A wind power blade operation and maintenance method and device based on aerodynamic thermal field inversion

Abstract

The application discloses a wind power blade operation and maintenance method and device based on aerodynamic thermal field inversion. The application comprises: collecting wind power blade multi-position temperature data and unit operation parameters, and obtaining standardized aerodynamic thermal response sequences through time synchronization and working condition normalization processing. Individualized aerodynamic thermal baseline models are constructed relying on blade normal operation historical data, and the change law of the blade thermal field under normal working conditions is described. The real-time standardized thermal data is compared with the baseline model, the thermal response deviation of each position is calculated, and the spatial and temporal characteristics are extracted to locate the persistent thermal anomaly area. The heat conduction diffusion constraint is constructed in combination with the geometric structure and material heat conduction characteristics of the blade, the surface thermal anomaly is mapped to the internal structure of the blade, and the probabilistic potential structural damage area is output. The application considers the whole blade thermal field monitoring and internal defect reverse deduction, accurately distinguishes the working condition fluctuation and the real damage, and realizes the early warning of the hidden structural defects of the blade in a closed loop.

Description

Technical Field

[0001] This invention relates to the field of wind turbine blade operation and maintenance technology, and in particular to a wind turbine blade operation and maintenance method and device based on aerodynamic thermal field inversion. Background Technology

[0002] As a key load-bearing component of wind turbine generators, the operating status of wind turbine blades directly affects the safety and power generation efficiency of the entire unit. Among related technologies, a blade condition assessment system has been constructed through the synergistic application of infrared thermal imaging, visual inspection, and vibration monitoring. Specifically, this system covers key aspects such as surface temperature distribution, identification of external defects, and overall dynamic response. Infrared thermal imaging technology indirectly identifies abnormal areas by collecting data on the blade surface temperature distribution; visual inspection targets obvious damage such as surface cracks; and vibration monitoring reflects changes in the overall structural dynamic characteristics.

[0003] However, existing monitoring methods directly process single observations without considering the inherent correlation between aerodynamic boundary layer development and structural state changes. This can lead to temperature field fluctuations masking structural anomalies under complex wind conditions, or difficulty in reliably extracting effective features related to early damage, thus affecting the reliability and engineering applicability of the detection results. Specifically, infrared thermal imaging is greatly affected by environmental radiation and wind speed disturbances, visual inspection has limited ability to identify early latent damage, vibration monitoring lacks sufficient ability to spatially locate local damage, and existing methods generally lack individualized baseline modeling mechanisms for single blades, resulting in poor comparability across different wind conditions and turbine units. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose a wind turbine blade operation and maintenance method based on aerothermal field inversion.

[0006] Another objective of this invention is to propose a wind turbine blade operation and maintenance device based on aerodynamic thermal field inversion.

[0007] The third objective of this invention is to provide a computer device.

[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, a first aspect of the present invention proposes a wind turbine blade operation and maintenance method based on aerothermal field inversion, comprising: S1. Collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating status parameters. Perform time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerodynamic thermal response data sequence. S2. Based on historical data from the healthy operation phase of the blade, an individualized aerodynamic baseline model is established. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. S3. The standardized aerodynamic thermal response data collected in real time is compared with the baseline model to calculate the thermal response deviation at each spatial location, and the spatial distribution offset characteristics and temporal stability characteristics are extracted to identify persistent thermal response anomaly areas. S4. Based on the blade structure geometry model and material thermal conductivity, establish the diffusion path constraint relationship of thermal response inside the blade, map the thermal response anomaly region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

[0010] In one embodiment of the present invention, S1 includes: The temperature response data and the operating status parameters are time-synchronized based on timestamp correction to obtain multi-source fused data under a unified time reference. The multi-source fused data is subjected to joint outlier removal based on statistical bias and temporal continuity to obtain a clean data sequence; The clean data sequence is normalized based on the thermal response benchmark model under healthy conditions to obtain a standardized aerodynamic thermal response data sequence.

[0011] In one embodiment of the present invention, the time synchronization processing of the temperature response data and the operating status parameters based on timestamp correction includes: The time offset of each acquisition device is calculated using a periodic synchronization signal, and the local time of each acquisition device is corrected to a unified coordinated time based on the time offset. The corrected temperature response data and operating status parameters are mapped to a unified time window, and data with different sampling frequencies are converted to a unified time axis through time interpolation, forming multi-source fused data under a unified time reference.

[0012] In one embodiment of the present invention, S2 includes: The leaf monitoring area is spatially discretized, dividing the leaf into multiple spatial nodes; Based on historical data from the healthy operation phase of the blades, a statistical model of the thermal response of each space node under different operating conditions is established, forming an individualized aerothermal baseline model that characterizes the inherent law of the change of the thermal field of the blades under normal service conditions with operating conditions.

[0013] In one embodiment of the present invention, S3 includes: Calculate the thermal response deviation between the real-time thermal response and the baseline model, and calculate the thermal field gradient change based on the thermal response deviation; The time stability index of thermal response deviation within the calculation time window.

[0014] In one embodiment of the present invention, it further includes: When a region simultaneously meets the conditions of persistent thermal response deviation, abnormal change in thermal field gradient, and time stability index exceeding a preset threshold, the region is identified as a persistent thermal response anomaly region.

[0015] In one embodiment of the present invention, S4 includes: Based on the blade structure geometry model and material thermal conductivity, the diffusion path constraint relationship of thermal response inside the blade is established. Using thermal response deviation and thermal field gradient change as inputs, combined with blade material parameters, the anomalous probability of each structural region inside the blade is calculated through a fusion mapping function, and the potential structural anomaly influence area is output as a probability distribution.

[0016] To achieve the above objectives, a second aspect of the present invention provides a wind turbine blade operation and maintenance device based on aerodynamic thermal field inversion, comprising: The multi-source data standardization module is used to collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating status parameters. The module performs time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerothermal response data sequence. The blade aerothermal baseline modeling module is used to establish an individualized aerothermal baseline model based on historical data from the healthy operation phase of the blade. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. The thermal response anomaly identification module is used to compare the standardized aerodynamic thermal response data collected in real time with the baseline model, calculate the thermal response deviation at each spatial location, extract spatial distribution offset features and temporal stability features, and identify persistent thermal response anomaly areas. The internal structural defect mapping module is used to establish the diffusion path constraint relationship of thermal response inside the blade based on the blade structure geometry model and material thermal conductivity characteristics, map the thermal response anomaly region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

[0017] The present invention provides a wind turbine blade operation and maintenance method and device based on aerodynamic thermal field inversion, which can improve the anti-interference ability and positioning accuracy of wind turbine blade structural anomaly identification and achieve stable regional anomaly identification under complex wind conditions.

[0018] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, for implementing the method described in the first aspect embodiment.

[0019] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] Figure 1 This is a flowchart of a wind turbine blade operation and maintenance method based on aerothermal field inversion according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a wind turbine blade operation and maintenance system based on aerothermal field inversion according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating a wind turbine blade operation and maintenance method based on aerothermal field inversion according to an embodiment of the present invention. Figure 4 This is a structural diagram of a wind turbine blade operation and maintenance device based on aerothermal field inversion according to an embodiment of the present invention; Figure 5 It is a computer device according to an embodiment of the present invention. Detailed Implementation

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] The following description, with reference to the accompanying drawings, describes a wind turbine blade operation and maintenance method and apparatus based on aerodynamic thermal field inversion according to an embodiment of the present invention.

[0025] Figure 1 This is a flowchart of a wind turbine blade operation and maintenance method based on aerothermal field inversion according to an embodiment of the present invention, as shown below. Figure 1 As shown, it includes: S1. Collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating status parameters. Perform time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerodynamic thermal response data sequence. S2. Based on historical data from the healthy operation phase of the blade, an individualized aerodynamic baseline model is established. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. S3. The standardized aerodynamic thermal response data collected in real time is compared with the baseline model to calculate the thermal response deviation at each spatial location, and the spatial distribution offset characteristics and temporal stability characteristics are extracted to identify persistent thermal response anomaly areas. S4. Based on the blade structure geometry model and material thermal conductivity, establish the diffusion path constraint relationship of thermal response inside the blade, map the thermal response anomaly region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

[0026] The data processing model in this invention is not a general artificial intelligence classification model, but an aerothermal field spatiotemporal physical constraint network (PSTN) designed for the wind turbine blade operation status perception scenario. This network includes an aerothermal field spatial coding module, a time evolution feature extraction module, a physical constraint mapping module, and an abnormal region probability output module.

[0027] like Figure 2 As shown, the present invention provides a wind turbine blade damage area identification system based on aerodynamic thermal field evolution, including a distributed temperature acquisition unit 1, an operating status acquisition unit 2, a data fusion processing unit 3, and an aerodynamic thermal field modeling and inversion analysis unit 4. The units are connected through data communication to jointly complete the blade operating status data acquisition, thermal response feature extraction, and abnormal area identification process.

[0028] Figure 2 This section primarily demonstrates the interconnections between the various modules of the system of this invention. The distributed temperature acquisition unit 1 is located inside the wind turbine blade's internal beam cap region or near the composite material region on the blade surface, used to acquire temperature response data at different spatial locations on the blade. This unit includes multiple temperature detection nodes, distributed along the blade's spanwise and chordal directions to form a spatial acquisition network for the blade's thermal response. The operating status acquisition unit 2 is connected to the wind turbine's main control system, used to acquire operating parameters during blade operation, including impeller speed, pitch angle, output power, and equivalent incoming wind speed parameters obtained from wind speed detection equipment or a power inversion model. The operating parameters output by the operating status acquisition unit 2 are synchronized with the temperature acquisition data, providing operating condition constraints for subsequent aerodynamic thermal response analysis. The data fusion processing unit 3 is connected to both the distributed temperature acquisition unit 1 and the operating status acquisition unit 2, used to receive data from different sources and sequentially perform time synchronization, abnormal data removal, and operating condition normalization processing, converting the raw monitoring data into aerodynamic thermal response characteristic data in a unified format. The aerodynamic thermal field modeling and inversion analysis unit 4 is connected to the data fusion processing unit 3. It is used to establish a baseline model of the blade health status based on the processed thermal response data, and to extract abnormal thermal response characteristics through the aerodynamic thermal field evolution analysis model. Combined with the thermal conduction constraint relationship of the blade structure, it realizes the identification of abnormal areas and status assessment. Figure 2 This embodies the overall system architecture of the invention, illustrating the complete process from data acquisition and fusion processing to model analysis and result output.

[0029] like Figure 3As shown, this invention establishes a healthy state thermal response baseline model by fusing blade operating status data and aerodynamic thermal response data, and extracts anomalous thermal response features using an aerodynamic thermal field spatiotemporal evolution model. Furthermore, by combining structural heat conduction constraints, it achieves a mapping from surface thermal response changes to the structural anomaly influence region. First, the system collects temperature field data and corresponding operating parameters during blade operation, forming a multi-source monitoring dataset. Then, the collected data undergoes time synchronization processing, ensuring that temperature data and wind turbine operating status parameters correspond to the same time reference. Simultaneously, an anomaly data detection algorithm removes invalid data caused by sensor communication anomalies or transient interference, and an operating condition normalization method eliminates the influence of wind speed, rotational speed, and ambient temperature changes on the thermal response. The processed data is then input into the aerodynamic thermal field spatiotemporal evolution model. This model first encodes the spatial temperature distribution of the blade to obtain spatial thermal response features at different spanwise and chordal positions; subsequently, a time evolution analysis module extracts the dynamic features of thermal response changes over operating time, forming thermal response evolution features that characterize the normal operating behavior of the blade. Furthermore, the real-time thermal response characteristics are compared with the health status baseline model to calculate the thermal response deviation. Then, the obtained thermal response deviation is input into the structural constraint mapping module, which uses a blade material heat conduction model to constrain the propagation law of the thermal response within the blade structure, obtaining the probability distribution of the anomalous influence area. Finally, the output is the blade anomalous area identification result, including the spatial location of the anomalous area, the degree of anomalousness, and the health status evaluation result.

[0030] Furthermore, in the aerothermal field spatial coding module, to address the issue of the obvious spatial continuity and spanwise and chordal distribution characteristics of the blade aerothermal response, this module adopts an improved two-dimensional spatial feature coding structure instead of directly extracting temperature features through traditional two-dimensional convolution.

[0031] The specific structure includes: Input layer: Input data is represented as: X t =[ T(x,y,t) , V(t) , Ohm(t) , θ(t) , T env ] in: T(x,y,t) This represents the temperature field in the blade space; V(t) Indicates the operating wind speed; Ohm(t) Indicates the impeller speed; i (t) Indicates the pitch angle; T env Indicates ambient temperature.

[0032] Understandably, the spanwise and chordwise positions of the blades are first introduced into the network through a spatial position encoding layer: P ( x,y )= MLP ( x,y ) Forming spatially enhanced input: X t '= X t + P (x,y) Understandably, this structure differs from ordinary convolutional networks in that it only utilizes local neighborhood information, enabling the network to perceive the differences in thermal response caused by aerodynamic load differences at different locations on the blade.

[0033] Subsequently, a multi-scale thermal field feature extraction structure was employed: F s = Concat ( F 1, F 2, F 3) in: F 1: Extract local temperature gradient changes; F 2: Extract regional thermal diffusion characteristics; F 3. Extracting the overall trend of change across the entire spectrum. This module improves the ability to identify small-scale local anomalies and long-distance thermal response changes through multi-scale spatial feature fusion.

[0034] Furthermore, in the time evolution feature extraction module, since the blade thermal response has obvious dynamic change characteristics, this invention does not adopt a single time point judgment method, but constructs a time evolution analysis network.

[0035] The input is a spatial feature sequence: F =[ F t n ,..., F t ] Adopt a gated time state update structure: H t = s ( W h [ F t, H t 1]+ b h ) in: H t Indicates the current time status; F t This indicates the thermal field characteristics at the current moment.

[0036] Furthermore, a time-attention mechanism is introduced to adaptively adjust the importance of different time stages:

[0037] The final time feature is represented as:

[0038] This improved structure differs from ordinary recurrent networks that rely solely on fixed time weights. It can automatically enhance the weights of time periods corresponding to persistent structural anomalies while reducing the impact of short-term disturbances such as gusts.

[0039] Furthermore, in the aerothermal constraint mapping module, since the blade thermal response is constrained by the heat conduction equation, this invention introduces the physical relationship of heat conduction into the network training process.

[0040] Heat transfer within the blade satisfies:

[0041] in: r The density of the material; c p Specific heat capacity; k Thermal conductivity; Q This is a heat source term.

[0042] Network output structure-state mapping results: S = f ( F T ) in: S ( x,y,z ) represents the probability of anomalies at different spatial locations.

[0043] To avoid the network relying solely on data fitting, a physical constraint loss is introduced:

[0044] The final network training loss function is designed as follows: L = L d + λL p +μ L s in: L d : Data prediction error;L p Physical consistency error; L s λ: Spatial continuity constraint; μ: Weighting coefficients. This design ensures that the network output simultaneously satisfies both data regularity and the physical laws of heat conduction, thereby improving the reliability of the inversion results.

[0045] Furthermore, the anomaly region probability output module outputs the final anomaly probability for different regions of the blade:

[0046] in: P i This represents the probability of an anomaly in the i-th structural region; N This indicates the number of regions divided into on the leaf blade. Based on probability results, these regions are categorized as: high-probability abnormal regions; medium-probability affected regions; and low-probability normal regions.

[0047] Furthermore, the network training process and model training process include the following steps: First, aerodynamic thermal response data of the blades under healthy operating conditions are collected, including parameters such as temperature field, wind speed, rotational speed, blade pitch angle, and ambient temperature, to establish a health state database. Second, the blade operating data is spatially discretized, dividing the blades into multiple computational regions, each corresponding to a temperature response node. Then, historical operating data is input into the PSTN network, and the predicted structural state is obtained through forward propagation. S = f ( X ; i )in: i This represents the network parameters. Subsequently, the error is calculated based on the real-state labels or simulated damage area labels: Combine thermal conduction constraint error to jointly optimize network parameters: i = i or L in: or This represents the learning rate. The final result is a network model that satisfies the aerodynamic thermal response law and the mapping relationship between the structural state.

[0048] It is understood that the method described in this invention addresses the problem of monitoring the structural state of wind turbine blades under actual operating conditions. By constructing a spatiotemporal evolution model of aerodynamic thermal response, it achieves regional-level identification and state assessment of potential structural anomalies in the blade. This method relies on the natural aerodynamic frictional heat transfer mechanism between the blade and the incoming airflow during operation, using the surface and near-surface temperature response of the blade as an indirect representation of the aerodynamic boundary layer development state. Combined with structural heat conduction constraints, it performs a restricted mapping of abnormal thermal responses, thereby obtaining the influence area of ​​internal structural anomalies within the blade.

[0049] It is understood that the aerodynamic thermal response spatiotemporal evolution model constructed in this invention is used to describe the dynamic relationship between aerodynamic load changes, surface frictional heat transfer, and internal heat conduction processes of wind turbine blades under different operating conditions. This model does not directly classify and judge temperature data, but rather establishes a mapping relationship between aerodynamic thermal response and time and space by integrating blade operating parameters, spatial temperature distribution characteristics, and the physical laws of heat conduction. The aerodynamic thermal response spatiotemporal evolution model includes a spatial thermal field feature extraction module, a time evolution feature update module, and a physical constraint state mapping module.

[0050] Furthermore, the aerodynamic thermal response data is input into the model. First, the monitoring area of ​​the blade is spatially discretized, dividing the blade into multiple spatial nodes: Ω={( x i , y i , z i ),i=1,2,...,N} in, x i , y i , z i Let represent the spatial location of the i-th monitoring area in the blade coordinate system, and N represent the number of spatially discrete nodes. The data collected in each time period is represented as follows: X t =[ T i (t), V (t), oh (t), β (t), T a (t)] in: T i (t): Temperature response of the i-th region; V (t): Operating wind speed; oh (t): Impeller speed; β (t): Pitch angle; T a (t): Ambient temperature. A multidimensional aerothermal response matrix for blade operation is constructed using the above data: X (t)=[ T 1(t) T 2(t)... T N (t)].

[0051] Furthermore, for the spatial thermal field feature extraction model, since the aerothermal response of the blades exhibits significant spatial continuity and is influenced by local flow states, curvature variations, and boundary layer development at different spanwise locations, a thermal field encoding model based on spatial neighborhood relationships is adopted. First, the spatial feature matrix is ​​constructed: F s (t)= f s ( X (t), P Where: P is the spatial position code of the blade; f s This is a spatial feature mapping function. The spatial location encoding is represented as: P i = MLP ( x i , y i , z i By incorporating spatial coordinate information into the thermal field data, the model can distinguish the differences in normal thermal response caused by different aerodynamic environments in the blade root, mid-blade, and blade tip regions. The spatial feature extraction process is represented as follows: F si = W s X i + b s .in: W s Spatial feature weights; b s This is the bias term. After spatial encoding, the aerothermal field features containing positional relationships are obtained: F s =[ F s 1 , F s 2 ,..., F s N ].

[0052] Furthermore, regarding the time evolution model, since the thermal response caused by abnormal blade structure typically exhibits continuous evolutionary characteristics, while temperature changes caused by factors such as gusts have short-term randomness, this invention further establishes a time evolution model. Spatial features within a continuous time window are input into the time update module: F T =[ F s ( t k),..., F s ( t The thermal response change is described using a gated state update method: H t = g ( F s ( t ), H t 1) Among them: H t Indicates the current time status; H t 1 represents the previous state. The specific update process is as follows: Z t =σ( W z [ F s (t), H t 1]+ b z ) R t =σ( W r [ F s (t), H t 1]+ b r )

[0053]

[0054] in: Z t To update the door; R t To reset the gate. Through the above calculations, the model is able to learn the trend of aerothermal field changes over time.

[0055] Furthermore, the aerothermal field anomaly characteristics are calculated, and a baseline model of the healthy state is established: T 0( x , y , t This represents the theoretical thermal response under the corresponding operating condition where the blades show no obvious abnormalities. The real-time operating thermal field is as follows: T r ( x , y ,t Calculate the thermal response deviation: Δ T ( x , y , t )= T r ( x , y,t ) T 0( x,y,t Further calculations of the thermal field gradient change:

[0056] Used to describe continuous changes in a local thermal field.

[0057] Simultaneously calculate the time stability index:

[0058] Where M represents the length of the time window. When a region simultaneously meets the following conditions: temperature deviation persists; gradient changes abnormally; and time stability exceeds a threshold, then the region is considered to have structural anomalies.

[0059] In one embodiment of the present invention, in order to avoid generating non-physical results by relying solely on data models, the present invention introduces thermal conduction constraints on the blade structure to limit the propagation process of thermal response to the internal structural region.

[0060] The composite material structure inside the blade satisfies the unsteady-state heat conduction equation:

[0061] in: r : Density of blade material; c p Specific heat capacity of the material; k Material thermal conductivity; Q Internal heat source item.

[0062] For wind turbine blades, the main heat source comes from: Q = q air in: q air = h ( T s T a This represents the equivalent heat flux generated by aerodynamic heat transfer. Where: h The aerodynamic heat transfer coefficient; T s This refers to the surface temperature of the blade. T a The ambient temperature.

[0063] Furthermore, in the physical constraint mapping process, according to the heat conduction equation, surface thermal anomalies do not directly correspond to an internal point, but propagate along the material's heat diffusion path. Therefore, a spatial mapping function is established: P s = f (Δ T , G , k , r , c p ) Where: Δ T This refers to thermal response deviation; G This is due to the change in thermal gradient; k、p、c p These are the material parameters. The final probability of anomalies in the internal region is obtained as follows: R ( x , y , z )= Softmax ( F ( H t , P s Where: R represents the anomaly probability of different structural regions; F represents the fusion mapping function. Model output: R =[ r 1, r 2,..., r m ]in ri This represents the probability that the i-th structural region has an abnormal influence.

[0064] Further, the model training process. During the model training phase, a training set is constructed using both healthy state data and artificially simulated anomaly data. Input: D ={ X(t ), T(x,y,t Output: Y ={ R ( x,y,z The training objective function consists of three parts: Data fitting loss:

[0065] This is used to ensure that the model output is consistent with the sample state.

[0066] Physical constraint loss of heat conduction:

[0067] This is used to ensure that the model conforms to the laws of heat conduction.

[0068] Spatial continuity constraints:

[0069] This is used to avoid random jumps in abnormal regions that do not conform to the continuity of the blade structure. The final comprehensive loss function is: L = L d + l 1 L p + l 2 L s By iteratively optimizing the network parameters, the model can simultaneously satisfy both data characteristics and physical constraints.

[0070] Furthermore, in terms of system structure, this invention includes a distributed temperature acquisition unit, an operational status acquisition unit, a data fusion processing unit, and an aerodynamic thermal field modeling and inversion analysis unit. The distributed temperature acquisition unit is deployed within the beam cap region or near-surface composite material structure of the wind turbine blade, forming a multi-point array along the blade spanwise and chordwise directions to ensure continuous acquisition of temperature response changes at different structural locations. The operational status acquisition unit interacts with the wind turbine main control system to synchronously acquire blade rotational speed, pitch angle, and equivalent incoming wind speed parameters estimated based on the nacelle anemometer or power inversion model, thereby providing operating condition constraints for thermal response analysis. The data fusion processing unit performs time synchronization, anomaly removal, and operating condition normalization on multi-source data to eliminate interference from sampling delays and wind condition fluctuations. The aerodynamic thermal field modeling and inversion analysis unit performs subsequent baseline modeling, feature extraction, and structural mapping calculations; this unit can be deployed on nacelle edge computing devices or cloud servers.

[0071] In one embodiment of the present invention, the operation status acquisition unit establishes a data interaction connection with the wind turbine main control system through an industrial communication interface to achieve synchronous correlation between blade thermal response data and unit operation status parameters. Specifically, the operation status acquisition unit obtains real-time operating parameters of the unit through a data communication interface reserved in the wind turbine control system. This communication interface can adopt industrial fieldbus or Ethernet communication methods, such as CAN communication, Modbus communication, or industrial Ethernet communication protocols. The blade thermal response data acquired by the temperature acquisition unit is edge-aggregated by the data acquisition module and matched with the operating parameters output by the main control system according to a unified timestamp, thereby forming a joint data sequence containing thermal field information and operation status information.

[0072] The specific data interaction process is as follows: The temperature acquisition unit acquires temperature signals from different spatial locations of the blade according to a preset sampling period: D T={ T i ( t k ),i=1,2,...,N} where, T i ( t k ) indicates that the i-th measurement point is at time t k Temperature values ​​collected at all times N This indicates the number of temperature measurement points.

[0073] Simultaneously, the operation status acquisition unit reads the corresponding operating parameters from the wind turbine main control system: D M ={ V ( t k ), oh ( t k ), β ( t k ), P ( t k )}in: V ( t k ): Equivalent incoming air velocity; oh ( t k ): Impeller speed; β ( t k ): Pitch angle; P ( t k Output power.

[0074] Subsequently, data associations were established using a unified time stamp: D ( t k )=[ T 1( t k ),..., T N ( t k ), V ( t k ), oh ( t k ), β ( t k ), P ( tk )] This results in multi-source fused data at a single time point. Since the temperature acquisition system and the fan main control system typically have different sampling frequencies, time alignment is required. For example, when the temperature sampling period is: Δ... t T The sampling period of the main control system is: Δ t M At the same time, data with different sampling frequencies are converted to a unified time axis using time interpolation methods: in, t 1. t 2 represents two adjacent sampling time points. After time matching, each temperature response data point corresponds to a unique blade operating state, achieving data coupling between aerothermal response and operating conditions.

[0075] In one embodiment of the present invention, since there are differences in the sampling clocks of the blade temperature sensor, the main control system, and the environmental monitoring equipment, the present invention adopts a synchronization method based on timestamp correction.

[0076] First, assign a unified coordination time to all data acquisition devices: t ′= t +Δ t in: t : Device local time; Δ t Device time offset. The time offset is calculated using a periodic synchronization signal: Δ t = t s t i in: t s Standard time; t i Equipment time.

[0077] After time calibration, data from different sensors are mapped to a unified time window:

[0078] Where: τ: Length of the analysis time window. Data within the same time window are considered to belong to the same operating state, thereby reducing errors caused by asynchronous sampling.

[0079] In one embodiment of the present invention, due to the complex operating environment of wind power, temperature sensors may experience instantaneous jumps, communication anomalies, or measurement drift; therefore, it is necessary to identify anomalies in the collected data. The present invention employs a joint anomaly judgment method based on statistical bias and temporal continuity. First, the average temperature within the sliding window is calculated:

[0080] Calculate the standard deviation:

[0081] If the current sampled value satisfies:

[0082] The data is then considered an outlier.

[0083] Where: K is the anomaly detection threshold; M is the sliding window length.

[0084] Further analysis is conducted based on the rate of temperature change.

[0085] when: | R i ( t )∣> R max If the change does not conform to the current pattern of wind changes, then the point is considered to be abnormal.

[0086] For identified anomalous data, interpolation replacement with data from nearby time periods is used:

[0087] To ensure the continuity of the thermal field.

[0088] In one embodiment of the present invention, since the aerodynamic thermal response of the blade is affected by wind speed, rotational speed and ambient temperature, the temperature level varies under different operating conditions, so it is necessary to normalize the operating conditions.

[0089] First, establish a baseline model of the thermal response under healthy conditions: T 0= f ( V , oh , β , T a ) in: T Wind speed; oh Rotational speed; β The pitch angle; T a The ambient temperature.

[0090] The difference between the real-time thermal response and the reference response is defined as: Δ T i ( t )= Ti ( t ) T 0( V , oh , β , T a ) This treatment eliminates temperature changes caused by variations in normal operating conditions.

[0091] Further define the normalized thermal response:

[0092] Where σ0 represents the standard deviation of temperature fluctuation under the corresponding operating condition in a healthy state. After normalization: Normal operating state: T i ≈0 indicates a persistent abnormal region: |T i The above method enables thermal response data obtained under different wind speeds and rotational speeds to be compared on a unified scale.

[0093] In one embodiment of the present invention, regarding its working principle, the invention is based on the coupling relationship between the boundary layer flow state and surface heat transfer behavior during blade operation. That is, the aerodynamic boundary layer formed during blade rotation directly affects the local frictional heat transfer intensity, thereby forming a thermal response field with spatial distribution characteristics inside the blade. When there is local damage, roughness change, or material property degradation on the blade surface, the boundary layer development state in the corresponding region will shift, resulting in a continuous deviation between the local thermal response and the baseline characteristics under healthy conditions. Therefore, this thermal response deviation can serve as an indirect characterization signal of structural state changes.

[0094] Understandably, during data acquisition, the system continuously collects temperature response data from various measuring points on the blades while the wind turbine is operating normally, and simultaneously records the corresponding operating status parameters, thus forming multidimensional time-series data containing temperature field and operating condition variables. To ensure data comparability, the acquired data is first time-synchronized to align data from different sampling channels under a unified time reference. Secondly, high-frequency noise is suppressed using a filtering algorithm, and further, operating condition normalization is performed on the temperature data by combining parameters such as rotational speed and wind speed, thereby obtaining a standardized data sequence that reflects the true changes in thermal response.

[0095] Understandably, during the baseline model construction process, historical data from when the blades are in a healthy operating phase are selected as training samples. By establishing statistical models of thermal response under different operating conditions, an individualized aerothermal baseline model is formed to characterize the inherent laws governing the change of the blade's thermal field with operating conditions under normal service. This baseline model not only includes the spatial characteristics of temperature distribution but also its dynamic evolution with changes in rotational speed and wind speed, thus providing a reference for subsequent anomaly detection.

[0096] Understandably, during the anomaly feature extraction process, real-time acquired thermal response data is compared and analyzed with the baseline model to calculate the thermal response deviation value at each spatial location. Based on this, spatial distribution asymmetry features and temporal stability features are extracted. The spatial distribution features describe the changes in temperature difference structure between different spanwise and tangential regions of the blade, while the temporal stability features determine whether the deviation is persistent, thus distinguishing between changes caused by transient aerodynamic disturbances and structural anomalies. Simultaneously, by combining the local thermal response intensity variation trend, potential anomaly regions are preliminarily identified.

[0097] Understandably, in the structural mapping and inversion analysis process, based on the blade's geometric model and the material's thermal conductivity, a constraint relationship for the diffusion path of the thermal response within the blade is established. Based on this, the surface or near-surface thermal response deviation is mapped to the structural domain space, thereby obtaining the distribution of the influence area of ​​potential structural anomalies. This mapping process does not aim to determine the location of a single point, but rather outputs regional results based on a probability distribution to adapt to the uncertainty of wind conditions and the information ambiguity brought about by the thermal diffusion process.

[0098] Understandably, during the condition assessment process, the results of abnormal regions within multiple consecutive time windows are cumulatively analyzed. If the abnormal characteristics of a certain region show stability across multiple time scales, it is identified as a structurally abnormal region; if the abnormal characteristics disappear rapidly with changes in operating conditions, it is classified as a transient anomaly caused by aerodynamic disturbances. Ultimately, the system outputs the overall blade health status assessment results and abnormal region distribution information, providing a basis for wind turbine operation and maintenance decisions.

[0099] In one embodiment of the present invention, the present invention fully considers the basic theory of fluid heat transfer, the anisotropic thermal conductivity of composite materials, and the actual working conditions of engineering applications during the aerodynamic thermal field modeling and defect inversion process, and improves the physical model boundary and long-term operation and maintenance adaptation logic: the aerodynamic heat generated by blade rotation is formed by the coupling of incoming flow deceleration, compression and heating and surface viscous frictional heat transfer; local structural damage will change the boundary layer transition and flow separation state, causing a continuous shift in heat transfer intensity; when constructing the three-dimensional transient unsteady-state heat conduction model, two types of heat source terms are distinguished, and the heat transfer of the airflow on the blade surface is defined as the boundary convection heat flux: q conv = h ( T s - T env ) in, q conv It is the convective heat flux over the blade surface. h It is the convective heat transfer coefficient on the blade surface. T s It is the thermodynamic temperature of the outer surface of the blade. T env It is the thermodynamic temperature of the mainstream airflow environment near the blade wall. T s -T env The excess temperature difference between the blade wall and the external airflow drives convective heat transfer. The thermal effect caused by heat stagnation due to internal defects such as blade delamination and detachment is defined as the volumetric heat source Q. defect Simultaneously, the thermal conductivity tensor is used to characterize the anisotropic thermal conductivity differences of glass fiber / carbon fiber composites along the spanwise, tangential, and thickness directions, avoiding the loss of inversion accuracy caused by a single scalar thermal conductivity coefficient. For the infrared temperature measurement acquisition path, blade surface emissivity and atmospheric environmental radiation correction terms are added to the operating condition normalization model to eliminate biases in the outdoor environment temperature measurement system. At the engineering implementation level, distributed temperature sensors are preferentially and densely deployed along the critical stress areas of the blade cap and web. Wireless passive temperature measurement nodes are used to replace wired sensors in the narrow inner cavity of the blade tip to avoid line damage caused by centrifugal force. An adaptive update mechanism is added to the baseline model, iteratively fine-tuning the individualized aerodynamic thermal baseline using monthly healthy operating interval data to adapt to the overall baseline drift caused by coating aging and uniform increase in surface roughness during blade service. For wind turbine shutdown and power-limited no-wind-load conditions, a static infrared thermal imaging acquisition module is linked to obtain the natural heat dissipation temperature field of the blade, combined with a pure solid heat conduction constraint model under no aerodynamic load to complete continuous defect monitoring, achieving full life-cycle and full-condition coverage. During the model training phase, a triple loss function is introduced, consisting of data fitting loss, thermal conduction physical consistency loss, and spatial continuity constraint loss. Relying on the aerodynamic thermal field spatiotemporal physical constraint network, multi-scale spatial features and temporal attention evolution features are integrated to distinguish between short-term gust turbulence disturbances and structural continuous thermal anomalies from the dual dimensions of spatial gradient and temporal stability, ensuring the physical interpretability and engineering reproducibility of defect identification results under complex wind conditions.

[0100] The embodiments of the present invention also have the following technical effects: This invention realizes an indirect characterization method for blade structural state based on operational aerodynamic thermal response. Compared with traditional detection methods based on infrared thermal imaging, this method does not rely on external heat sources or static imaging conditions. Instead, it utilizes the aerodynamic thermal field information naturally formed during blade operation and combines it with individualized baseline models and structural physical constraints to achieve continuous monitoring and identification of potential damage areas of the blade. It has better adaptability and stability under complex wind conditions.

[0101] This invention introduces a thermal response analysis method based on aerodynamic boundary layer heat transfer mechanisms, combined with an individualized health baseline model and a multi-condition normalization processing mechanism. This transforms thermal response characteristics from single temperature observations into evolutionary feature expressions with physical constraints, significantly improving the anti-interference capability of anomaly identification. Compared to traditional infrared detection methods, this invention does not rely on external thermal excitation or static imaging conditions, maintaining stable feature extraction capabilities even under conditions of large wind speed fluctuations. According to comparative analysis of simulated operating conditions, under wind speed fluctuations ranging from 8–15 m / s, the Interpretation of Value (IoU) for anomaly region identification using this invention is improved from approximately 0.62 using traditional infrared threshold segmentation methods to over 0.80, while the false alarm rate is reduced from approximately 18% to less than 7%. Furthermore, by introducing a thermal response time-scale decomposition mechanism and a structural constraint mapping method, this invention separates transient wind disturbances from structural anomalies, resulting in better stability and repeatability of anomaly identification results. In continuous operation testing, for simulated local structural stiffness degradation or surface damage conditions, the method of this invention can stably capture persistent thermal response shifts within a short time window (approximately 3-5 operating cycles). For abnormal responses caused by short-term gust disturbances, it can effectively suppress their impact on the final judgment result. This invention combines aerodynamic thermal response characteristics with structural geometric constraints to achieve a restricted mapping from surface thermal anomalies to the affected area of ​​the blade structure. This elevates the anomaly identification result from the traditional "point-based judgment" to a "regional probability expression," improving engineering interpretability and operational decision-making reference value. Overall, this invention achieves a simultaneous improvement in the accuracy and stability of wind turbine blade structural anomaly identification without adding additional external excitation, and enhances its engineering applicability in complex meteorological environments.

[0102] To achieve the above embodiments, such as Figure 4 As shown, this embodiment also provides a wind turbine blade operation and maintenance device 10 based on aerothermal field inversion, including: The multi-source data standardization module 100 is used to collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating state parameters. The module performs time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerothermal response data sequence. The blade aerothermal baseline modeling module 200 is used to establish an individualized aerothermal baseline model based on historical data from the healthy operation phase of the blade. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. The thermal response anomaly identification module 300 is used to compare the standardized aerodynamic thermal response data collected in real time with the baseline model, calculate the thermal response deviation at each spatial location, extract spatial distribution offset features and temporal stability features, and identify persistent thermal response anomaly areas. The internal structural defect mapping module 400 is used to establish the diffusion path constraint relationship of thermal response inside the blade based on the blade structure geometric model and material thermal conductivity characteristics, map the thermal response abnormal region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

[0103] An embodiment of the present invention provides a wind turbine blade operation and maintenance device based on aerodynamic thermal field inversion, which can improve the anti-interference capability and positioning accuracy of wind turbine blade structural anomaly identification, and achieve stable regional anomaly identification under complex wind conditions.

[0104] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 5 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the method described above.

[0105] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0106] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0107] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A wind turbine blade operation and maintenance method based on aerothermal field inversion, characterized in that, include: S1. Collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating status parameters. Perform time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerodynamic thermal response data sequence. S2. Based on historical data from the healthy operation phase of the blade, an individualized aerodynamic baseline model is established. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. S3. The standardized aerodynamic thermal response data collected in real time is compared with the baseline model to calculate the thermal response deviation at each spatial location, and the spatial distribution offset characteristics and temporal stability characteristics are extracted to identify persistent thermal response anomaly areas. S4. Based on the blade structure geometry model and material thermal conductivity, establish the diffusion path constraint relationship of thermal response inside the blade, map the thermal response anomaly region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

2. The method as described in claim 1, characterized in that, S1 includes: The temperature response data and the operating status parameters are time-synchronized based on timestamp correction to obtain multi-source fused data under a unified time reference. The multi-source fused data is subjected to joint outlier removal based on statistical bias and temporal continuity to obtain a clean data sequence; The clean data sequence is normalized based on the thermal response benchmark model under healthy conditions to obtain a standardized aerodynamic thermal response data sequence.

3. The method as described in claim 2, characterized in that, The time synchronization processing of the temperature response data and the operating status parameters based on timestamp correction includes: The time offset of each acquisition device is calculated using a periodic synchronization signal, and the local time of each acquisition device is corrected to a unified coordinated time based on the time offset. The corrected temperature response data and operating status parameters are mapped to a unified time window, and data with different sampling frequencies are converted to a unified time axis through time interpolation, forming multi-source fusion data under a unified time reference.

4. The method as described in claim 1, characterized in that, The S2 includes: The leaf monitoring area is spatially discretized, dividing the leaf into multiple spatial nodes; Based on historical data from the healthy operation phase of the blades, a statistical model of the thermal response of each space node under different operating conditions is established, forming an individualized aerothermal baseline model that characterizes the inherent law of the change of the thermal field of the blades under normal service conditions with operating conditions.

5. The method as described in claim 1, characterized in that, The S3 includes: Calculate the thermal response deviation between the real-time thermal response and the baseline model, and calculate the thermal field gradient change based on the thermal response deviation; The time stability index of thermal response deviation within the calculation time window.

6. The method as described in claim 5, characterized in that, The method further includes: When a region simultaneously meets the conditions of persistent thermal response deviation, abnormal change in thermal field gradient, and time stability index exceeding a preset threshold, the region is identified as a persistent thermal response anomaly region.

7. The method as described in claim 1, characterized in that, The S4 includes: Based on the blade structure geometry model and material thermal conductivity, the diffusion path constraint relationship of thermal response inside the blade is established. Using thermal response deviation and thermal field gradient change as inputs, combined with blade material parameters, the anomalous probability of each structural region inside the blade is calculated through a fusion mapping function, and the potential structural anomaly influence area is output as a probability distribution.

8. A wind turbine blade operation and maintenance device based on aerothermal field inversion, characterized in that, include: The multi-source data standardization module is used to collect multi-source data of wind turbine blades under operating conditions. The multi-source data includes temperature response data of different spatial positions of the blades and corresponding operating status parameters. The module performs time synchronization and operating condition normalization processing on the multi-source data to form a standardized aerothermal response data sequence. The blade aerothermal baseline modeling module is used to establish an individualized aerothermal baseline model based on historical data from the healthy operation phase of the blade. The baseline model is used to characterize the inherent law of the change of the thermal field of the blade with operating conditions under normal service conditions. The thermal response anomaly identification module is used to compare the standardized aerodynamic thermal response data collected in real time with the baseline model, calculate the thermal response deviation at each spatial location, extract spatial distribution offset features and temporal stability features, and identify persistent thermal response anomaly areas. The internal structural defect mapping module is used to establish the diffusion path constraint relationship of thermal response inside the blade based on the blade structure geometry model and material thermal conductivity characteristics, map the thermal response anomaly region to the internal structural domain of the blade, and output the potential structural anomaly influence region characterized in the form of probability distribution.

9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images