A wind power blade icing identification method and device

Abstract

The application discloses a wind power blade icing identification method and device. The collected blade thermal response is a composite signal coupled by aerodynamic viscous shear dissipation and wall heat exchange. The composite signal is superimposed with environmental temperature, radiation, turbulence, coating aging, dust and dew and other interferences. The interferences are inhibited through multiple pretreatments and working condition compensation. Noise is removed through filtering and temperature drift calibration. Environmental temperature and humidity, wind speed turbulence, rotating speed and variable pitch angle are used as compensation quantities to solve the baseline offset of convective heat exchange and correct the baseline offset. Local thermal distortion caused by boundary layer disturbance is retained. A two-dimensional thermal field is reconstructed according to the local thermal distortion. Spatial structure and periodic similarity are extracted. The deviation from the reference thermal field is compared to generate an abnormal index and monitor the multi-period trend. When the trend continuously increases and the abnormal regions are gathered, icing is determined and early warning is given. Through physical cause and effect driving and double time sequence constraint, the application realizes accurate identification and spatial positioning of early and weak icing under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of wind turbine blade operating status sensing and early icing identification technology, and in particular to a method, device, equipment and storage medium for wind turbine blade icing identification. Background Technology

[0002] As the core component of wind turbines for capturing wind energy, the aerodynamic shape and surface condition of wind turbine blades directly determine the overall power generation efficiency and operational safety of the turbine. In low-temperature, high-humidity, or ultra-high-altitude environments, blade surfaces are highly susceptible to icing. Ice buildup alters the blade's aerodynamic shape and increases surface roughness, leading to decreased lift, increased drag, and exacerbated load fluctuations. In severe cases, it can cause abnormal turbine vibration or even unplanned shutdowns. Therefore, accurate and timely identification of blade icing is crucial for ensuring the safe and stable operation of wind turbines.

[0003] Currently, blade icing monitoring methods are mainly divided into two categories. One category is the indirect judgment method based on environmental parameters and unit operating data. This method assesses icing risk by monitoring parameters such as temperature, humidity, wind speed, and power changes, combined with empirical models or threshold criteria. This type of method relies on the statistical correlation between the external environment and operating data, is significantly affected by fluctuations in unit operating conditions, and the icing process exhibits localized and nonlinear characteristics, making stable and reliable identification difficult in the early stages of icing. The other category is the surface detection method based on infrared thermal imaging or visible light images. This method determines the icing state by imaging and analyzing the temperature distribution or ice layer morphology on the blade surface. However, this type of method is easily affected by meteorological conditions such as rain, snow, fog, and background radiation, and the detection results mostly reflect visible surface icing, lacking sensitivity to initial, weak ice layers. Furthermore, the difficulty in imaging and positioning due to the high-speed rotation of the blades also restricts its engineering application. The aforementioned methods fail to address the following technical challenges: the blade surface thermal response signal is a composite thermal signal formed by the coupling of aerodynamic viscous shear dissipation and multi-source heat transfer on the wall surface. This signal is simultaneously superimposed with various interference components, including ambient temperature, solar radiation, turbulent disturbances, blade coating aging, and surface dust and dew. The thermal characteristic changes caused by icing are weak signals, easily overwhelmed by the aforementioned interferences. Furthermore, in the early stages of icing, there are no significant temperature amplitude changes, but rather subtle changes such as disruption of the thermal gradient topology and phase drift. Existing methods rely on a single temperature threshold or image morphology judgment, making it difficult to effectively separate icing disturbances from non-icing interferences from the composite thermal signal. They also cannot establish an identification mechanism driven by the physical causal chain of icing, resulting in low early-stage weak icing recognition rates and high false alarm rates.

[0004] As wind turbines continue to evolve towards larger capacities and more complex high-altitude environments, icing issues are becoming increasingly uncertain and insidious, making traditional detection methods insufficient for identifying early-stage, subtle icing conditions. There is an urgent need to utilize the physical response information generated by the blades themselves during operation to construct a monitoring mechanism that is more sensitive to changes in the aerodynamic boundary layer, thereby effectively improving the predictability and stability of early-stage icing identification. Summary of the Invention

[0005] The present invention aims to at least partially solve one of the technical problems in the related art.

[0006] To address this issue, this invention proposes a method for identifying icing on wind turbine blades. This method involves collecting the thermal response time-series signal and rotation angle signal from distributed sensing units on the blade surface. The angle is mapped to the corresponding interval, and the two-dimensional thermal field distribution is reconstructed by combining the spanwise position. Spatial structural features and temporal similarity features of adjacent periods of the thermal field are extracted and compared with a preset benchmark thermal field to obtain local structural deviation information. The above information is comprehensively evaluated to generate an overall anomaly index, and its cumulative trend is continuously monitored. When the trend continues to increase and the anomaly shows regional clustering, the icing area is determined, and an early warning signal is output. This invention achieves accurate identification and spatial positioning of early-stage, weak icing.

[0007] Another objective of this invention is to provide a wind turbine blade icing identification device.

[0008] The third objective of this invention is to provide a computer device.

[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0010] To achieve the above objectives, the present invention provides a method for identifying icing on wind turbine blades, comprising: The original thermal response timing signal and blade rotation angle signal are acquired by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, the timing signal of each sensing unit is mapped to the corresponding blade rotation angle range. Combined with the spanwise position information of each sensing unit, the two-dimensional spatial thermal field distribution of the current rotation cycle is reconstructed. Extract the spatial structure features of the two-dimensional thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field. The spatial structural features, temporal similarity features, and local structural deviation information are comprehensively evaluated to generate an evaluation index characterizing the overall degree of thermal field anomaly. The cumulative change trend of the evaluation index is continuously monitored over multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly is regionally clustered in space, it is determined that the corresponding region is in an icing evolution state, and an icing warning signal is output.

[0011] The wind turbine blade icing identification method of this invention may also have the following additional technical features: In one embodiment of the present invention, the process of establishing the reference thermal field distribution includes: Historical thermal response data of the blades were acquired over multiple rotation cycles during which the blades were ice-free and operating stably. This data was then combined with the rotation angle signal and spanwise position information corresponding to each cycle, and analyzed using the formula... Reconstructing the two-dimensional spatial thermal field distribution across multiple historical cycles; The two-dimensional spatial thermal field distributions of each historical period are statistically averaged according to the same angular interval and spanwise position to generate the initial reference thermal field distribution; Acquire new thermal field data of the blade during the long-term stable operation phase, calculate the difference between the new thermal field data and the initial reference thermal field distribution, and when the difference value is less than the preset update threshold, perform progressive weighted correction on the initial reference thermal field distribution using the new thermal field data to generate an updated reference thermal field distribution. The preset update threshold is dynamically generated based on the 3σ interval of the thermal field residual under normal operating conditions. The reference thermal field adopts a layered verification-based progressive update strategy: before updating, the uniformity of the global thermal field offset and the distribution of local gradient distortion are verified simultaneously. Weighted correction is only performed when the deviation is uniform across the entire domain and there are no local concentrated anomalies. If permanent local thermal field offset caused by coating peeling or blade collision is detected, the reference model is immediately locked and the update is stopped, and a blade surface structure anomaly prompt is output simultaneously.

[0012] In one embodiment of the present invention, the process of extracting spatial structural features includes: The two-dimensional spatial thermal field distribution of the current rotation period is obtained as input data, according to the formula. Calculate the rate of change of thermal response between adjacent positions along the spanwise direction of the blade to obtain the spanwise thermal gradient distribution, and then apply the formula... Calculate the rate of change of thermal response between adjacent angular intervals along the blade rotation angle to obtain the circumferential thermal gradient distribution; The spanwise thermal gradient distribution and the circumferential thermal gradient distribution are respectively compared with the reference thermal gradient distribution corresponding to the reference thermal field distribution to obtain the spanwise gradient difference value and the circumferential gradient difference value. Spatially accumulate the spanwise gradient difference value and the circumferential gradient difference value to output a thermal gradient anomaly index that characterizes the degree of disruption of the spatial continuity of the current thermal field.

[0013] In one embodiment of the present invention, the extraction process of temporal similarity features includes: Obtain the two-dimensional spatial thermal field distribution during the current rotation period. Two-dimensional spatial thermal field distribution of the adjacent previous period As input data, the thermal field distribution of the two cycles is mapped point by point according to the same angular interval and spanwise position; The periodic consistency index between two periodic thermal field distributions is calculated using the following formula:

[0014] in, This represents the covariance between two periodic thermal fields. and These represent the standard deviations of the thermal field data for the two periods, respectively. The period consistency index of multiple adjacent period pairs is continuously acquired, arranged in chronological order to form a period consistency sequence, and the period consistency sequence is output as a temporal similarity feature. Before input feature extraction, the thermal response time series signal undergoes multi-layer signal preprocessing and operating condition coupling compensation: second-order low-noise filtering and component temperature drift calibration are performed on the original time series to eliminate sensor jumps and power frequency noise; real-time ambient temperature, ambient humidity, wind speed and turbulence intensity, impeller speed and pitch angle are used as compensation parameters to quantify the baseline offset of global convective heat transfer on the reconstructed thermal field, and global translation correction is performed on the two-dimensional spatial thermal field distribution.

[0015] In one embodiment of the present invention, the process of acquiring local structural deviation information includes: Obtain the two-dimensional spatial thermal field distribution during the current rotation period. and the pre-established reference thermal field distribution As input data, the two thermal field distributions are matched point by point according to the same angular interval and spanwise position; According to the formula Calculate the absolute difference in thermal response between corresponding points to generate the differential thermal field matrix; Spatial connectivity analysis is performed on the differential thermal field matrix to extract connected regions formed by consecutive differential values ​​exceeding a preset threshold. The spatial distribution range and cumulative differential amount of each connected region are used as local structural deviation information output.

[0016] In one embodiment of the present invention, the process of generating evaluation indicators includes: Obtain the degree of thermal gradient anomaly in the current period The degree of decline in periodic consistency and the degree of thermal response phase shift As input data, the three are normalized to ensure that each indicator is on a uniform measurement scale. Calculate the comprehensive anomaly evaluation index using the following formula:

[0017] in, , , These represent the weighting coefficients corresponding to the degree of thermal gradient anomaly, the degree of periodic consistency degradation, and the degree of thermal response phase shift, respectively. Output a comprehensive anomaly evaluation index for the current rotation cycle. ; Among them, all algorithm parameters on which the comprehensive anomaly evaluation index depends are adaptively calibrated based on historical data of stable, ice-free operation of the unit; the weights of three types of features are: thermal gradient anomaly, decreased periodic consistency, and thermal response phase shift. , , The baseline update weights are determined by least squares fitting based on typical icing conditions; during periods of high incidence of low-temperature icing, the baseline update weights are automatically reduced to suppress the interference of waterlogging baseline drift on the evaluation indicators.

[0018] In one embodiment of the present invention, the process for determining the output conditions of the icing warning signal includes: Obtain the comprehensive anomaly evaluation index corresponding to each of multiple consecutive rotation cycles. And arranged in cyclical order to form a time series of evaluation indicators; According to the formula Calculate the current period relative to the previous period Overall trend value of each cycle ,like If the value is positive and maintains positive growth for multiple consecutive periods, then the cumulative change trend of the evaluation indicator is determined to be a continuous growth trend; among which, the number of consecutive rotation periods involved in the trend calculation... The value range is 5 to 30; The spatial distribution of connected regions in the differential thermal field matrix where the difference value continuously exceeds a preset threshold is obtained. When the connected regions exist continuously for multiple periods at the same spatial location and the area shows an expanding trend, it is determined that the anomaly is regionally clustered in space, and an icing warning signal containing the spanwise position and circumferential angle range of the anomaly region is output. The judgment logic incorporates dual time constraints: only when the positive growth trend of the comprehensive anomaly evaluation index is continuously maintained for at least [time period missing]... Only when the abnormally connected regions at the same location have undergone a rotation cycle and simultaneously and continuously expanded, with their areas gradually increasing, is it determined that the corresponding region has entered the icing evolution state.

[0019] To achieve the above objectives, another aspect of the present invention provides a wind turbine blade icing identification device, comprising: The thermal field reconstruction module is used to acquire the original thermal response timing signal and blade rotation angle signal collected by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, the timing signal of each sensing unit is mapped to the corresponding blade rotation angle range, and combined with the spanwise position information of each sensing unit, the two-dimensional spatial thermal field distribution of the current rotation cycle is reconstructed. The feature extraction and comparison module is used to extract the spatial structure features of the two-dimensional spatial thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field. The fusion judgment module is used to comprehensively evaluate the spatial structure features, the temporal similarity features, and the local structure deviation information, generate evaluation indicators that characterize the overall degree of thermal field anomaly, and continuously monitor the cumulative change trend of the evaluation indicators within multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly shows a regional clustering in space, it is determined that there is an icing evolution state in the corresponding region, and an icing warning signal is output.

[0020] This invention discloses a method and apparatus for identifying icing on wind turbine blades. It reconstructs the two-dimensional spatial thermal field distribution of the blade using distributed sensing units, extracts spatial structural features, temporal similarity features, and local structural deviation information relative to a reference thermal field, comprehensively generates an overall anomaly evaluation index, and continuously monitors its cumulative trend. When the trend continues to grow and anomalies exhibit regional clustering, the icing area is determined and an early warning signal is output. This effectively solves the problems of insufficient sensitivity and difficulty in spatial positioning for identifying early, weak icing states in existing technologies. This invention achieves an integrated solution for the entire process from thermal field reconstruction, multi-dimensional feature extraction, anomaly quantification assessment to icing area determination, significantly improving the pre-emptiveness and stability of icing identification, and enhancing the method's adaptability and engineering practical value under complex operating conditions and environmental interference.

[0021] 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 a wind turbine blade icing identification method as described in the first aspect embodiment.

[0022] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements a wind turbine blade icing identification method as described in the first aspect embodiment.

[0023] 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

[0024] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a wind turbine blade icing identification method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the aerothermal field processing flow for early identification of icing on wind turbine blades according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the overall system structure according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a wind turbine blade icing identification device 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

[0025] 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.

[0026] 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.

[0027] The following description, with reference to the accompanying drawings, describes a method, apparatus, device, and storage medium for identifying icing on wind turbine blades according to an embodiment of the present invention.

[0028] Example 1 like Figure 1 , Figure 2 As shown, a wind turbine blade icing identification method of the present invention includes the following steps: S1. Acquire the original thermal response timing signal and blade rotation angle signal collected by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, map the timing signal of each sensing unit to the corresponding blade rotation angle range. Combine the spanwise position information of each sensing unit to reconstruct the two-dimensional spatial thermal field distribution of the current rotation cycle.

[0029] In one embodiment of the present invention, from a physical mechanism perspective, wind turbine blades undergo continuous viscous shearing with the air during rotation, forming a stable boundary layer structure near the blade surface. This boundary layer not only undertakes momentum exchange processes but also involves a certain degree of energy dissipation, resulting in a weak-amplitude but stable structural thermal response field on the blade surface. This thermal response is not a uniform temperature distribution but a spatially non-uniform thermal structure determined by local flow states, blade geometry, and angle of attack variations. It exhibits strong repeatability across multiple rotation cycles and can be considered a "thermal fingerprint" naturally formed by the aerodynamic process.

[0030] It should be noted that, under normal operating conditions, the essential characteristic of this thermal fingerprint lies in its structural stability, that is, the topological relationship of the thermal gradient in space remains stable. Although there are fluctuations in local temperature, the relative relationship between different spatial locations is highly repeatable. This stability comes from the stable coupling relationship between the boundary layer shear structure and the energy dissipation path.

[0031] It is worth noting that when ice forms on the blade surface, the ice layer primarily affects the boundary layer structure itself, rather than directly manifesting as a temperature change. Ice crystal adhesion alters the equivalent roughness of the wall surface, transforming the boundary layer from a relatively stable shear structure to a state of enhanced local disturbance, thus causing a reconstruction of the momentum exchange mode. Simultaneously, as a low-thermal-conductivity medium, the ice layer introduces additional thermal resistance between the wall and the external flow field, shifting the original aerodynamic frictional heat transfer path. This process does not immediately manifest as a significant temperature difference, but rather first causes a slight shift in the spatial phase relationship of heat release; that is, the thermal response at the same rotation angle is delayed or misaligned.

[0032] From a physical causal perspective, the initial thin ice on the blade surface increases the equivalent roughness of the wall, causing the boundary layer transition position to shift forward and abrupt changes in local turbulent heat transfer intensity. Simultaneously, the ice layer, as a low thermal conductivity medium, introduces additional thermal resistance between the wall and the external flow field, shifting the original aerodynamic frictional heat transfer path and causing a synchronous drift in the thermal response phase. This process does not immediately produce a large temperature amplitude change, but rather first disrupts the continuity of the spanwise and circumferential thermal gradients. While short-term deposits such as dew and floating dust cause instantaneous gradient distortion, they do not exhibit multi-period continuous expansion characteristics and can be effectively distinguished from icing disturbances through multi-period cumulative trend constraints. The thermal field deviation caused by coating aging and slight dust accumulation is a globally uniform and gradually varying shift, without forming localized concentrated distortions, and can be distinguished through benchmark update pre-verification logic.

[0033] Based on the above mechanism, this invention no longer treats the blade thermal response as a single-point temperature signal, but analyzes it as a structured thermal field that evolves continuously with blade rotation. By introducing a blade rotation angle synchronization mechanism, the time-series thermal signal is mapped to a quasi-spatial thermal field based on the blade angle, transforming the originally time-varying signal into a spatially comparable thermal structure distribution, thereby enabling the observation of the continuous evolution of the thermal field in the spanwise and circumferential directions.

[0034] Moreover, in this thermal structural field, the key changes in the early stages of icing do not manifest as an overall temperature shift, but rather as three coupled changes: first, a decrease in the continuity of the local thermal gradient structure, i.e., the originally smooth gradient distribution becomes locally distorted; second, a weakening of the periodic repetition consistency of the thermal structure, i.e., the response at the same spatial location deviates in different rotation cycles; and third, a phase drift in the boundary layer thermal response, i.e., the peak position of the thermal response shifts systematically relative to the blade angle position. These three types of changes collectively reflect the local reconstruction of the boundary layer energy dissipation path.

[0035] It is important to further explain that, from a more fundamental perspective, this process can be understood as the disruption of the stability of the path for the conversion of aerodynamic shear energy into thermal energy. Under normal conditions, this conversion path has a stable spatial mapping relationship, meaning there is a stable phase correspondence between energy input, dissipation, and heat release. However, in the early stages of icing, this correspondence first decouples in local regions, causing the thermal field to evolve from a continuous topological structure to a locally unstable structure. This instability does not depend on changes in temperature amplitude but on the disruption of structural correlations, thus exhibiting earlier sensitivity.

[0036] Therefore, based on this mechanism, this invention constructs an identification framework centered on "thermal uniformity evolution stability." Thermal uniformity does not refer to uniform temperature distribution in the traditional sense, but rather to the predictability and repeatability of the thermal gradient structure in space. Under normal operating conditions, although the blade thermal gradient is not uniform, its spatial distribution exhibits a stable topological relationship; however, in the early stages of icing, this topological relationship first decouples in local regions, manifesting as a decrease in gradient correlation and disruption of structural continuity.

[0037] Furthermore, in terms of methodology, a three-dimensional thermal field data structure of "angle-span-time" is constructed by synchronously calibrating the distributed thermal response signal on the blade surface with the rotation angle information. This transforms the thermal response from the time domain to the spatial structural domain, and on this basis, thermal gradient changes, periodic consistency indices, and local structural offsets are extracted. Moreover, by establishing a thermal structural benchmark model under normal operating conditions and introducing a blade-specific thermal fingerprint correction mechanism, the inherent deviations caused by manufacturing errors or differences in surface condition between different blades are eliminated, thus ensuring that the judgment criteria focus on changes in operating conditions rather than structural differences.

[0038] The core of this invention does not rely on a single fixed data processing algorithm, but rather on establishing a mapping relationship between the blade's aerothermal response and icing state to convert the original thermal response signal into thermal field structural characteristics. The specific signal processing procedure is as follows: This invention first acquires raw thermal response data during blade operation using distributed thermal response sensing units arranged on the blade surface. The raw signal can be a temperature signal or an equivalent heat flow response signal. The data collected at time t by the i-th measuring point is denoted as Si(t). Since the wind turbine blade is in a periodic rotation state, the signal collected by a single measuring point simultaneously includes the effects of blade rotation period changes, wind speed fluctuations, ambient temperature changes, and sensor noise. Therefore, the raw signal is preprocessed first. This includes outlier removal, noise filtering, and environmental factor correction. It should be noted that the blade thermal response acquired in this invention is a composite thermal signal formed by aerodynamic viscous shear dissipation and multi-source heat transfer coupling on the wall surface. Multiple interference components, such as ambient temperature, solar radiation, turbulence disturbance, blade coating aging, and surface dust and dew, are simultaneously superimposed on the signal. Therefore, a multi-layer signal preprocessing and operating condition coupling compensation mechanism is implemented. At the signal acquisition end, a second-order low-noise filter is performed on the raw thermal response time sequence to suppress power frequency and high-frequency noise interference. Simultaneously, a component temperature drift calibration algorithm is used to eliminate data deviations caused by sensor transient jumps or communication anomalies. Furthermore, real-time ambient temperature, ambient humidity, wind speed and turbulence intensity, impeller speed, and pitch angle are used as input parameters to couple the compensation model. The baseline offset of the thermal field caused by changes in global convective heat transfer intensity is quantitatively solved, and a global translation correction is performed on the reconstructed two-dimensional spatial thermal field distribution to eliminate the overall thermal field drift caused by environmental and unit dynamic operating conditions, retaining only local thermal structure distortions induced by boundary layer flow disturbances as effective identification criteria. Specifically, outlier removal is used to eliminate data deviations caused by sensor transient jumps or communication anomalies; filtering is used to retain effective thermal response changes within the blade rotation cycle; and environmental correction, by combining ambient temperature, wind speed, and blade operating parameters, compensates for the overall thermal response drift caused by changes in external conditions, ensuring that subsequent analysis focuses primarily on local thermal structure changes caused by changes in blade surface conditions.

[0039] After signal preprocessing, this invention further utilizes blade rotation angle information to spatially map the thermal response data. Specifically, the blade rotation angle θ(t) is obtained through a hub angle encoder, pitch system feedback signal, or equivalent rotation angle measurement signal, and a correspondence between time sampling points and blade spatial positions is established. Since the same measurement point will periodically pass through the same spatial position during blade rotation, each rotation cycle is divided into several angle intervals. The thermal response data within the same angle interval are categorized and averaged to form a periodic thermal response sequence indexed by the blade rotation angle. H i ( i )= f ( S i ( t ), i (t )) in, H i ( i () represents the thermal response value of the i-th measuring point under different blade rotation angles. Further combining the positional relationships of multiple spanwise measuring points, the angle mapping results of different measuring points are combined to form a two-dimensional thermal field matrix: M ( i , x )=[ H 1( i ), H 2( i ),…, H n ( i )] in, i Indicates the direction of the blade's rotation angle. x This matrix represents the spanwise position of the blade. It is used to convert the original time series into a quasi-spatial thermal field of "angle-span" to obtain a data structure that can reflect the aerodynamic thermal distribution characteristics of the blade surface.

[0040] S2, extract the spatial structure features of the two-dimensional thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field.

[0041] In one embodiment of the present invention, after obtaining the reconstructed thermal field, the present invention further extracts the thermal gradient change characteristics. Since the initial stage of blade icing primarily affects the boundary layer flow state and heat transfer path, it mainly manifests as a change in the continuity of local heat distribution, rather than a significant change in overall temperature. Based on this, the thermal gradient distribution is obtained by calculating the difference in thermal response between adjacent spanwise or adjacent angular positions: G x = M ( i , x ) / x G θ = M ( i , x ) / i Among them, G x G represents the change in thermal gradient along the spanwise direction of the blade. θThis represents the change in thermal gradient along the direction of rotation. When the blade surface remains in a normal state, the thermal gradient distribution exhibits strong continuity; however, when icing occurs in a localized area, the ice layer alters the surface roughness and thermal resistance distribution, leading to localized gradient enhancement, abrupt changes, or a decrease in spatial continuity. Therefore, by comparing the thermal gradient distribution between the current cycle and the normal baseline state, icing-related anomalies can be identified.

[0042] It should be understood that, for the calculation of the periodic consistency index, this invention utilizes the periodic repetition characteristics during blade rotation for analysis. Under normal operating conditions, the thermal response at the same spatial location in different rotation cycles exhibits high repeatability; therefore, its stability can be evaluated by the correlation between the thermal fields of adjacent rotation cycles. Specifically, let the first... k The thermal field matrix corresponding to each rotation cycle is M. k ( i , x ), No. k The thermal field matrix corresponding to +1 period is M. k+1 ( i , x The periodic consistency index is obtained by calculating the similarity between the two: C= Cov ( M k , M k+1 ) / σ k σ k+1 in, Cov This represents the covariance between two periodic thermal fields. s This represents the standard deviation of the thermal field data. This index is used to describe the stability of the thermal response structure during continuous rotation cycles. When early icing occurs on the blade surface, the spatial distribution of the thermal response gradually deviates from its original pattern due to the disturbance of the boundary layer structure, causing the periodicity index to decrease.

[0043] It should be further explained that, regarding the extraction of local structural offsets, this invention is mainly used to describe the degree of positional deviation of the current thermal field relative to the normal state thermal fingerprint model. First, a baseline thermal field model is established using data collected under normal operating conditions. M 0( θ,x This model, obtained through statistical analysis of data from multiple stable operating cycles, characterizes the thermal response of the blades under icing-free conditions. During real-time operation, the thermal field of the current cycle is... M ( i , x Differences between the thermal field and the reference thermal field are calculated: D ( i , x = |M ( i , x ) M 0( i , x )∣ Furthermore, by analyzing the distribution, duration, and spatial continuity of anomalous regions in the differential thermal field, local structural offsets are obtained. This index does not represent a simple temperature difference, but rather describes the degree of deviation of the thermal structure from the normal thermal fingerprint. When a local region exhibits a persistently high offset, and this offset occurs simultaneously with changes in the thermal gradient and a decrease in periodic consistency, it is considered that the region contains an icing-induced thermal structural anomaly.

[0044] S3. The spatial structure features, temporal similarity features and local structural deviation information are comprehensively evaluated to generate an evaluation index characterizing the overall degree of thermal field anomaly. The cumulative change trend of the evaluation index is continuously monitored over multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly is regionally clustered in space, it is determined that the corresponding region is in an icing evolution state, and an icing warning signal is output.

[0045] In one embodiment of the present invention, to reduce the impact of differences in manufacturing processes, materials, and initial surface conditions of different blades on the identification results, the present invention further introduces an individualized thermal fingerprint correction mechanism. When establishing the baseline thermal field model, a uniform template is not directly used; instead, thermal response data under stable operating conditions is collected for each blade to form its own baseline characteristic parameters. When the surface condition of the blade changes slowly during long-term operation, the baseline model is progressively updated using data from the stable operating phase, enabling the model to adapt to the natural changes during long-term service of the blade, while avoiding mistaking sudden abnormal states for new normal states.

[0046] It should be noted that during the identification process, the system does not rely on a single threshold for judgment, but rather makes a comprehensive judgment based on the evolution trend of the thermal structure over time. When a local area simultaneously exhibits continuous instability of the thermal gradient structure, a decrease in periodic consistency, and an increase in the cumulative phase drift, the area is determined to have entered the early stage of icing evolution. This process is achieved through the cumulative analysis of multiple rotation cycles, thereby avoiding misjudgments caused by short-term disturbances or environmental fluctuations.

[0047] This invention does not use a single temperature threshold as the basis for icing determination. Instead, it establishes a multi-feature fusion determination mechanism based on the continuous evolution of the aerothermal field structure over time. Its core idea is to identify the continuous structural changes that occur on the blade surface during the transition from a normal aerothermal equilibrium state to an icing disturbance state, rather than judging a certain instantaneous abnormal signal.

[0048] Specifically, after completing the thermal field reconstruction, the system first calculates the deviation of the thermal field in the current operating cycle from the baseline state based on the individualized thermal fingerprint model established under normal operating conditions, and obtains the thermal gradient anomaly index, the period consistency change index, and the thermal response phase drift index. The thermal gradient anomaly index describes the degree of change in the spatial distribution of the current thermal field relative to the normal state, reflecting the local thermal structure reconstruction caused by boundary layer disturbances and changes in heat transfer paths after ice formation. The period consistency index evaluates the stability of the thermal response at the same blade position in continuous rotation cycles, reflecting whether the aerodynamic thermal field still maintains its normal periodicity. The phase drift index describes the offset of the current thermal response characteristic position relative to the normal thermal response position, characterizing the changes in energy dissipation and heat release processes caused by icing.

[0049] In the specific judgment process, the degree of change of the above three characteristic quantities relative to the baseline state is first calculated, and then normalized to ensure that different physical quantities can be compared on a uniform scale. Let the degree of thermal gradient anomaly be... A g The degree of decrease in periodic consistency is A c The thermal response phase shift is A p Then, a comprehensive evaluation index for icing evolution is constructed: I ice = w g A g + w c A c + w p A p in, w g , w c , w p These represent the weighting coefficients for different characteristic parameters, and the weights can be determined based on historical operating data or characteristic sensitivity under typical operating conditions. This index is not used to reflect the icing state at a specific moment, but rather to describe the overall deviation of the current blade thermal structure from its normal state.

[0050] All algorithm parameters are adaptively calibrated based on historical data of the unit's stable, icing-free operation. Among these, the degree of thermal gradient anomaly... The degree of decline in periodic consistency Thermal response phase shift degree Three types of feature weights , , The least-squares fitting method was used to adaptively adjust the parameters based on typical icing conditions. The thresholds for determining the connectivity of the differential thermal field and updating the baseline thermal field were both based on the residuals of the thermal field under normal operating conditions. The intervals are dynamically generated to avoid insufficient model compatibility issues caused by manually setting thresholds. During periods of high incidence of low-temperature icing, the baseline update weight is automatically reduced to prevent baseline drift caused by residual water stains after ice melt from being continuously absorbed by the baseline model.

[0051] To avoid instantaneous anomalies caused by wind speed changes, turbulence disturbances, and ambient temperature fluctuations, this invention further introduces a time continuity constraint mechanism. That is, when an abnormal increase in thermal gradient, a decrease in periodic consistency, or a phase drift occurs in a certain monitoring area, the system will not immediately trigger an icing alarm. Instead, it will perform trend analysis on the evaluation indicators over multiple consecutive rotation cycles. Specifically, this is achieved by calculating the changing trend of the comprehensive evaluation indicators over consecutive cycles:

[0052] in, k Indicates the current rotation period. n This indicates the number of historical cycles involved in the trend analysis. K This indicates the icing evolution trend. When the comprehensive evaluation index continues to rise and the trend of change remains positive, it indicates that the anomaly is not a random disturbance, but a continuous evolutionary process caused by changes in the blade surface state.

[0053] Based on the above trend analysis results, this invention establishes a hierarchical state determination logic. Overall, the determination logic of this invention sets dual temporal constraints. The first constraint is the time dimension: only when the positive growth trend of the comprehensive anomaly evaluation index is continuously maintained for at least [time period missing]... A valid anomaly is considered to occur only after a certain number of rotation cycles. The number of historical periods participating in the trend calculation ranges from 5 to 30. The second constraint is the spatial dimension: abnormally connected regions at the same location must expand synchronously and continuously, with their areas gradually increasing. Only when both constraints are satisfied simultaneously will the system determine that the region has entered the icing evolution state and output a warning signal with spatial positioning. Compared with traditional single temperature threshold judgment or single-frame image detection schemes, the dual temporal constraints can significantly reduce false alarms caused by short-term meteorological changes and unit operating condition fluctuations, and greatly improve the predictability and stability of early weak icing identification under complex conditions of low temperature, high humidity, and low visibility. Furthermore, the graded state determination logic includes the following details: First, when the thermal field characteristics deviate slightly from the normal baseline model, but the anomaly lasts for a short period and does not show a significant upward trend, it is judged as a normal fluctuation state. At this time, the system only records data and does not output alarm information. When a change in the thermal gradient structure occurs in a local area, and the periodic consistency decreases, but the comprehensive evaluation index is still at a low level, it is judged as the initial stage of icing risk. The system outputs a warning message and increases the data sampling frequency of the corresponding area. When the three characteristics of thermal gradient anomaly, decreased periodic consistency, and phase drift occur simultaneously, and the comprehensive evaluation index exceeds the set reference range for multiple consecutive rotation cycles, it is judged as entering the early icing state. The system outputs an icing alarm and marks the location of the abnormal area. When the abnormal area continues to expand and the degree of thermal field deviation further increases, it is judged as the icing development stage, providing a basis for operation control or de-icing strategies.

[0054] Furthermore, to improve spatial positioning capabilities, this invention not only assesses the overall icing state of the blades but also performs spatial clustering analysis on anomalous thermal field regions. By analyzing the distribution of continuous anomalous regions in the differential thermal field matrix, the spanwise and circumferential locations where icing may occur are determined, thereby expanding the assessment from "whether icing has occurred" to "icing location and degree of evolution."

[0055] Furthermore, the baseline model in this invention employs a dynamic update mechanism, but the update process only occurs under long-term stable operating conditions. When the system detects that the blade is in a normal state and its thermal structural characteristics are continuously stable, the baseline model is slowly corrected using new stable data to adapt to gradual changes such as surface aging and contamination during long-term blade operation. When the system detects an abnormal evolution trend, the baseline update is stopped to avoid incorrectly including the icing state in the normal model, thereby ensuring the reliability of the judgment.

[0056] It should be understood that the reference thermal field adopts a layered verification-based progressive update strategy. Before updating, the uniformity of the global thermal field offset and the distribution of local gradient distortion are verified simultaneously: the first step is to determine whether the global thermal field offset is uniformly distributed across the entire domain, and the second step is to determine whether there is local concentrated gradient distortion. Progressive weighted correction is only allowed when the deviation is uniform across the entire domain and there are no local concentrated anomalies; if permanent local thermal field offsets caused by coating peeling, blade impact, etc., are detected, the reference model is immediately locked and automatic updates are stopped, and a blade surface structure anomaly warning is output simultaneously. At the same time, a periodic manual calibration and reset interface is provided. During the annual operation and maintenance phase of the unit, the reference thermal field can be reinitialized based on ice-free stable operation data, ensuring the sensitivity of icing identification and the reliability of the reference model in the long term from three dimensions: update logic, scene differentiation, and periodic reset.

[0057] In summary, the signal processing flow of this invention forms a complete technical chain consisting of "raw thermal response acquisition—rotational synchronization mapping—thermal field reconstruction—thermal gradient calculation—periodic consistency analysis—local structural offset evaluation—individualized benchmark correction—icing state determination." Among these, the thermal gradient change reflects the change in boundary layer heat exchange space, the periodic consistency index reflects the stability of the aerodynamic thermal structure, and the local structural offset reflects the deviation of the current state from the normal thermal fingerprint of the blade. These three types of features together constitute the basis for early icing identification, thus avoiding the problem of the single temperature threshold method being insensitive to the initial icing state. Through the above determination logic, this invention achieves a shift from single-state judgment to continuous state evaluation based on the evolution trend of aerodynamic thermal structure. Compared with traditional icing detection methods based on fixed temperature thresholds or power reduction, this method utilizes the cumulative effect of thermal field structure changes over multiple rotational cycles to identify the icing process, effectively reducing false alarms caused by short-term environmental disturbances and improving the stability and reliability of early icing identification under complex meteorological conditions.

[0058] Example 2 It is worth mentioning that this invention further incorporates thermal structure data into the digital twin framework at the system level, specifically: By constructing the aerodynamic-thermal coupling mapping relationship of the blades, a dynamic comparison between the actual operating thermal structure and the reference thermal structure is realized, so that the icing evolution process can be presented synchronously in the model space, thereby providing a continuous state basis for subsequent operation and maintenance decisions.

[0059] like Figure 3 As shown, the system of the present invention mainly includes a blade aerodynamic thermal response acquisition unit set on the wind turbine blade, a rotation state synchronization and data aggregation unit set in the hub area, a data processing unit set inside the nacelle, and a digital twin analysis platform set in a remote server or ground control center. The functional units interact with each other through wired or wireless communication links to jointly complete the acquisition of aerodynamic thermal response signals, spatial positioning, thermal field reconstruction, and early icing state identification during the operation of the wind turbine blade.

[0060] The blade aerodynamic thermal response acquisition unit is installed on the surface of the wind turbine blade or inside the blade composite material. It is used to acquire the thermal response signal formed by the combined effects of aerodynamic friction, boundary layer shear, and surface convection heat transfer during blade operation. The blade aerodynamic thermal response acquisition unit includes multiple distributed thermal response detection nodes, arranged along the spanwise and chordwise directions of the blade. The spanwise detection nodes are used to acquire the variation of the thermal response along the blade length, while the chordwise detection nodes are used to acquire the differences in thermal response between different aerodynamic regions of the blade. Preferably, the detection nodes are located near the leading edge of the blade, in the suction surface region, and in locations where aerodynamic thermal response changes are more sensitive, to improve the ability to sense boundary layer disturbances and surface state changes.

[0061] The blade aerodynamic thermal response acquisition unit is connected to the data aggregation module in the hub area via a signal transmission line. It transmits temperature or heat flow response signals collected by multiple detection nodes to the rotating side data processing node. Since the wind turbine blades are rotating, the data collected at the blade tip needs to correspond to the blade's spatial position. Therefore, the system further includes a rotation state synchronization unit. This unit can acquire real-time blade rotation angle information using the impeller azimuth angle signal from the existing wind turbine control system, the hub angle encoder signal, or a separately installed angle detection device. This angle information is then synchronized with the thermal response data to ensure that each thermal response data point corresponds to a specific spatial position of the blade during operation.

[0062] It should be understood that the data aggregation module located in the hub area is used to realize data exchange between the rotating and stationary components. One end of the module connects to multiple thermal response detection nodes at the blade end, and the other end connects to the data processing module on the nacelle side. The data aggregation module is used to perform preliminary processing on the data collected by multiple sensor nodes, including data synchronization, time stamping, and communication conversion. In the rotating state, data transmission between the blade end and the stationary end of the nacelle can be achieved through a slip ring communication structure located at the hub position, or by using a rotating-side wireless communication module to interact with the nacelle-side receiving module, to ensure that thermal response data can be continuously transmitted to the back-end processing system.

[0063] It should be noted that the nacelle-side data processing module receives thermal response data, rotation angle data, and wind turbine operating parameter data during blade operation, and performs thermal field data reconstruction and preliminary state analysis. Based on the blade rotation angle information, this module spatially maps the thermal response signals collected at different times, converting the original time series into an angle-span thermal response data structure corresponding to the blade's spatial position. It further calculates characteristic parameters such as thermal gradient changes, periodic consistency changes, and local thermal structure shifts, providing a data foundation for subsequent icing state identification.

[0064] It is worth mentioning that the digital twin analysis platform is set up on a remote server or ground control center to establish an individualized thermal fingerprint model of the wind turbine blade under normal operating conditions, and to dynamically compare the real-time acquired thermal field structure with the baseline thermal structure. The digital twin analysis platform integrates blade structural parameters, historical operating data, and real-time aerodynamic-thermal response data to construct an aerodynamic-thermal coupling mapping relationship for the blade, achieving synchronous updates between the actual operating state and the model space state. When a persistent thermal gradient anomaly, decreased periodic consistency, or cumulative changes in thermal structure offset are detected in a localized area of ​​the blade, the system outputs an icing risk level based on the comprehensive evaluation results and feeds the identification results back to the wind turbine operation control system or maintenance platform.

[0065] pass Figure 3 The system structure shown in this invention realizes a complete technical chain from acquiring the aerodynamic thermal response of the blade surface, rotational spatial positioning, data transmission, thermal field analysis to icing state identification. This system utilizes the naturally generated aerodynamic frictional heat during blade operation as a source of state perception information, without relying on external infrared radiation imaging conditions. Compared to traditional methods based on surface temperature images or environmental parameters, it can more directly reflect changes in the blade boundary layer state, providing a physical basis for early icing identification.

[0066] Furthermore, the digital twin framework of this invention is not an independent data processing system, but rather a data-driven model built upon blade operating status perception, rotational synchronization measurement, and thermal field reconstruction. Therefore, the hardware system is primarily used to acquire aerodynamic thermal response information and corresponding spatial location information during blade operation, providing physical data sources for subsequent thermal structure mapping and icing evolution analysis.

[0067] Specifically, the hardware system of this invention mainly includes a blade thermal response acquisition unit, a rotation state synchronization unit, a nacelle-side data processing unit, and a ground-based digital twin analysis platform. The blade thermal response acquisition unit acquires thermal response information generated on the blade surface by the combined effects of aerodynamic frictional heat and boundary layer heat transfer; the rotation state synchronization unit acquires the real-time rotation angle and operating state parameters of the blade; the nacelle-side data processing unit performs data reception, synchronization, preprocessing, and thermal field reconstruction; and the ground-based digital twin platform establishes an aerodynamic-thermal coupling mapping model of the blade and enables dynamic comparison between the operating state and the baseline state.

[0068] It should be noted that, in terms of blade thermal response acquisition, this invention can employ distributed temperature sensors or flexible thermal sensor arrays as aerodynamic thermal response detection units. Depending on actual engineering conditions, thin-film thermocouples, flexible resistive temperature sensors, fiber Bragg grating temperature sensors, or other sensing elements with surface temperature change detection capabilities can be selected. Preferably, a flexible thin-film temperature sensing structure is used, allowing the sensing unit to conform to the curvature changes of the blade composite material surface and reducing the impact on the original aerodynamic shape of the blade. If fiber optic sensing is used, a continuous thermal response measurement path can be formed by laying optical fibers along the blade, improving spatial sampling capabilities.

[0069] It should be further clarified that the sensor placement is determined based on the aerodynamic characteristics of the blade, rather than simply being uniformly laid out. Since the leading edge region of the blade experiences strong airflow impact and is highly sensitive to boundary layer formation and surface state changes, thermal response measurement points are preferentially placed near the leading edge. Simultaneously, to obtain the development pattern of icing along the blade's length, multiple measurement sections are set in the spanwise direction, ensuring that sensor data covers the blade root, middle, and tip regions. In practice, several spanwise measurement areas can be set according to the blade's length and structural characteristics, with multiple measurement points set along the chordal direction within each area. For example, measurement positions can be placed in the high-shear region at the leading edge, near the maximum thickness, and in the trailing edge region to obtain the differences in thermal response across different aerodynamic regions of the blade.

[0070] At the hardware level, the sensing unit employs a high-precision flexible thin-film sensor array with a temperature range of ±0.01℃, coupled with a composite encapsulation protection structure that is resistant to freeze-thaw cycles and centrifugal delamination, ensuring long-term reliable operation under high-speed blade rotation and low-temperature, high-humidity environments. The sensing unit features densely distributed measurement points along the blade's leading edge and suction surface—areas with high icing risk—and differentiated sensor spacing at the blade root, middle, and tip to prioritize the accuracy of thermal field sampling in high-icing areas.

[0071] Furthermore, an aerodynamic prior constraint interpolation algorithm is introduced: relying on the aerodynamic heat transfer CFD simulation prior database corresponding to the blade airfoil and spanwise position, refined interpolation calculations with boundary layer constraints are performed between adjacent sensing points to restore the continuous thermal gradient distribution of aerodynamically sensitive regions such as the leading edge and suction surface. Simultaneously, a small connected domain enhancement operator is added at the algorithm level to enhance the features of discrete anomalous regions with small scales but significant gradient abrupt changes in the interpolated differential thermal field matrix. This effectively captures local thermal structure distortions corresponding to small-scale nascent thin ice and avoids the problem of missed early icing identification due to insufficient measuring point density.

[0072] The chordal sensing layout is primarily used to identify thermal structural changes caused by alterations in local surface roughness and boundary layer conditions due to ice layer variations, while the spanwise sensing layout is mainly used to analyze the expansion process of the icing region along the blade length. Joint sampling in both the chordal and spanwise directions allows for the formation of a two-dimensional thermal response distribution on the blade surface, providing a spatial basis for subsequent thermal field reconstruction.

[0073] Furthermore, to address the power supply and signal transmission issues during the rotating blade operation, this invention employs a combination of localized power supply at the blade tip and communication between the rotating components. Specifically, the blade tip sensing unit can be configured with a miniature low-power acquisition module, powered by a power supply module installed near the hub, or powered by existing electrical channels within the blade. For long-distance blade arrangements, low-power signal acquisition nodes can be set up inside the blade to locally converge signals from multiple sensors, and then transmit them to the nacelle-side processing unit via a data interface in the hub area.

[0074] It should be understood that, for data transmission between rotating and stationary components, a slip ring communication structure can be used to achieve data exchange between the rotating blades and the nacelle control system. The slip ring is installed at the connection point between the hub and the nacelle, allowing data collected at the blade tip to be continuously transmitted to the stationary equipment during impeller rotation. For implementations using wireless communication, a wireless data aggregation node can also be set up in the hub area to transmit sensor data to the nacelle-side receiving module via low-power wireless communication. Both communication methods do not alter the main blade structure and can meet the data acquisition requirements under long-term operating conditions.

[0075] It is worth mentioning that, to achieve an accurate correspondence between thermal response data and blade spatial position, this invention further includes a rotation angle acquisition unit. Rotation angle information can be obtained through encoder signals from the existing wind turbine control system, hub angle sensors, pitch system position feedback signals, or a separately configured inertial measurement unit. Preferably, the impeller azimuth angle signal provided by the existing wind turbine control system is used, as this signal directly reflects the blade's spatial position during rotation and requires no additional complex mechanical structures. When the control system angle signal cannot be directly obtained, the blade rotation period and real-time angular position can be obtained through an angle encoder installed in the hub area.

[0076] During data processing, the rotation angle information is synchronized with the thermal response data collected by each sensor. Based on the blade rotation cycle, the system maps continuously collected temperature or heat flux data to the corresponding blade angle position, thus establishing a three-dimensional thermal response data structure of "rotation angle—blade spanwise position—time". This method eliminates the influence of spatial position changes caused by blade rotation, making the thermal response data of the same blade region comparable under different rotation cycles.

[0077] During the construction of the digital twin model, the ground analysis platform establishes an aerodynamic-thermal coupling mapping model of the blades based on blade structural parameters, operational status parameters, and historical stable operating thermal response data. This model is not intended to replace actual flow field calculations, but rather to describe the changing patterns of the blade's thermal response structure under different operating conditions. During real-time operation, the collected thermal field data is input into the model and dynamically matched with the baseline thermal structure under the corresponding operating conditions. By analyzing changes in thermal gradients, periodic consistency, and the degree of local structural offset, the synchronous mapping of the icing evolution state in the digital space is achieved.

[0078] Through the aforementioned hardware deployment, this invention establishes a complete technical chain from blade surface thermal response acquisition, rotational spatial positioning, data transmission, thermal field reconstruction to digital twin analysis. This enables the continuous sensing and analysis of aerodynamic frictional heat, a naturally occurring physical response during blade operation, providing a reliable data foundation for early icing identification based on thermal structural anomalies. This approach differs from traditional external observation methods using infrared imaging equipment; its detection information originates from the blade's own operational status, making it more suitable for long-term online monitoring scenarios.

[0079] Example 3 It is worth mentioning that this technical solution is significantly different from existing technologies and actually solves some technical problems, specifically: The core innovation of this solution lies in utilizing the naturally generated aerodynamic frictional heat and boundary layer thermal response characteristics during wind turbine blade operation to achieve early icing identification, distinct from traditional infrared temperature detection. Addressing the insufficient description of the hardware system architecture in the current technical disclosure, the signal acquisition and system components will be further supplemented in the subsequent specification. However, it should be noted that this invention is not limited to a specific fixed sensor type, but rather is protected based on the function of "blade aerodynamic thermal response information acquisition." Depending on the actual engineering requirements, distributed temperature sensors, thin-film thermocouples, flexible thermistor arrays, or sensing units with equivalent thermal response detection capabilities can be used for signal acquisition. These sensing units can be arranged in the leading edge region, high-shear region of the suction surface, or typical aerodynamically sensitive regions of the blade, and can be arranged at multiple points along the blade spanwise to obtain the variation law of aerodynamic frictional heat distribution along the blade space. The sensing units can be connected to the blade using flexible attachment, surface embedding, or integrated composite material encapsulation methods, achieving thermal response information acquisition during operation without altering the main aerodynamic shape of the blade. This solution does not impose a fixed limit on the specific number and spacing of sensors, but configures them according to the blade size, monitoring accuracy requirements, and installation conditions to ensure the scope of patent protection and engineering applicability.

[0080] To address the issue of insufficient description between raw data and thermal field structural features in the signal processing process, future work will further refine the data processing workflow, ensuring it embodies a complete technical chain of "physical signal acquisition—spatial mapping—feature construction—anomaly identification." Specifically, sensors acquire time-varying temperature or thermal response sequences at different blade locations. First, combining impeller speed signals, blade azimuth angle signals, and sampling time information, data from different measurement points are synchronously calibrated, converting the time-domain signal into quasi-spatial distribution data corresponding to the blade's rotational position. Subsequently, a two-dimensional or three-dimensional thermal response matrix is ​​established based on the blade's spanwise position and rotation period, achieving thermal field reconstruction. Based on this thermal field reconstruction, by calculating the spatial response differences between different rotation periods, structural features reflecting boundary layer state changes are extracted, including local thermal gradient changes, thermal field homogeneity changes, periodic consistency changes, and thermal response phase shifts. Specifically, the thermal gradient describes the changing trend of aerodynamic frictional heat spatial distribution, thermal field homogeneity characterizes the degree of disruption to the thermal structure continuity under normal operating conditions, periodic consistency distinguishes between continuous state changes and random disturbances, and thermal response phase shifts reflect changes in heat transfer paths caused by icing. The above processing is not a simple data statistics method, but a feature extraction method based on the changes in energy dissipation paths caused by the influence of ice on the aerodynamic boundary layer of the blade.

[0081] To address the lack of specific judgment thresholds and trend models, this solution employs a dynamic discrimination method based on the degree of deviation from a baseline state, rather than a fixed temperature threshold. Since different wind turbine models, blade materials, operating environments, and meteorological conditions vary, using a uniform absolute threshold can easily lead to insufficient adaptability. Therefore, this invention is more suitable for anomaly identification using individualized baseline models. Specifically, under conditions of no blade icing and stable operation, aerodynamic thermal response data are collected under different wind speeds, rotational speeds, and ambient temperatures to establish a thermal field characteristic database under normal operating conditions and form a corresponding thermal response baseline model for each blade. During real-time operation, the thermal field characteristics collected in the current cycle are compared with the baseline model to calculate the degree of deviation of the current state from the normal state. When a local area exhibits a combination of changes in multiple dimensions, such as continuous changes in thermal gradient, decreased thermal field uniformity, reduced periodic repetition consistency, and accumulated phase shift, it is determined that the area has an icing evolution trend.

[0082] It should be noted that the "non-single threshold judgment" in this invention does not mean the absence of criteria, but rather that a comprehensive anomaly evaluation mechanism is composed of multiple physical features. In specific implementation, a feature fusion model can be established based on actual operating data. For example, a weighted scoring model, statistical distance model, or machine learning classification model can be used to fuse multiple feature quantities to form an icing risk assessment index. When this assessment index exceeds the dynamic reference range determined based on historical normal operating data, the icing risk level is output. This judgment method can avoid misjudgments caused by single temperature changes, instantaneous wind speed disturbances, or environmental changes, and it aligns with the technological development direction of trend analysis and condition assessment in wind power equipment condition monitoring.

[0083] It should be further noted that this technical solution includes some compensation measures compared to existing technologies (but these are not the core innovations of this invention): An integrated compensation and avoidance scheme for weak thermal signal-to-noise ratio and coupling interference from varying environmental operating conditions. Addressing the baseline drift interference caused by weak aerodynamic frictional heat amplitude, environmental turbulence, temperature and humidity variations, and pitch control conditions, this invention incorporates a multi-layered signal preprocessing and operating condition coupling compensation mechanism to prevent misjudgments caused by insufficient signal-to-noise ratio from the outset. First, a second-order low-noise filtering and component temperature drift calibration algorithm are integrated at the sensor acquisition end to remove jump anomalies and suppress power frequency noise in the original thermal response time series, thereby improving the effective signal-to-noise ratio of weak frictional heat signals. Second, a multi-dimensional operating condition coupled compensation model is constructed, using real-time ambient temperature, wind speed and turbulence intensity, unit pitch angle, and impeller speed as compensation inputs to quantify the offset of convective heat transfer intensity from the reference thermal field. Global baseline translation correction is performed on the real-time reconstructed two-dimensional thermal field to eliminate the overall thermal field drift caused by environmental and unit dynamic operating conditions. At the same time, the characteristics of short-term / gradually changing attachments such as dew and ash and icing evolution are distinguished: the thermal gradient distortion caused by dew and floating ash has no continuous expansion trend and cannot form a multi-period stable connected anomaly zone. This invention uses the dual judgment constraint of "multi-period cumulative growth trend + continuous expansion of spatial connected region" to filter out the instantaneous local anomalies caused by non-icing attachments, greatly reducing the icing classification judgment error under extreme variable operating conditions and realizing stable extraction of thermal field characteristics under complex meteorological conditions.

[0084] Super-resolution reconstruction optimization of the thermal field under sparse sensor array layout avoids the omission defect caused by insufficient measurement point density. To address the spatial resolution shortcomings caused by sparse sensor point arrangement and the masking of small-scale nascent thin ice thermal distortion by interpolation smoothing, this invention introduces an aerodynamic prior constraint interpolation algorithm in the two-dimensional thermal field reconstruction stage to compensate for the loss of recognition accuracy caused by sensor array density limitations. Based on the prior database of aerodynamic heat transfer CFD simulation corresponding to the blade airfoil and spanwise position, fine interpolation calculations with boundary layer constraints are performed between adjacent sensing points to restore the continuous thermal gradient distribution of aerodynamically sensitive areas such as the leading edge and suction surface. This improves the restoration capability of local small anomaly areas without unlimited densification of sensing points. At the same time, a sensing point densification logic is set for high-icing-risk areas such as the blade leading edge and suction surface. The sensing unit density is configured differently in segments at the blade root, blade middle, and blade tip to prioritize the thermal field sampling accuracy of high-icing areas. At the algorithm level, a small connected domain enhancement operator is added to enhance the features of discrete anomaly areas with small scale but significant gradient abrupt changes in the interpolated differential thermal field matrix. This can effectively capture the local thermal structure distortion corresponding to small-scale nascent thin ice, avoid the problem of early icing missed identification caused by excessive spacing between sensing points, and balance hardware deployment cost and icing spatial positioning accuracy.

[0085] A hierarchical verification mechanism for dynamic updating of the reference thermal field avoids baseline drift and structural distortion problems caused by long-term blade service. Addressing the distortion defects in the reference model caused by dust accumulation, coating aging, local structural damage, and residual melt-ice water stains during long-term blade service, this invention designs a hierarchical verification-based adaptive reference update strategy, which distinguishes three types of thermal field shift scenarios: normal gradual change, structural abrupt change, and icing disturbance. A pre-verification branch is added before the progressive weighted correction of the reference thermal field: The first step distinguishes between slow and uniform gradual changes such as blade coating aging and slight dust accumulation and local structural abrupt changes such as coating peeling and surface impact, using dual criteria of global thermal field offset and spatial gradient distortion distribution. Progressive weighted updates are only allowed when the overall thermal field deviation is uniform and there is no local concentrated distortion. If a local permanent thermal field offset area is detected, the reference model is locked and automatic updates are stopped, and a blade surface structure anomaly warning is output simultaneously to avoid structural damage interfering with the icing judgment reference. For the slow baseline offset caused by residual water stains in the icing-melting cycle, a seasonal adaptive update threshold is introduced. During the high incidence of low temperature and high humidity icing, the reference update weight is automatically reduced to prevent the water stain deviation from being continuously absorbed by the reference model. At the same time, a periodic manual calibration and reset interface is provided. During the annual operation and maintenance phase of the unit, the reference thermal field can be reinitialized based on ice-free stable operation data. From the three dimensions of update logic, scenario differentiation, and periodic reset, the sensitivity of icing recognition and the reliability of the reference model are guaranteed in the long term.

[0086] The technical effects of this solution, which differ from existing technologies, are explained below: Compared to traditional infrared thermal imaging methods, the core difference of this invention lies in the different nature of the observed object: traditional methods are based on passive measurement of external radiation temperature fields, and the results are easily affected by environmental radiation, lighting conditions and surface obstruction; while this invention is based on the frictional heat and boundary layer dissipation process naturally formed during the aerodynamic operation of the blades, and observes the endogenous thermal structure field, which essentially reflects the changes in flow structure and energy path. Therefore, it has higher pre-sensitivity and stronger physical interpretability for early weak icing states.

[0087] In summary, this invention transforms the problem of wind turbine blade icing from "temperature anomaly detection" to "identification of boundary layer aerodynamic and thermal structural stability failure," enabling continuous perception and early identification of the early evolution of icing without altering the blade structure. It also provides a novel technical approach for wind turbine blade condition monitoring based on endogenous physical fields.

[0088] The technical effects brought about by this technical solution, which differ from existing technologies, are as follows: Compared with existing wind turbine blade icing monitoring technologies, the beneficial effects of this invention are reflected in the overall improvement of its ability to identify weak states in the early stages of icing, its ability to resist environmental interference, and its ability to reflect the true operating state of the blade. This improvement comes from the utilization of the endogenous physical process of aerodynamic boundary layer frictional heat and the introduction of thermal field structure evolution characteristics, rather than relying on external radiation imaging or indirect inference from environmental parameters.

[0089] First, this invention utilizes the aerodynamic frictional heat and boundary layer energy dissipation processes naturally generated during blade operation as an information source, transforming the monitored object from a traditional external temperature field or image field to an endogenous thermal structure field. This mechanism avoids the dependence of infrared detection methods on environmental radiation, rain and snow obstruction, and changes in viewing angle. Under complex meteorological conditions, this method is insensitive to changes in external light and exhibits more stable adaptability to low visibility conditions (such as fog, snow, and low light at night), thereby significantly reducing the probability of detection failure caused by imaging conditions.

[0090] Secondly, this invention does not use absolute temperature or a single threshold as the criterion, but rather comprehensively judges based on multi-dimensional thermal field structural characteristics such as thermal gradient structure, periodic consistency, and phase shift, transforming the identification target from "temperature anomalies" to "thermal structural stability degradation processes." This mechanism can capture subtle perturbations in the boundary layer flow state before ice forms observable thickness changes, theoretically improving the pre-emptiveness of ice formation identification. According to mechanistic analysis, this method can exhibit a decrease in thermal field uniformity and deviation from periodic repetition consistency in the initial ice adhesion stage, demonstrating earlier response capability compared to traditional methods based on power loss or temperature anomalies.

[0091] Furthermore, this invention establishes a complete icing thermal response evolution chain from a physical causal logic perspective: "increased roughness → forward shift of boundary layer transition → abrupt change in turbulent heat transfer + thermal conduction phase drift → disruption of spanwise / circumferential thermal gradient continuity." This provides clear physical interpretability for the judgment criteria, distinguishing it from traditional data-driven black-box models. By synchronously mapping the blade rotation angle to the thermal response signal, this invention reconstructs the quasi-spatial thermal field from a time series, transforming the previously unobservable boundary layer thermal evolution process into an analyzable structured thermal field distribution, thereby improving the spatial resolution of icing identification. This approach allows the thermal structural disturbances caused by localized icing to be amplified and accumulated in the spatial dimension, avoiding the problem of single-point measurements being insensitive to local anomalies.

[0092] This invention effectively suppresses interference components such as ambient temperature, solar radiation, and turbulence disturbances through a multi-layer signal preprocessing and operating condition coupling compensation mechanism. It distinguishes between icing disturbances and non-icing gradual changes (coating aging, dust accumulation) and local structural abrupt changes (coating peeling, impact) through a layered verification-based progressive update strategy. A dual temporal constraint judgment logic (a trend lasting at least n cycles in the time dimension + synchronous and continuous expansion of the connected region in the spatial dimension) filters out instantaneous local anomalies caused by short-term non-icing deposits (dew, floating dust), forming an anti-interference closed loop covering the entire chain of "signal acquisition → interference suppression → feature extraction → anomaly judgment." Theoretical analysis shows that under conditions of fluctuating wind speed or small changes in ambient temperature, traditional temperature threshold-based methods are prone to misjudgment. This invention, however, uses a thermal structure consistency index for judgment, based on multi-cycle cumulative change trends, thus exhibiting a significant suppression effect on transient disturbances.

[0093] In terms of recognition performance, under typical low-temperature and high-humidity conditions, combined with a ±0.01℃ high-precision sensor array, aerodynamic CFD prior constraint interpolation, and a small connected domain enhancement operator, this invention can provide early warnings through thermal gradient structural instability and phase drift characteristics before the ice layer has undergone significant geometric thickness changes (corresponding to the stage where the wall roughness has just changed and the boundary layer transition position has shifted forward). Based on the mechanism derivation, the sensitivity of this method to the initial state of icing stems from changes in the boundary layer energy dissipation path; therefore, its detection target is closer to the physical starting point of ice formation than the resulting state.

[0094] Furthermore, by introducing an individualized thermal fingerprint benchmark model, this invention eliminates the inherent thermal response deviations caused by manufacturing differences and surface roughness differences among different blades, thereby improving cross-blade identification consistency and avoiding the problem of insufficient adaptability of a unified model between different units, enabling the system to maintain stable identification performance when used in multiple units.

[0095] In summary, this invention, without altering the blade structure, achieves proactive identification of early icing conditions on wind turbine blades by introducing a thermal structure analysis mechanism driven by aerodynamic frictional heat. Compared to existing technologies that rely on environmental parameters or infrared imaging, this invention demonstrates significant improvements in anti-interference capabilities, early sensitivity, and structural consistency identification.

[0096] Example 4 To achieve the above invention, such as Figure 4 As shown, this embodiment also provides a wind turbine blade icing identification device 10, which includes: The thermal field reconstruction module 100 is used to acquire the original thermal response timing signal and blade rotation angle signal collected by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, the timing signal of each sensing unit is mapped to the corresponding blade rotation angle range, and combined with the spanwise position information of each sensing unit, the two-dimensional spatial thermal field distribution of the current rotation cycle is reconstructed.

[0097] The feature extraction and comparison module 200 is used to extract the spatial structure features of the two-dimensional thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field.

[0098] The fusion judgment module 300 is used to comprehensively evaluate the spatial structure features, the temporal similarity features, and the local structure deviation information, generate evaluation indicators that characterize the overall degree of thermal field anomaly, and continuously monitor the cumulative change trend of the evaluation indicators within multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly shows a regional clustering in space, it is determined that there is an icing evolution state in the corresponding region, and an icing warning signal is output.

[0099] The device described in this invention effectively solves the problems of insufficient sensitivity to early, weak icing conditions and difficulty in spatial positioning in existing technologies. It achieves an integrated solution for the entire process, from thermal field reconstruction, multi-dimensional feature extraction, anomaly quantification assessment to icing region determination, significantly improving the pre-emptiveness and spatial positioning accuracy of icing identification, and enhancing the adaptability and engineering practical value of the method under complex operating conditions and environmental interference.

[0100] Example 5 To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 5As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads the 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 wind turbine blade icing identification method described above.

[0101] Example 6 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 a wind turbine blade icing identification method as described in the foregoing embodiments.

[0102] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are 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.

[0103] 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 method for identifying icing on wind turbine blades, characterized in that, include: The original thermal response timing signal and blade rotation angle signal are acquired by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, the timing signal of each sensing unit is mapped to the corresponding blade rotation angle range. Combined with the spanwise position information of each sensing unit, the two-dimensional spatial thermal field distribution of the current rotation cycle is reconstructed. Extract the spatial structure features of the two-dimensional thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field. The spatial structural features, temporal similarity features, and local structural deviation information are comprehensively evaluated to generate an evaluation index characterizing the overall degree of thermal field anomaly. The cumulative change trend of the evaluation index is continuously monitored over multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly is regionally clustered in space, it is determined that the corresponding region is in an icing evolution state, and an icing warning signal is output.

2. The method as described in claim 1, characterized in that, The process of establishing the baseline thermal field distribution includes: Historical thermal response data of the blades were acquired over multiple rotation cycles during which the blades were ice-free and operating stably. This data was then combined with the rotation angle signal and spanwise position information corresponding to each cycle, and analyzed using the formula... Reconstructing the two-dimensional spatial thermal field distribution across multiple historical cycles; The two-dimensional spatial thermal field distributions of each historical period are statistically averaged according to the same angular interval and spanwise position to generate the initial reference thermal field distribution; Acquire new thermal field data of the blade during the long-term stable operation phase, calculate the difference between the new thermal field data and the initial reference thermal field distribution, and when the difference value is less than the preset update threshold, perform progressive weighted correction on the initial reference thermal field distribution using the new thermal field data to generate an updated reference thermal field distribution. Among them, the preset update threshold is based on the thermal field residual under normal operating conditions. Interval dynamic generation; the reference thermal field adopts a layered verification-based progressive update strategy: before updating, the uniformity of global thermal field offset and the distribution of local gradient distortion are verified simultaneously. Weighted correction is only performed when the deviation is uniform across the entire domain and there are no local concentrated anomalies. If permanent local thermal field offset caused by coating peeling or blade collision is detected, the reference model is immediately locked and the update is stopped, and a blade surface structure anomaly prompt is output simultaneously.

3. The method as described in claim 1, characterized in that, The process of extracting spatial structural features includes: The two-dimensional spatial thermal field distribution of the current rotation period is obtained as input data, according to the formula. Calculate the rate of change of thermal response between adjacent positions along the spanwise direction of the blade to obtain the spanwise thermal gradient distribution, and then apply the formula... Calculate the rate of change of thermal response between adjacent angular intervals along the blade rotation angle to obtain the circumferential thermal gradient distribution; The spanwise thermal gradient distribution and the circumferential thermal gradient distribution are respectively compared with the reference thermal gradient distribution corresponding to the reference thermal field distribution to obtain the spanwise gradient difference value and the circumferential gradient difference value. Spatially accumulate the spanwise gradient difference value and the circumferential gradient difference value to output a thermal gradient anomaly index that characterizes the degree of disruption of the spatial continuity of the current thermal field.

4. The method as described in claim 1, characterized in that, The process of extracting temporal similarity features includes: Obtain the two-dimensional spatial thermal field distribution during the current rotation period. Two-dimensional spatial thermal field distribution of the adjacent previous period As input data, the thermal field distribution of the two cycles is mapped point by point according to the same angular interval and spanwise position; The periodic consistency index between two periodic thermal field distributions is calculated using the following formula: in, This represents the covariance between two periodic thermal fields. and These represent the standard deviations of the thermal field data for the two periods, respectively. The period consistency index of multiple adjacent period pairs is continuously acquired, arranged in chronological order to form a period consistency sequence, and the period consistency sequence is output as a temporal similarity feature. Before input feature extraction, the thermal response time series signal undergoes multi-layer signal preprocessing and operating condition coupling compensation: second-order low-noise filtering and component temperature drift calibration are performed on the original time series to eliminate sensor jumps and power frequency noise; real-time ambient temperature, ambient humidity, wind speed and turbulence intensity, impeller speed and pitch angle are used as compensation parameters to quantify the baseline offset of global convective heat transfer on the reconstructed thermal field, and global translation correction is performed on the two-dimensional spatial thermal field distribution.

5. The method as described in claim 1, characterized in that, The process of acquiring information about local structural deviations includes: Obtain the two-dimensional spatial thermal field distribution during the current rotation period. and the pre-established reference thermal field distribution As input data, the two thermal field distributions are matched point by point according to the same angular interval and spanwise position; According to the formula Calculate the absolute difference in thermal response between corresponding points to generate the differential thermal field matrix; Spatial connectivity analysis is performed on the differential thermal field matrix to extract connected regions formed by consecutive differential values ​​exceeding a preset threshold. The spatial distribution range and cumulative differential amount of each connected region are used as local structural deviation information output.

6. The method as described in claim 1, characterized in that, The process of generating evaluation indicators includes: Obtain the degree of thermal gradient anomaly in the current period The degree of decline in periodic consistency and the degree of thermal response phase shift As input data, the three are normalized to ensure that each indicator is on a uniform measurement scale. Calculate the comprehensive anomaly evaluation index using the following formula: in, , , These represent the weighting coefficients corresponding to the degree of thermal gradient anomaly, the degree of periodic consistency degradation, and the degree of thermal response phase shift, respectively. Output a comprehensive anomaly evaluation index for the current rotation cycle. ; Among them, all algorithm parameters on which the comprehensive anomaly evaluation index depends are adaptively calibrated based on historical data of stable, ice-free operation of the unit; the weights of three types of features are: thermal gradient anomaly, decreased periodic consistency, and thermal response phase shift. , , The baseline update weights are determined by least squares fitting based on typical icing conditions; during periods of high incidence of low-temperature icing, the baseline update weights are automatically reduced to suppress the interference of waterlogging baseline drift on the evaluation indicators.

7. The method according to claim 1, characterized in that, The process for determining the output conditions of an icing warning signal includes: Obtain the comprehensive anomaly evaluation index corresponding to each of multiple consecutive rotation cycles. And arranged in cyclical order to form a time series of evaluation indicators; According to the formula Calculate the current period relative to the previous period Overall trend value of each cycle ,like If the value is positive and maintains positive growth for multiple consecutive periods, then the cumulative change trend of the evaluation indicator is determined to be a continuous growth trend; among which, the number of consecutive rotation periods involved in the trend calculation... The value range is 5 to 30; The spatial distribution of connected regions in the differential thermal field matrix where the difference value continuously exceeds a preset threshold is obtained. When the connected regions exist continuously for multiple periods at the same spatial location and the area shows an expanding trend, it is determined that the anomaly is regionally clustered in space, and an icing warning signal containing the spanwise position and circumferential angle range of the anomaly region is output. The judgment logic incorporates dual time constraints: only when the positive growth trend of the comprehensive anomaly evaluation index is continuously maintained for at least [time period missing]... Only when the abnormally connected regions at the same location have undergone a rotation cycle and simultaneously and continuously expanded, with their areas gradually increasing, is it determined that the corresponding region has entered the icing evolution state.

8. A wind turbine blade icing identification device, characterized in that, include: The thermal field reconstruction module is used to acquire the original thermal response timing signal and blade rotation angle signal collected by the distributed sensing units arranged on the blade surface. Based on the rotation angle signal, the timing signal of each sensing unit is mapped to the corresponding blade rotation angle range, and combined with the spanwise position information of each sensing unit, the two-dimensional spatial thermal field distribution of the current rotation cycle is reconstructed. The feature extraction and comparison module is used to extract the spatial structure features of the two-dimensional spatial thermal field distribution and the temporal similarity features between adjacent rotation cycles, and compare the current thermal field distribution with the pre-established reference thermal field distribution under normal blade operation to obtain the local structural deviation information of the current thermal field relative to the reference thermal field. The fusion judgment module is used to comprehensively evaluate the spatial structure features, the temporal similarity features, and the local structure deviation information, generate evaluation indicators that characterize the overall degree of thermal field anomaly, and continuously monitor the cumulative change trend of the evaluation indicators within multiple rotation cycles. When the cumulative change trend shows a continuous growth trend and the anomaly shows a regional clustering in space, it is determined that there is an icing evolution state in the corresponding region, and an icing warning signal is output.

9. An electronic device, characterized in that, include: processor; The memory stores executable instructions; when the processor executes the instructions, it implements the wind turbine blade icing identification method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The device contains a computer program, which, when executed by a processor, implements a wind turbine blade icing identification method as described in any one of claims 1-7.

Images