wherein:

[0029] The resonance frequency confidence weight Wv is determined based on signal signal-to-noise ratio and acquisition integrity;

[0030] The acoustic signal confidence weight Wa is determined based on background noise level and spectral clarity;

[0031] The surface feature confidence weight Wi is determined based on lighting conditions and image resolution;

[0032] A weighted fusion decision function is set ;

[0033] Wherein, Pv is the resonance frequency confidence, Pa is the acoustic signal confidence, Pi is the surface feature confidence, and P is the integrated anomaly confidence;

[0034] Pt is set as the determination threshold, which is dynamically adjusted based on local environmental conditions and equipment conditions. When the integrated anomaly confidence satisfies P≥Pt, it is determined that the blade anomaly is preliminarily identified, and the feature data is uploaded to the cloud platform.

[0035] Further, when an anomaly is identified, the associated feature data is uploaded to a cloud platform, and the digital twin is driven to perform simulation deduction, outputting diagnosis and life prediction of the fault, specifically including:

[0036] The detection node packs and uploads the multi-modal feature data, corresponding raw data segments, and equipment operating condition data within a certain time window before and after the trigger time of the anomaly to the cloud platform;

[0037] After receiving the above data, the cloud platform inputs the data as boundary conditions and loads to drive the fan blade digital twin to perform high-fidelity simulation calculation, simulating the evolution process of the abnormal feature under actual operating conditions;

[0038] Based on the simulation results, combined with the historical fault case library, the diagnosis results of crack type, location, depth, or ice thickness and area are output through a combination of physical mechanisms and data-driven methods;

[0039] Based on the diagnosis results, the remaining useful life and failure probability of the fan blade under a given operating strategy are predicted through accelerated life test simulation of the digital twin.

[0040] Further, the simulation of the evolution process of the abnormal feature under actual operating conditions specifically includes:

[0041] Mapping the equipment operating condition data to the simulation boundary conditions of the digital twin;

[0042] Based on the identified abnormal type and confidence, initializing the abnormal feature at the corresponding position of the digital twin;

[0043] Performing coupled simulation under the boundary conditions:

[0044] Based on the coupled simulation results, accelerating the time scale to deduce the evolution of the abnormal feature;

[0045] Recording key physical parameters during the entire evolution process, including maximum equivalent stress, fatigue damage index, and crack tip stress intensity factor, as the basis for diagnosis and life prediction.

[0046] Further, predicting the remaining useful life and failure probability of the fan blade under a given operating strategy specifically includes:

[0047] Based on the historical operating data of the fan and the local environmental conditions, constructing a typical load spectrum representing future operating conditions;

[0048] Performing simulation in the digital twin to apply the load spectrum in an accelerated time scale to quantify the damage level of the fan;

[0049] The failure criterion is continuously evaluated during the simulation process;

[0050] The simulation time experienced from the current state to reaching any failure criterion is recorded, which is converted into the predicted remaining service life;

[0051] The curve of the remaining service life and the failure probability changing over time is output, a remaining life prediction report is generated, and uploaded to the cloud platform.

[0052] In order to achieve the above purpose, the present application also provides a small wind power equipment operation and maintenance system, which further comprises:

[0053] The data acquisition module is used to collect vibration, acoustic and image data of the blade through the detection node deployed on the wind turbine site;

[0054] The model construction module is used to construct a digital twin of the wind turbine blade;

[0055] The data analysis module uses a lightweight AI model deployed on the detection node to perform real-time analysis on the collected data, and realizes preliminary identification of blade abnormalities;

[0056] The simulation and deduction module uploads the associated feature data to the cloud platform when an abnormality is identified, and drives the digital twin to perform simulation and deduction, and outputs the fault diagnosis and life prediction;

[0057] The operation and maintenance decision module is used to automatically generate and issue operation and maintenance decision instructions according to the results of the diagnosis and life prediction.

[0058] Compared with the prior art, the present application has the beneficial effects that through real-time analysis based on multi-modal sensor data and lightweight AI models, the timeliness and accuracy of blade abnormality identification are effectively improved; through high-fidelity simulation and accelerated life testing of the digital twin, fault diagnosis and quantitative prediction of the remaining life are realized, providing a scientific basis for predictive maintenance; finally, through the automatic operation and maintenance decision loop, the operation and maintenance cost and downtime risk are greatly reduced, and the reliability and life cycle economic benefit of small wind power equipment operation are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0059] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for purposes of illustration only and are not considered a limitation of the present application. Moreover, like reference numerals are used to designate identical components throughout the specification. In the drawings:

[0060] Figure 1 A flowchart of a small wind power equipment operation and maintenance method in an embodiment of the present application is shown;

[0061] Figure 2 Fig. 1 shows a structural schematic diagram of an operation and maintenance system of a small wind power device according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] The specific embodiments of the present application will be further described in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0063] In the description of the present application, it should be understood that the terms “center”, “upper”, “lower”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inner”, “outer” and the like indicate the orientation or positional relationship shown in the drawings based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0064] The terms “first”, “second” are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with “first”, “second” can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of “a plurality of” is two or more.

[0065] In the description of the present application, it should be noted that, unless otherwise explicitly specified and limited, the terms “mounting”, “connection”, “connecting” should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0066] The following is a description of the preferred embodiments of the present application in conjunction with the accompanying drawings.

[0067] S110: collecting vibration, acoustic and image data of the blade through the detection node deployed at the fan site.

[0068] In some embodiments of the present application, vibration, acoustic and image data of the blade are collected through the detection node deployed at the fan site, specifically including:

[0069] vibration signals representing the dynamic response of the blade structure are collected through vibration acceleration sensors installed at specific positions of the fan hub or nacelle head;

[0070] The aerodynamic acoustic signals and impact sound signals in the blade rotation process are collected through an audio collection array deployed outside the cabin and pointing to the blade rotation plane;

[0071] Periodic image sequences of the blade surface are collected under specific lighting conditions or auxiliary lighting through a weather-resistant high-definition camera installed on the tower or nearby tower;

[0072] The detection node synchronously collects and encapsulates the one-dimensional convolutional neural network, the time series anomaly detection model and the lightweight visual Transformer model through an integrated multi-protocol data interface.

[0073] In the embodiment, the specific lighting conditions or auxiliary lighting specifically include: the ambient light intensity is monitored in real time through a light sensor, when the daytime natural light is lower than the minimum threshold required for clear imaging, or in the night without light, the light compensation lamp group arranged in the same group with the weather-resistant high-definition camera deployed on the tower or nearby tower is automatically triggered to uniformly illuminate the blade surface at a specific angle and color temperature, so that the blade surface image sequence with clear details and moderate contrast can be obtained at any time, and a reliable data basis is provided for subsequent visual anomaly recognition.

[0074] In the embodiment, the one-dimensional convolutional neural network, the time series anomaly detection model and the lightweight visual Transformer model are synchronously collected and encapsulated through an integrated multi-protocol data interface, which specifically includes: the data collection unit integrated in the detection node can be compatible with and simultaneously receive multi-channel data streams from the vibration acceleration sensor, the audio collection array and the weather-resistant high-definition camera; the unit applies a synchronous time stamp to all accessed sensor data through a unified hardware clock to eliminate the data time sequence deviation caused by transmission delay or independent collection; then, the multi-modal data such as vibration signals, acoustic signals and image sequences with synchronous time sequence are integrated and data packed according to the preset encapsulation format, to provide a time sequence consistent and correlation accurate data basis for subsequent real-time analysis and fault diagnosis of the lightweight AI model.

[0075] The beneficial effects of the above technical solutions are: by constructing an intelligent perception system integrating multi-modal data detection, data integration and unified time stamp encapsulation, the industry problems such as unstable image collection quality in complex outdoor environment and time sequence misplacement of multi-source sensing data are effectively solved, the correlation and consistency of vibration, acoustic and visual data in time-space dimension are ensured, a high-quality and high-reliability data basis is provided for subsequent real-time analysis and accurate diagnosis of the lightweight AI model, and the accuracy of blade state perception and fault recognition is improved.

[0076] S120: Construct a digital twin of the wind turbine blade.

[0077] In some embodiments of the application, a digital twin of a fan blade is constructed, specifically comprising:

[0078] Based on the three-dimensional geometric model of the fan blade, the material physical properties, and the aerodynamic load characteristics, a structural mechanics simulation model is established.

[0079] The structural mechanics simulation model is bidirectionally coupled with the vibration, acoustic, and image data of the blade.

[0080] The structural mechanics simulation model is driven by the vibration, acoustic, and image data of the blade to dynamically simulate the stress distribution, deformation state, and vibration mode of the fan blade under actual working conditions.

[0081] In this embodiment, the structural mechanics simulation model is a high-fidelity calculation model based on the three-dimensional geometry of the fan blade, the composite material layer properties, and the aerodynamic load distribution characteristics. It discretizes the actual blade into a large number of element grids through finite element analysis method, and establishes a mathematical model that can reflect the structural response of the blade under complex load according to the material mechanics constitutive relation and aerodynamics principle, thereby realizing the digital simulation and calculation of key mechanical behaviors such as blade stress distribution, dynamic deformation, and vibration mode.

[0082] The above-mentioned scheme has the beneficial effects that: by establishing a high-fidelity structural mechanics simulation model that integrates geometry, material, and aerodynamic characteristics, and realizing bidirectional dynamic coupling with the detected vibration, acoustic, and image data of the blade, a digital twin that can accurately map the physical entity state is constructed, effectively breaking through the limitation of traditional monitoring methods that can only obtain surface data, and realizing the visualized quantitative analysis and dynamic tracking of key mechanical states such as blade internal stress distribution, dynamic deformation, and vibration mode, providing a digital analysis foundation based on physical mechanism for blade structural health assessment, damage evolution prediction, and accurate life judgment.

[0083] S130: Using the lightweight AI model deployed on the detection node, real-time analysis is performed on the collected data to realize preliminary identification of blade abnormalities.

[0084] In some embodiments of the application, a lightweight AI model is deployed on the detection node to perform real-time analysis on the collected data to realize preliminary identification of blade abnormalities, specifically comprising:

[0085] The collected raw data is preprocessed, including time-frequency transformation of vibration data and acoustic data to generate frequency domain features, and normalization and enhancement operations on image data.

[0086] The processed multi-modal feature data is input into the lightweight AI model.

[0087] In this embodiment, the lightweight AI model specifically includes:

[0088] A one-dimensional convolutional neural network for analyzing vibration frequency domain characteristics, identifying specific resonance frequencies related to blade imbalance or structural cracks, and obtaining resonance frequency confidence;

[0089] A time series anomaly detection model for processing acoustic signals, capturing abnormal acoustic signals such as peeling and crack propagation embedded in aerodynamic noise, and obtaining acoustic signal confidence;

[0090] A lightweight visual Transformer model for identifying image data, detecting crack, icing, and lightning damage features on the blade surface, and obtaining surface feature confidence;

[0091] The confidence outputs of the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model are fused by weighted decision, and when the integrated anomaly confidence after fusion exceeds the dynamic adaptive threshold, it is determined that the blade anomaly is preliminarily identified and data upload is triggered.

[0092] In this embodiment, the lightweight AI model uses large-scale small fan historical running data sets on a cloud platform server to train standard-scale one-dimensional convolutional neural networks, time series anomaly detection models, and visual Transformer models as baseline models; then through knowledge distillation, channel pruning, and parameter quantization, etc. Model compression technology, the above baseline model is lightweight processed, and finally the lightweight AI model is obtained.

[0093] In this embodiment, the confidence outputs of the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model are fused by weighted decision, specifically including:

[0094] A weighted decision fusion model is established, and based on the quality evaluation results of each sensor data and the historical diagnosis accuracy, appropriate weight coefficients are assigned to the one-dimensional convolutional neural network, the time series anomaly detection model, and the lightweight visual Transformer model, wherein:

[0095] The resonance frequency confidence weight Wv is determined based on the signal signal-to-noise ratio and the acquisition integrity;

[0096] The acoustic signal confidence weight Wa is determined based on the background noise level and the frequency spectrum definition;

[0097] The surface feature confidence weight Wi is determined based on the lighting conditions and the image resolution;

[0098] The weighted fusion decision function is set ;

[0099] Where Pv is the confidence level of the resonant frequency, Pa is the confidence level of the acoustic signal, Pi is the confidence level of the surface features, and P is the confidence level of the overall anomaly.

[0100] Pt is set as the judgment threshold, which is dynamically adjusted based on local environmental conditions and equipment conditions. When the comprehensive anomaly confidence level meets P≥Pt, it is determined that a blade anomaly has been initially identified, and the feature data is triggered to be uploaded to the cloud platform.

[0101] In this embodiment, the judgment threshold Pt is dynamically adjusted based on local environmental conditions and equipment conditions. The environmental conditions specifically include local wind speed, temperature, and humidity; the equipment conditions include the equipment's cumulative operating time and historical failure frequency. When the system detects severe environmental conditions or when the equipment enters the late stage of its lifespan, the system will appropriately lower the judgment threshold Pt according to the actual situation to improve the system's early warning sensitivity under high-risk conditions. Conversely, in a stable environment and under healthy equipment conditions, it will return to the standard threshold, thereby achieving intelligent adaptive adjustment of anomaly detection sensitivity while ensuring a low false alarm rate.

[0102] The beneficial effects of the above scheme are as follows: by constructing a multimodal lightweight AI analysis model based on vibration, acoustics and vision, and combining it with a data quality adaptive weighted decision fusion mechanism, the limitations and reliability of single sensors in complex environments are effectively overcome. This enables the identification of multiple types of faults such as blade imbalance, structural cracks, and icing. At the same time, through dynamic threshold control and confidence assessment, the false alarm and false negative rates are significantly reduced, providing a highly reliable early warning basis for triggering deep cloud-based diagnostics, thereby greatly improving the automation and accuracy of edge-side intelligent monitoring.

[0103] S140: When an anomaly is detected, the associated feature data is uploaded to the cloud platform, and the digital twin is driven to perform simulation and deduction, outputting fault diagnosis and life prediction.

[0104] In some embodiments of the present invention, when an anomaly is detected, the associated feature data is uploaded to the cloud platform, and the digital twin is driven to perform simulation and deduction, outputting fault diagnosis and lifespan prediction, specifically including:

[0105] The detection node will package and upload the multimodal feature data, the corresponding original data fragments, and the equipment operating condition data within a specific time window before and after the triggering of the abnormal moment to the cloud platform;

[0106] After receiving the above data, the cloud platform uses the data as boundary conditions and load inputs to drive the digital twin of the wind turbine blade to perform high-fidelity simulation calculations, simulating the evolution process of abnormal features under actual operating conditions.

[0107] Based on the simulation results, combined with the historical fault case library, through the combination of physical mechanism and data-driven, the diagnostic results of crack type, location, depth or icing thickness, area are output;

[0108] Based on the diagnostic results, the accelerated life test simulation is performed through the digital twin to predict the remaining useful life and failure probability of the fan blade under the given operation strategy.

[0109] In this embodiment, the simulation of the evolution process of abnormal characteristics under actual operating conditions specifically includes:

[0110] Mapping the equipment operating condition data to the simulation boundary conditions of the digital twin;

[0111] Based on the identified abnormal type and confidence, initializing the abnormal characteristics at the corresponding position of the digital twin;

[0112] Performing coupled simulation under the boundary conditions:

[0113] Based on the coupled simulation results, accelerating the time scale to deduce the evolution of abnormal characteristics;

[0114] Recording the key physical parameters in the entire evolution process, including the maximum equivalent stress, fatigue damage index, and crack tip stress intensity factor, as the basis for diagnosis and life prediction.

[0115] In this embodiment, predicting the remaining useful life and failure probability of the fan blade under the given operation strategy specifically includes:

[0116] Based on the historical operating data of the fan and the local environmental conditions, constructing a typical load spectrum representing future operating conditions;

[0117] Performing simulation in the digital twin to cyclically apply the load spectrum on the accelerated time scale to quantify the damage degree of the fan;

[0118] Continuously evaluating the failure criteria during the simulation process;

[0119] Recording the simulation time experienced from the current state to reaching any failure criterion, which after conversion is the predicted remaining useful life;

[0120] Outputting the curve of the remaining useful life and failure probability over time, generating a remaining life prediction report, and uploading it to the cloud platform.

[0121] The beneficial effects of the above scheme are as follows: a major breakthrough in fault diagnosis and prediction capabilities has been achieved through simulation evolution: high-fidelity simulation based on digital twins transforms the abnormal features of edge detection into fault quantification parameters with clear physical meaning; through accelerated life test simulation, the performance evolution of the blade throughout its entire life cycle is pre-enacted in virtual space, and the remaining life prediction curve with both time and probability dimensions is output, providing a decision-making basis that combines physical mechanisms and data-driven approaches for precise predictive maintenance, and significantly improving the foresight and economy of operation and maintenance strategies.

[0122] S150: Based on the results of the diagnosis and life prediction, automatically generate and issue operation and maintenance decision instructions.

[0123] In this embodiment, based on the results of the diagnosis and lifespan prediction, operation and maintenance decision instructions are automatically generated and issued. Specifically, the system automatically matches the preset operation and maintenance strategy library based on the remaining lifespan prediction report through the built-in decision rule engine, generates the optimal operation and maintenance plan that includes specific maintenance measures, suggested execution time windows, and a list of required resources, and issues the plan to the relevant operation and maintenance personnel's mobile terminals or scheduling system in real time in the form of standardized work orders, thereby realizing full automation and intelligence from status diagnosis to maintenance execution.

[0124] The beneficial effects of the above solution are: by transforming the traditional experience-based passive maintenance into data-driven proactive early warning through diagnosis and prediction results, unplanned downtime can be effectively avoided; by accurately quantifying the urgency of maintenance and resource requirements, operation and maintenance costs can be significantly reduced and resource utilization can be improved, and operation and maintenance response efficiency and decision-making can be greatly enhanced.

[0125] To further illustrate the technical concept of this invention, the technical solution of this invention will now be described in conjunction with specific application scenarios.

[0126] Correspondingly, such as Figure 2 As shown, this application also provides an operation and maintenance system for small wind power equipment, including:

[0127] The data acquisition module is used to collect vibration, acoustic, and image data of the blades through detection nodes deployed at the wind turbine site;

[0128] The model building module is used to build digital twins of wind turbine blades;

[0129] The data analysis module uses a lightweight AI model deployed on the detection node to perform real-time analysis on the collected data, enabling preliminary identification of leaf anomalies.

[0130] The simulation and deduction module, when an anomaly is detected, uploads the associated feature data to the cloud platform and drives the digital twin to perform simulation and deduction, outputting fault diagnosis and lifespan prediction.

[0131] An operation and maintenance decision module is configured to automatically generate and issue operation and maintenance decision instructions based on the results of the diagnosis and life prediction.

[0132] In the description of the above-described embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0133] Although the present application has been described with reference to the embodiments above, various modifications can be made to the present application and equivalents thereof without departing from the scope of the present application. In particular, the features of the disclosed embodiments can be used in any combination without departing from the scope of the present application, and the combinations of these features are not described in the present specification only for the sake of brevity and conciseness.

[0134] It should be understood by those of ordinary skill in the art that the above are only preferred embodiments of the present application and are not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art can still make modifications to the technical solutions described in the foregoing embodiments or make equivalent replacements to some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for the operation and maintenance of a small wind power device, characterized in that, include: Vibration, acoustic, and image data of the blades are collected through detection nodes deployed at the wind turbine site; Constructing digital twins of wind turbine blades; By utilizing a lightweight AI model deployed on the detection node, the collected data is analyzed in real time to achieve preliminary identification of leaf anomalies; When an anomaly is detected, the associated feature data is uploaded to the cloud platform, and the digital twin is driven to perform simulation and deduction, outputting fault diagnosis and lifespan prediction. Based on the results of the diagnosis and lifespan prediction, operation and maintenance decision instructions are automatically generated and issued.

2. The operation and maintenance method for a small wind power device according to claim 1, characterized in that, Vibration, acoustic, and image data of the blades are collected through monitoring nodes deployed at the wind turbine site, specifically including: Vibration signals characterizing the dynamic response of the blade structure are collected by vibration acceleration sensors installed at specific locations in the wind turbine hub or nacelle head. By deploying an audio acquisition array outside the nacelle and pointing towards the plane of blade rotation, aeroacoustic signals and impact sound signals are collected during the blade rotation process. A weather-resistant high-definition camera installed on the tower or wind turbine tower is used to collect a series of periodic images of the blade surface under specific lighting conditions or auxiliary lighting. The detection node synchronously acquires and encapsulates data from the vibration acceleration sensor, the audio acquisition array, and the weather-resistant high-definition camera through an integrated multi-protocol data interface.

3. The operation and maintenance method for a small wind power device according to claim 1, characterized in that, Constructing digital twins of wind turbine blades specifically includes: A structural mechanics simulation model is established based on the three-dimensional geometric model, material physical properties, and aerodynamic load characteristics of the wind turbine blades. The structural mechanics simulation model is bidirectionally coupled with the vibration, acoustic, and image data of the blade. The structural mechanics simulation model is driven by the vibration, acoustic and image data of the blades to dynamically simulate the stress distribution, deformation state and vibration mode of the wind turbine blades under actual working conditions.

4. The operation and maintenance method for a small wind power device according to claim 1, characterized in that, Using a lightweight AI model deployed on the detection node, the collected data is analyzed in real time to achieve preliminary identification of leaf anomalies, specifically including: The collected raw data is preprocessed, including time-frequency transformation of vibration and acoustic data to generate frequency domain features, and normalization and enhancement operations on image data. The processed multimodal feature data is synchronously input into the lightweight AI model.

5. The operation and maintenance method for a small wind power device according to claim 4, characterized in that, The lightweight AI model specifically includes: A one-dimensional convolutional neural network is used to analyze the vibration frequency domain characteristics, identify specific resonant frequencies related to blade imbalance or structural cracks, and obtain the confidence level of the resonant frequencies. A time-series anomaly detection model for processing acoustic signals is used to capture anomalous acoustic signals such as peeling and crack propagation mixed in aerodynamic noise, and to obtain the confidence level of the acoustic signals. A lightweight visual Transformer model for recognizing image data is used to detect cracks, icing, and lightning damage features on the blade surface and to derive the confidence level of surface features. The confidence scores of the one-dimensional convolutional neural network, the temporal anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model are weighted and fused. When the combined anomaly confidence score after fusion exceeds the dynamic adaptive threshold, it is determined that a blade anomaly has been initially identified and data upload is triggered.

6. The operation and maintenance method for a small wind power device according to claim 5, characterized in that, The weighted decision fusion of the confidence scores output by the one-dimensional convolutional neural network, the temporal anomaly detection model, and the lightweight visual Transformer model in the lightweight AI model specifically includes: A weighted decision fusion model is established. Based on the quality assessment results and historical diagnostic accuracy of each sensor data, appropriate weight coefficients are assigned to the one-dimensional convolutional neural network, the temporal anomaly detection model, and the lightweight visual Transformer model, wherein: The confidence weight Wv of the resonant frequency is determined based on the signal-to-noise ratio and the integrity of the acquisition. The acoustic signal confidence weight Wa is determined based on the background noise level and spectral clarity. The confidence weight Wi for surface features is determined based on illumination conditions and image resolution. Set the weighted fusion decision function ; Where Pv is the confidence level of the resonant frequency, Pa is the confidence level of the acoustic signal, Pi is the confidence level of the surface features, and P is the confidence level of the overall anomaly. Pt is set as the judgment threshold, which is dynamically adjusted based on local environmental conditions and equipment conditions. When the comprehensive anomaly confidence level meets P≥Pt, it is determined that a blade anomaly has been initially identified, and the feature data is triggered to be uploaded to the cloud platform.

7. The operation and maintenance method for a small wind power device according to claim 1, characterized in that, When an anomaly is detected, the associated feature data is uploaded to the cloud platform, and the digital twin is driven to perform simulation and deduction, outputting fault diagnosis and lifespan prediction, specifically including: The detection node will package and upload the multimodal feature data, the corresponding original data fragments, and the equipment operating condition data within a specific time window before and after the triggering of the abnormal moment to the cloud platform; After receiving the corresponding raw data fragments and equipment operating condition data, the cloud platform uses the data as boundary conditions and load inputs to drive the digital twin of the wind turbine blades to perform high-fidelity simulation calculations, simulating the evolution process of abnormal features under actual operating conditions. Based on simulation results and combined with a historical failure case library, the system outputs diagnostic results on crack type, location, depth, or icing thickness and area by combining physical mechanisms with data-driven approaches. Based on the diagnostic results, accelerated life test simulations are performed using the digital twin to predict the remaining service life and failure probability of the wind turbine blades under a given operating strategy.

8. The operation and maintenance method for a small wind power device according to claim 7, characterized in that, The simulation covers the evolution of abnormal characteristics under actual operating conditions, specifically including: The operating condition data of the equipment is mapped to the simulation boundary conditions of a digital twin; Based on the identified anomaly type and confidence level, anomaly features are initialized at the corresponding positions in the digital twin; Perform coupled simulation under the stated boundary conditions: Based on the results of coupled simulation, the evolution of anomalous features can be extrapolated over time. Record key physical parameters throughout the entire evolution process, including maximum equivalent stress, fatigue damage index, and crack tip stress intensity factor, as a basis for diagnosis and life prediction.

9. A method for operation and maintenance of a small wind power device according to claim 7, characterized in that, Predicting the remaining service life and failure probability of the wind turbine blades under a given operating strategy, specifically including: Based on the historical operating data of the wind turbine and the local environmental conditions, a typical load spectrum representing future operating conditions is constructed; Simulations are performed in a digital twin to accelerate the application of the load spectrum over time and quantify the degree of damage to the wind turbine. Failure criteria are continuously evaluated during the simulation process; Record the simulation time from the current state to reaching any failure criterion. This time, after conversion, is the predicted remaining service life. Output the curves showing how the remaining useful life and failure probability change over time, generate a remaining useful life prediction report, and upload it to the cloud platform.

10. A small wind power equipment operation and maintenance system, applied to the operation and maintenance method of a small wind power equipment as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect vibration, acoustic, and image data of the blades through detection nodes deployed at the wind turbine site; The model building module is used to build digital twins of wind turbine blades; The data analysis module uses a lightweight AI model deployed on the detection node to perform real-time analysis on the collected data, enabling preliminary identification of leaf anomalies. The simulation and deduction module, when an anomaly is detected, uploads the associated feature data to the cloud platform and drives the digital twin to perform simulation and deduction, outputting fault diagnosis and lifespan prediction. The operation and maintenance decision module is used to automatically generate and issue operation and maintenance decision instructions based on the results of the diagnosis and life prediction.

Images