Dynamic early warning system for icing of blades of wind turbine generator

Abstract

According to the wind turbine generator blade icing dynamic early warning system provided by the invention, through omnibearing monitoring of the frequency modulation continuous wave radar group, the orthogonal weaving piezoelectric fiber array and the infrared temperature measurement unit, airspace water drop distribution, blade vibration characteristics and multi-dimensional dynamic changes of a surface temperature field are accurately captured; by combining improved Kalman filtering, wavelet transformation and a dynamic weight distribution algorithm of an intelligent decision module, traditional single threshold judgment is upgraded to comprehensive threat index deduction based on digital twinborn bodies, and the temporal-spatial resolution and reliability of icing threat assessment are remarkably improved; the execution control module achieves heating network partition accurate regulation and control and unit operation state collaborative optimization according to the generated anti-icing strategy, energy waste of traditional global heating is avoided, material damage caused by local overheating or underheating is effectively restrained, finally, anti-icing energy consumption is reduced under the complex meteorological condition, and the anti-icing efficiency is improved. And meanwhile, the power generation efficiency loss of the unit under the icing working condition is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a dynamic early warning system for icing on wind turbine blades. Background Technology

[0002] As an important component of clean and renewable energy, wind energy has seen its development and utilization scale continue to expand, with the single-unit capacity of wind turbines exceeding 15MW. Units operating in cold and humid climates generally face the problem of blade icing. Accumulated ice can lead to decreased aerodynamic performance, load imbalance, and loss of power generation efficiency, and in severe cases, it can cause structural resonance or even fracture accidents.

[0003] Traditional anti-icing technologies are mainly divided into two categories: passive coating and active heating. The former uses superhydrophobic materials to delay icing but cannot cope with continuous freezing rain conditions; the latter uses electrothermal films for full-area heating, resulting in excessive energy consumption and prominent local overheating and underheating phenomena. With the introduction of monitoring methods such as millimeter-wave radar and fiber optic sensing, existing systems mostly use single parameter threshold judgments and fail to effectively integrate multi-physics dynamic data, resulting in additional energy loss and reduced energy conversion efficiency. Summary of the Invention

[0004] In view of this, the present invention provides a dynamic early warning system for icing of wind turbine blades to solve the technical defects existing in the prior art.

[0005] Specifically, the present invention provides a dynamic early warning system for icing of wind turbine blades, comprising: an environmental situation perception module, an intelligent decision-making module, and an execution control module; The environmental situation awareness module acquires airspace water droplet distribution data, blade vibration characteristics, and surface temperature field, and transmits the data to the intelligent decision-making module. The intelligent decision-making module performs fusion calculations on the water droplet distribution data, vibration characteristics, and temperature field to generate a threat assessment result. Based on the assessment result, it outputs an anti-icing strategy containing heating power distribution and unit control commands to the execution control module. The execution control module controls the heating network partitions to operate according to the anti-icing strategy and coordinates the adjustment of the unit's operating status.

[0006] In some implementations, the environmental situation awareness module includes: The frequency-modulated continuous wave radar array scans the airspace to form a three-dimensional concentration distribution model that includes particle size differentiation, and performs moving target filtering on the radar echo signal to generate the water droplet concentration gradient change rate. An array of orthogonally woven piezoelectric fibers is used to collect vibration signals, which are then processed by a frequency feature extraction algorithm to generate harmonic distortion rate. The infrared temperature measurement unit is used to dynamically calibrate the surface temperature and calculate the rate of change of the temperature difference slope at each measurement point.

[0007] In some implementations, the intelligent decision-making module is further configured as follows: Improved Kalman filtering is applied to millimeter-wave signals to generate denoised water droplet distribution data; Vibration spectrum is generated by extracting time-frequency features from piezoelectric signals through wavelet transform processing; A dynamic weight allocation model is constructed to process the rate of change of water droplet concentration gradient, harmonic distortion rate, and rate of change of temperature difference slope to generate weighted parameters; Anti-icing strategies are generated by processing weighted parameters using digital twins.

[0008] In some implementations, the improved Kalman filter uses constant false alarm rate detection to eliminate moving noise, and the wavelet transform performs time-frequency analysis on a specific frequency band of the piezoelectric signal to generate spectral data containing harmonic frequency components and vibration amplitude components.

[0009] In some implementations, the dynamic weight allocation model uses vibration harmonic distortion rate, supercooled water droplet concentration gradient, and surface temperature difference slope as core input parameters, wherein the harmonic distortion rate weight is adaptively adjusted to generate the frequency band effectiveness weight according to wind speed changes.

[0010] In some implementations, the digital twin initiates strategy simulation when the comprehensive threat index exceeds a threshold, simulating the impact of different heating schemes on aerodynamic performance and outputting an optimized scheme that includes regional topological intensity distribution and time series.

[0011] In some implementations, the comprehensive threat index is calculated by integrating the absolute value of the water droplet concentration gradient, the logarithmic value of the harmonic distortion rate, and the rate of change of the temperature difference slope, and its threshold is determined by training with historical icing accident data.

[0012] In some implementations, the first formula for calculating the comprehensive threat index includes: in, This represents the comprehensive threat index. The rate of change of water droplet concentration gradient in the i-th spatial unit is derived from the three-dimensional concentration distribution model of the millimeter-wave detection unit. The harmonic distortion rate of the j-th time window is derived from the frequency feature extraction of the piezoelectric sensing unit. The slope of the temperature difference at the k-th temperature measurement point is derived from the dynamic calibration data of the infrared temperature measurement unit. The time decay coefficient of the k-th temperature measurement point is derived from historical icing accident data training; N represents the total number of spatial units, M represents the total number of time windows, and P represents the total number of temperature measurement points.

[0013] In some implementations, the second formula for calculating the harmonic distortion rate includes: in, Indicates harmonic distortion rate. The nth vibration harmonic frequency component is derived from the wavelet transform result of the piezoelectric signal. The reference energy value for the m-th frequency band is derived from a historical vibration spectrum database. This represents the p-th vibration amplitude component, derived from the time-domain analysis of the piezoelectric signal; The environmental disturbance coefficient of the p-th vibration component is derived from the unit's operating status parameters; The effective weight of the m-th frequency band is derived from the dynamic weight allocation model; Q represents the total number of frequency bands, R represents the total number of harmonic frequency components, and S represents the total number of vibration amplitude components.

[0014] In some implementations, the execution control module includes a distributed plus network and system linkage unit, wherein, The heating network implements zoned energy control based on an anti-icing strategy; The linkage unit interacts with the unit control system to perform operational adjustments.

[0015] At least one embodiment of the present invention uses a frequency-modulated continuous wave radar array, an orthogonal braided piezoelectric fiber array, and an infrared thermometry unit to comprehensively monitor and accurately capture the multidimensional dynamic changes in airborne water droplet distribution, blade vibration characteristics, and surface temperature field. Combined with the improved Kalman filter, wavelet transform, and dynamic weight allocation algorithm of the intelligent decision-making module, the traditional single threshold judgment is upgraded to a comprehensive threat index deduction based on digital twins, which significantly improves the spatiotemporal resolution and reliability of icing threat assessment. The execution control module realizes precise regional control of the heating network and coordinated optimization of the unit's operating status based on the generated anti-icing strategy. This avoids the energy waste of traditional whole-area heating and effectively suppresses material damage caused by local overheating or underheating. Ultimately, it reduces anti-icing energy consumption under complex meteorological conditions and significantly reduces the power generation efficiency loss of the unit under icing conditions. Attached Figure Description

[0016] Figure 1 This is a structural block diagram of a dynamic early warning system for icing of wind turbine blades provided by the present invention. Detailed Implementation

[0017] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0018] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items. The modifications “a” and “a plurality” as used in this disclosure are illustrative and not restrictive, and those skilled in the art will understand that they should be understood as “one or more” unless the context clearly indicates otherwise.

[0019] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0020] Traditional wind turbine anti-icing technology has long faced two major bottlenecks: First, the threshold alarm mechanism that relies on a single sensor (such as temperature or vibration) cannot capture the coupling relationship between water droplet distribution, structural response and temperature field during the icing process; second, the mismatch between static heating strategy and dynamic icing situation results in both energy waste and blind spots in protection.

[0021] Analysis of multiple blade icing incidents revealed significant nonlinear spatiotemporal characteristics in ice growth, a complexity that conventional techniques struggle to quantify. Therefore, this study incorporated moving target filtering algorithms from radar signal processing, piezoelectric fiber sensing networks from materials science, and temperature slope analysis from thermodynamics. By integrating the dynamic weighting model from aero-engine anti-icing control with digital twin technology, and through experimental verification, a three-module collaborative architecture was ultimately determined, achieving a paradigm shift from "passive response de-icing" to "active predictive anti-icing."

[0022] See Figure 1 , Figure 1A structural block diagram of a dynamic early warning system for wind turbine blade icing according to some embodiments of this specification is shown. The system includes an environmental situation awareness module, an intelligent decision-making module, and an execution control module. The environmental situation awareness module acquires airborne water droplet distribution data, blade vibration characteristics, and surface temperature field, and transmits the data to the intelligent decision-making module. The intelligent decision-making module performs fusion calculations on the water droplet distribution data, vibration characteristics, and temperature field to generate a threat assessment result, and outputs an anti-icing strategy containing heating power distribution and turbine control commands to the execution control module based on the assessment result. The execution control module controls the heating network to operate in zones according to the anti-icing strategy and coordinates the adjustment of the turbine's operating status.

[0023] The environmental situation awareness module refers to the system components used to collect environmental parameters related to blade icing. It works in concert through three methods: frequency-modulated continuous wave radar scanning the leading edge airspace of the blade, piezoelectric fiber array detecting vibration harmonics, and infrared thermography unit monitoring the surface temperature gradient. This enables three-dimensional capture of icing precursor factors. Airspace water droplet distribution data refers to parameters reflecting the liquid water content in the air surrounding the blade. By analyzing radar echo intensity and Doppler frequency shift, the spatial concentration of water droplets of different sizes is analyzed to quantify the distribution characteristics of icing material sources. Blade vibration characteristics refer to the spectral information reflecting the dynamic response of the structure. Piezoelectric fiber arrays collect changes in charge signals and extract characteristic frequency band energy through Fourier transform, which can identify structural stiffness degradation caused by ice accumulation. The surface temperature field refers to the two-dimensional temperature distribution on the outer surface of the blade. Using an infrared thermography unit to scan the skin temperature difference at a millisecond refresh rate, it can accurately locate the critical temperature region of phase transition.

[0024] The intelligent decision-making module can refer to a processor that performs multi-source data fusion calculations. It uses an improved Kalman filter to align the spatiotemporal reference, extract features using wavelet transform, and generate a threat index using a dynamic weighting algorithm. This index is used to transform raw data into executable protection commands. The threat assessment result can refer to a quantified icing risk level. Based on the coupling relationship between water droplet concentration, temperature change rate, and vibration distortion, an index in the zero-to-one range is calculated, enabling trend prediction of icing development. Heating power distribution can refer to the energy allocation scheme in the anti-icing strategy. The blade surface is divided into different heating level zones according to the threat index, enabling precise energy consumption control. Unit control commands can refer to operating parameter adjustment commands to maintain power generation efficiency. Dynamically correcting yaw angle and speed limits based on the icing threat level can reduce aerodynamic performance losses.

[0025] The execution control module can refer to the strategy execution mechanism, which independently adjusts the duty cycle of the heating elements in each zone through multiple PWM signals, achieving millimeter-level precision temperature field control. The heating network zoning can refer to modular electrothermal execution units, dividing the blade surface into hundreds of independently temperature-controllable micro-regions to meet the protection requirements of non-uniform icing. Unit operating status adjustment can refer to the adaptive control of the power generation system, which synchronously adjusts the pitch system, yaw motor, and other actuators according to the protection strategy to maintain stable output of the unit under icing conditions.

[0026] The present invention will be further described below through a specific embodiment: A wind power station is equipped with the wind turbine blade icing dynamic early warning system described in this invention. This system comprises a closed-loop system consisting of an environmental situation awareness module, an intelligent decision-making module, and an execution control module. The environmental situation awareness module uses a frequency-modulated continuous wave radar array to scan a 30-meter airspace at the leading edge of the blade in the 76-81 GHz band. It analyzes the spatial distribution density of water droplets with diameters of 50-500 micrometers using the Doppler effect, achieving a measurement accuracy of ±3%. An orthogonally woven piezoelectric fiber array is embedded in the blade skin with a 5×5 cm grid density, detecting vibration harmonic distortion in the 0.01-100 Hz range through changes in charge signal. An infrared thermography unit monitors the temperature gradient change on the blade surface from -40℃ to 80℃ in real time with a resolution of 0.1℃. These three types of sensors establish a data channel with the intelligent decision-making module via a CAN bus, with the sampling period synchronously controlled within 10 ms.

[0027] The intelligent decision-making module employs an improved Kalman filter algorithm to spatiotemporally align radar reflection intensity, vibration spectrum characteristics, and temperature slope. When a region is detected to have a water droplet concentration exceeding 0.5 g / m³ and a temperature within the critical range of -5°C to 0°C, wavelet transform analysis is initiated to analyze energy abrupt changes in the 2-15 Hz frequency band of the vibration signal. A dynamic weighting algorithm assigns initial weighting coefficients of 70% ambient humidity, 20% temperature change rate, and 10% vibration distortion based on historical icing data, generating an icing threat index in the 0-1 range within the digital twin. Specifically, when the threat index for a leaf segment exceeds 0.7, a level-three warning is triggered, and the coordinates of that region are marked.

[0028] The execution control module comprises 192 independent temperature control zones, each configured with three sets of redundant heating units. Upon receiving a decision command, a gradient heating strategy is implemented based on the threat index: the base power of 40W / m² is maintained in the 0.3-0.5 zone, increased to 120W / m² in the 0.5-0.7 zone, and the peak power of 240W / m² is activated in the zone exceeding 0.7, while simultaneously reducing the yaw speed by 15% in conjunction with the yaw system. Actual operation data shows that under -8℃ freezing fog conditions, this system reduces the average daily energy consumption of traditional full-area heating from 58kWh to 19kWh, while keeping the ice thickness at the blade leading edge below 2mm.

[0029] In a typical implementation scenario, when a 2.5MW unit encountered intermittent freezing rain, the radar first detected a cluster of water droplets with a concentration of 1.2 g / m³ at a distance of 15 meters from the leading edge of the blades. Simultaneously, the temperature in zone 3 dropped from -2℃ to -4.5℃ within 10 minutes. The decision module, considering the 35% energy increase in the 7.8Hz component of the vibration signal, determined the threat index of this area to be 0.82. The execution module then initiated 210 W / m² directional heating in the area from the blade root to 25% of the chord length, while the remaining areas remained in standby mode. After 20 minutes, monitoring showed that only a 0.5 mm frost layer had formed in the target area, and the unit's power curve deviation was controlled within 3%, verifying the system's precise protection capabilities.

[0030] The beneficial effects of one of the embodiments in this specification include at least the following: through the all-round monitoring of frequency-modulated continuous wave radar group, orthogonal braided piezoelectric fiber array and infrared temperature measurement unit, the multi-dimensional dynamic changes of airborne water droplet distribution, blade vibration characteristics and surface temperature field are accurately captured. Combined with the improved Kalman filter, wavelet transform and dynamic weight allocation algorithm of the intelligent decision module, the traditional single threshold judgment is upgraded to the comprehensive threat index inference based on digital twin, which significantly improves the spatiotemporal resolution and reliability of icing threat assessment. The execution control module realizes precise control of heating network partition and coordinated optimization of unit operation status according to the generated anti-icing strategy. This not only avoids the energy waste of traditional whole-area heating, but also effectively suppresses material damage caused by local overheating or underheating. Ultimately, it will reduce anti-icing energy consumption under complex meteorological conditions, and at the same time, it will significantly reduce the power generation efficiency loss of the unit under icing conditions.

[0031] In some embodiments, the environmental situation awareness module includes: a frequency-modulated continuous wave radar array that scans the airspace to form a three-dimensional concentration distribution model including particle size differentiation, and performs moving target filtering on the radar echo signal to generate the water droplet concentration gradient change rate; an orthogonally woven piezoelectric fiber array for collecting vibration signals and processing them through a frequency feature extraction algorithm to generate harmonic distortion rate; and an infrared temperature measurement unit for dynamically calibrating the surface temperature and calculating the temperature difference slope change rate at each temperature measurement point.

[0032] Frequency-modulated continuous wave radar arrays can refer to millimeter-wave radar arrays used to detect water droplets in the airspace. They transmit 76-81 GHz electromagnetic waves in a linear frequency-modulated manner and receive reflected signals. Fourier transform is used to convert the time-domain signal into frequency-domain information, enabling the creation of spatial distribution heat maps containing water droplets of different sizes. Three-dimensional concentration distribution models can refer to digital twins that quantify water droplet distribution. Based on radar echo intensity and propagation attenuation characteristics, the liquid water content per unit volume is calculated, allowing visualization of the spatial accumulation of freezing substances. Moving target filtering processing refers to algorithms that eliminate interference signals. Adaptive Kalman filters separate background noise from effective echoes, improving the signal-to-noise ratio for detecting the trajectory of tiny water droplets. Derivative parameters reflecting the freezing rate can be used to calculate spatial derivatives by comparing the differences in three-dimensional concentration distribution between adjacent time frames, providing early warning of rapid freezing risks in local areas.

[0033] Orthogonally woven piezoelectric fiber arrays refer to embedded vibration sensing networks. Lead zirconate titanate fibers are woven into the composite material skin at 0° / 90° angles, converting mechanical vibrations into electrical signals through the piezoelectric effect, thus capturing minute vibrations across the entire blade area. Harmonic distortion rate (HCR) is a quantitative indicator of the degree of structural vibration anomalies. By performing a fast Fourier transform on the acquired broadband signal and calculating the ratio of characteristic frequency band energy to fundamental frequency energy, it reflects the nonlinear changes in structural stiffness caused by ice accumulation.

[0034] Infrared temperature measurement units refer to non-contact temperature monitoring devices that use 8-14μm band thermal imaging sensors to scan the surface at a rate of 30 frames per second. By inverting the true temperature using the blackbody radiation law, they can achieve sub-second response for full-field temperature measurement. The rate of change of the temperature difference slope refers to a parameter characterizing the dynamic properties of the temperature field. By performing spatiotemporal difference calculations on multiple consecutive frames of infrared images, the acceleration of temperature change at each pixel can be calculated, identifying abnormal temperature regions caused by the release of latent heat of phase transition.

[0035] By combining the collaborative sensing of multimodal sensors with the fusion of characteristic parameters, comprehensive monitoring of the icing process, from macroscopic material distribution to microscopic phase transition characteristics, has been achieved. This provides a high-precision data foundation for intelligent anti-icing decision-making and effectively solves the technical problem of large blind spots in traditional single-sensor monitoring.

[0036] In some implementations, the intelligent decision-making module is further configured to: perform improved Kalman filtering on the millimeter-wave signal to generate denoised water droplet distribution data; extract time-frequency features from the piezoelectric signal through wavelet transform to generate a vibration spectrum; construct a dynamic weight allocation model to process the rate of change of water droplet concentration gradient, harmonic distortion rate, and rate of change of temperature slope to generate weighted parameters; and generate an anti-icing strategy by processing the weighted parameters through digital twin deduction.

[0037] Improved Kalman filtering refers to an optimized signal denoising algorithm that iteratively processes the original millimeter-wave echo signal by introducing an adaptive noise covariance matrix and a state prediction correction mechanism. This effectively eliminates measurement interference caused by atmospheric turbulence and mechanical vibration. The denoised water droplet distribution data refers to environmental parameters enhanced by signal enhancement. The state estimation function of the Kalman filter is used to reconstruct the true spatial distribution characteristics of water droplets, thereby improving the accuracy of icing risk assessment.

[0038] Wavelet transform processing can refer to non-stationary signal analysis methods. It employs Daubechies wavelet basis functions to perform multi-scale decomposition of piezoelectric signals, extracting energy distribution characteristics of specific frequency bands through joint time-frequency analysis, thus capturing structural resonance phenomena induced by ice accumulation. Time-frequency features refer to the joint domain characterization parameters of vibration signals. Identifying energy accumulation regions and instantaneous frequency change trajectories in characteristic frequency bands on the time-frequency spectrum generated by wavelet transform can reflect the time-varying laws of structural dynamic characteristics. The vibration spectrum refers to the frequency domain expression of the structural response. Power spectral density estimation and dominant frequency extraction are performed on the denoised time-frequency features to quantify the relative distribution ratio of vibration energy in different frequency bands.

[0039] The dynamic weight allocation model can refer to a multi-parameter fusion algorithm. Based on historical icing data, a neural network is trained to establish a nonlinear mapping relationship between water droplet concentration, vibration distortion, and temperature changes, enabling adaptive adjustment of parameter weights under different operating conditions. Weighting parameters can refer to the quantization coefficients of feature fusion. By weighting and normalizing the three types of monitoring parameters using the model's output weight values, a unified benchmark for assessing icing threats can be constructed. Digital twin simulation can refer to a virtual simulation process. Weighting parameters are injected into a three-dimensional blade model, and the thermodynamic-structural mechanics coupling equations are solved. By simulating the ice growth and shedding process under different anti-icing strategies in real time, the actual effectiveness of protective measures can be predicted. Anti-icing strategy can refer to the optimal control command set. Based on the twin simulation results, the heating power combination and unit adjustment scheme that minimizes energy consumption and achieves the required anti-icing effect are selected, enabling optimal allocation of protective resources.

[0040] By organically combining multi-level signal processing and intelligent decision-making algorithms, raw sensor data is transformed into executable and precise protection strategies. This avoids the blindness of traditional experience-based anti-icing and ensures the scientific nature and reliability of protection measures under different meteorological conditions.

[0041] In some implementations, the improved Kalman filter uses constant false alarm rate detection to eliminate moving noise, and the wavelet transform performs time-frequency analysis on a specific frequency band of the piezoelectric signal to generate spectral data containing harmonic frequency components and vibration amplitude components.

[0042] Constant false alarm rate (CFAR) detection refers to anti-interference techniques in radar signal processing. It maintains a constant false alarm probability by automatically adjusting the detection threshold, statistically analyzing local background noise power using a sliding window, and dynamically setting the amplitude threshold. This effectively suppresses interference from moving clutter such as birds on water droplet detection. Moving clutter refers to interference signals from non-target reflectors. Doppler frequency shift characteristics are used to distinguish between stationary backgrounds and moving objects. A dual discrimination mechanism of velocity filtering and amplitude screening eliminates false echoes from irrelevant moving objects in the environment. Specific frequency bands for piezoelectric signals refer to the key frequency range of structural vibration characteristics. Based on blade modal analysis, a monitoring frequency band of 20-200Hz is determined. A bandpass filter is used to preprocess the original signal, focusing on the characteristic frequency shift phenomenon caused by ice accumulation. Time-frequency analysis refers to joint domain processing methods for non-stationary signals. Complex Morlet wavelet basis functions are used to perform multi-resolution decomposition of the signal. By constructing a time-spectrum graph to track the trajectory of characteristic frequencies over time, transient vibration events and slowly changing processes can be captured. Harmonic frequency components refer to the harmonics of a vibration signal. By performing peak search on the time-frequency spectrum to extract frequency components that are integer multiples of the fundamental frequency, and then calculating the instantaneous phase using Hilbert transform, the nonlinear vibration characteristics of the structure can be quantitatively characterized. Vibration amplitude components refer to the energy parameters of mechanical vibration. By integrating in the time-frequency domain to calculate the root mean square amplitude of each characteristic frequency band, and then normalizing it in conjunction with the damping characteristics of the blade material, the influence of ice mass on the dynamic response of the structure can be reflected.

[0043] By extracting high signal-to-noise ratio feature parameters through advanced signal processing algorithms, the sensitivity and reliability of icing monitoring are significantly improved, providing accurate data support for intelligent anti-icing decision-making, while effectively reducing the false alarm rate caused by environmental interference.

[0044] In some implementations, the dynamic weight allocation model uses vibration harmonic distortion rate, supercooled water droplet concentration gradient, and surface temperature difference slope as core input parameters, wherein the harmonic distortion rate weight is adaptively adjusted to generate the frequency band effectiveness weight according to wind speed changes.

[0045] The supercooled water droplet concentration gradient refers to the spatial rate of change of liquid water content. It is calculated by normalizing the concentration difference between adjacent spatial units using millimeter-wave radar data to reflect the regional distribution characteristics of icing risk. The surface temperature difference slope is a parameter characterizing the dynamic properties of the temperature field. Spatiotemporal difference operations are performed on continuous temperature field data collected by infrared thermography units to calculate the second derivative of the rate of temperature change per unit time, which characterizes the intensity of latent heat release during phase change. The frequency band effectiveness weight refers to the importance coefficient of vibration characteristics. A frequency band sensitivity database is established based on real-time wind speed sensor data. The contribution weight of each harmonic frequency band is dynamically allocated using a fuzzy logic algorithm, which can optimize the accuracy of icing judgment under different wind conditions.

[0046] Through a multi-parameter dynamic weighted fusion mechanism, adaptive optimization of icing risk assessment was achieved. This not only considered the influence of ambient wind speed on vibration characteristics but also maintained the sensitivity of temperature field and water droplet distribution parameters, significantly improving the reliability of anti-icing decisions under complex meteorological conditions.

[0047] In some implementations, the digital twin initiates strategy simulation when the comprehensive threat index exceeds a threshold, simulating the impact of different heating schemes on aerodynamic performance and outputting an optimized scheme that includes regional topological intensity distribution and time series.

[0048] The comprehensive threat index refers to a comprehensive assessment of icing risk. It uses a weighted fusion of parameters such as water droplet concentration gradient, harmonic distortion rate, and temperature difference slope, and employs a fuzzy comprehensive evaluation algorithm to calculate a normalized risk score, enabling a quantitative characterization of multi-dimensional threats. Strategy simulation refers to the simulation verification process of protection schemes. Current environmental parameters and structural states are injected into a digital twin environment, and virtual experiments with multiple heating power combinations are executed in parallel. The optimal scheme is selected by comparing the degree of aerodynamic performance improvement. Aerodynamic performance refers to the aerodynamic characteristics of the airfoil. Key parameters such as lift-to-drag ratio, pressure distribution, and flow separation point are obtained through computational fluid dynamics simulation, quantitatively assessing the impact of ice accumulation and heating measures on flight performance. Regional topology intensity distribution refers to the spatial configuration scheme of heating power. Based on a heat conduction model, the required energy density for each zone is calculated. Combined with the material's temperature resistance limit, the heat flux distribution is optimized to achieve the best balance between protection effectiveness and structural safety. Time series optimization scheme refers to the timing strategy of heating control. A dynamic programming algorithm is used to arrange the activation sequence and duration time interval of each zone, considering thermal inertia and energy storage characteristics, maximizing thermal energy utilization efficiency.

[0049] The overall benefits of this system are as follows: by using digital twin technology to achieve virtual verification and optimization of protection strategies, it not only ensures the scientific validity and effectiveness of anti-icing measures, but also avoids the safety risks and energy waste caused by actual trial and error, and significantly improves the active protection capabilities of aircraft under complex weather conditions.

[0050] In some implementations, the comprehensive threat index is calculated by integrating the absolute value of the water droplet concentration gradient, the logarithmic value of the harmonic distortion rate, and the rate of change of the temperature difference slope, and its threshold is determined by training with historical icing accident data.

[0051] The absolute value of the water droplet concentration gradient can refer to the intensity index of changes in liquid water content. It is calculated by taking the absolute value of the water droplet concentration difference between adjacent spatial units using millimeter-wave radar detection data to eliminate directional influences and highlight the intensity characteristics of icing risk. The logarithmic value of the harmonic distortion rate can refer to the quantitative characterization of the nonlinearity of the vibration signal. A base-10 logarithmic transformation is applied to the original harmonic distortion rate to compress the dynamic range of the data, improving the resolution of small signals and enhancing the sensitivity of identifying weak icing features. The rate of change of the temperature difference slope can refer to a higher-order parameter of the dynamic characteristics of the temperature field. A second-order difference operation is performed on continuously acquired surface temperature data to calculate the change in the derivative of the temperature difference slope per unit time, which can capture abrupt changes in thermodynamic properties during phase transitions. Historical icing accident data refers to the basic sample set for system reliability verification. It collects records of icing events under typical meteorological conditions, including multi-dimensional information such as environmental parameters, structural response, and accident consequences. Risk feature patterns are extracted using machine learning algorithms.

[0052] The comprehensive threat index constructed by the multi-parameter nonlinear fusion algorithm realizes a full-dimensional quantitative assessment of icing risk. Combined with the dynamic threshold setting method driven by historical data, it significantly improves the adaptability and accuracy of the early warning system and provides reliable decision support for aviation safety.

[0053] In some implementations, the first formula for calculating the comprehensive threat index includes: in, This represents the comprehensive threat index. The rate of change of water droplet concentration gradient in the i-th spatial unit is derived from the three-dimensional concentration distribution model of the millimeter-wave detection unit. The harmonic distortion rate of the j-th time window is derived from the frequency feature extraction of the piezoelectric sensing unit. The slope of the temperature difference at the k-th temperature measurement point is derived from the dynamic calibration data of the infrared temperature measurement unit. The time decay coefficient of the k-th temperature measurement point is derived from historical icing accident data training; N represents the total number of spatial units, M represents the total number of time windows, and P represents the total number of temperature measurement points.

[0054] The first calculation formula can refer to the mathematical model construction method of the threat index. It uses square root operations to handle the spatial cumulative effect of the water droplet concentration gradient change rate, employs logarithmic transformation and cube root operations to adjust the contribution of harmonic distortion rate, and combines exponential functions to amplify the time-dependent characteristics of the temperature difference slope, enabling the fusion of heterogeneous data from multiple sources. The time decay coefficient can refer to the time-dependent parameter of temperature influence. Based on historical data, it statistically analyzes the relaxation time characteristics of temperature changes at each measurement point, and determines the decay rate under different environmental conditions through exponential fitting, reflecting the inertial characteristics of the heat conduction process. The total number of spatial units can refer to the resolution parameter of the monitoring area. Based on the angular resolution and detection range of the millimeter-wave radar, the monitoring airspace is divided into equal-volume cubic grids, and the number of these grids determines the accuracy of the concentration gradient calculation. The total number of time windows can refer to the number of segments in the monitoring duration, determined by dividing the total monitoring duration by the duration of a single window, affecting the temporal resolution of the harmonic distortion rate statistical characteristics. The total number of temperature measurement points can refer to the sampling density of the temperature field. Based on the field of view and scanning frequency of the infrared sensor, it determines the number of effective temperature data points that can be acquired per unit time, relating to the spatial representativeness of the temperature difference slope.

[0055] By establishing a first calculation formula that integrates spatiotemporal multidimensional features, an accurate quantitative assessment of icing threats was achieved. This formula considers both the spatial distribution characteristics of parameters and the temporal evolution patterns, providing a scientific and reliable decision-making basis for aircraft anti-icing systems.

[0056] In some implementations, the second formula for calculating the harmonic distortion rate includes: in, Indicates harmonic distortion rate. The nth vibration harmonic frequency component is derived from the wavelet transform result of the piezoelectric signal. The reference energy value for the m-th frequency band is derived from a historical vibration spectrum database. This represents the p-th vibration amplitude component, derived from the time-domain analysis of the piezoelectric signal; The environmental disturbance coefficient of the p-th vibration component is derived from the unit's operating status parameters; The effective weight of the m-th frequency band is derived from the dynamic weight allocation model; Q represents the total number of frequency bands, R represents the total number of harmonic frequency components, and S represents the total number of vibration amplitude components.

[0057] The second calculation formula can refer to a composite quantization model of harmonic distortion rate. It integrates the characteristic contributions of each frequency band through multiplication, uses the square root to process the weighted sum of frequency energy ratio and amplitude cube, and combines an effectiveness weight for exponential adjustment, enabling a multi-dimensional fusion assessment of vibration characteristics. Vibration harmonic frequency components can refer to the spectral characteristics of structural vibration. Wavelet packet decomposition is performed on the piezoelectric sensor signal to extract characteristic frequency components within each scale subspace. Instantaneous frequency values ​​are calculated using Hilbert transform, reflecting changes in stiffness characteristics caused by icing. Wavelet transform results can refer to the processed output of non-stationary signals. The db4 wavelet basis function is selected to perform multi-scale decomposition on the original vibration signal. Threshold denoising and reconstruction algorithms are used to obtain the time-frequency characteristics of each frequency band, supporting accurate extraction of harmonic components. The frequency band reference energy value can refer to the reference standard for vibration characteristics. Typical spectral energy distributions under different working conditions are statistically analyzed from historical databases to establish a normal distribution model of energy values ​​for each frequency band. The normal fluctuation range is determined using the 3σ principle. Vibration amplitude components refer to the intensity parameters of the time-domain signal. Peak detection and envelope analysis are performed on the preprocessed piezoelectric signal to calculate the maximum offset within each vibration cycle. Random interference is eliminated through moving average filtering. Environmental interference coefficients refer to correction factors for operating conditions. Flight parameters such as engine speed, airflow velocity, and angle of attack are collected, and a mapping relationship between vibration amplitude and operating conditions is established through multivariate regression analysis to eliminate the influence of non-icing factors. The total number of frequency bands refers to the resolution parameter of the spectrum analysis. Based on the Nyquist sampling theorem and sensor frequency response characteristics, the effective frequency band is divided into several sub-intervals of equal bandwidth or proportion. The number of sub-intervals affects the precision of feature extraction. The total number of harmonic frequency components refers to the dimension of the spectral features, determined by the number of wavelet decomposition levels and the number of effective coefficients in each level, reflecting the number of independent frequency components contained in the vibration signal. The total number of vibration amplitude components refers to the sampling density of the time-domain features. The number of effective amplitude data points that can be extracted per unit time is determined based on the signal sampling rate and analysis duration, which relates to the statistical reliability of vibration intensity.

[0058] By establishing a second calculation formula that comprehensively considers frequency domain and time domain characteristics, the scientific extraction of icing characteristics from vibration signals is realized. This not only preserves the independent contribution of each frequency band but also achieves the organic fusion of characteristics, providing a highly sensitive diagnostic basis for aircraft icing monitoring.

[0059] In some implementations, the execution control module includes a distributed heating network and a system linkage unit, wherein the heating network implements zoned energy control according to the anti-icing strategy; and the linkage unit interacts with the unit control system to perform operational adjustments.

[0060] Distributed heating networks refer to the topology of anti-icing actuators, using a CAN bus to connect heating elements in key areas such as the wing leading edge and engine intakes. Gradient power allocation is implemented based on threat index zoning calculations, enabling precise energy supply positioning. Anti-icing strategies refer to the decision rules for de-icing operations. Warning levels are categorized based on a comprehensive threat index threshold, establishing heating power curves and duration parameters corresponding to different risk levels, and achieving dynamic adjustment through fuzzy control algorithms. Zonal energy control refers to optimized thermal distribution methods, dividing the aircraft surface into several independent temperature-controlled zones. A PID controller adjusts the duty cycle of heating elements in each zone in real time, balancing de-icing effectiveness with energy consumption. System linkage units refer to interface devices for multi-system collaboration. Data exchange is established with the flight control computer via the ARINC429 bus, translating icing threat levels into recommended adjustments to flight parameters such as airspeed and angle of attack, supporting crew decision-making and execution. The flight control system can be considered the core of flight management. It receives icing risk parameters from the linkage unit, comprehensively assesses the current flight status, and generates control commands such as control surface deflection and thrust adjustment. It then optimizes flight attitude through the fly-by-wire system. Operational adjustments refer to the adaptation and modification of flight parameters. Based on the icing risk level, it automatically triggers protective logic such as airspeed maintenance and altitude limitation, while providing operational suggestions and prompts on the human-machine interface, balancing safety and operational comfort requirements.

[0061] Through the intelligent linkage between the distributed heating network and the flight control system, a closed-loop control of the entire process from icing monitoring to protection execution is achieved, which not only ensures timely de-icing of key components but also optimizes flight performance parameters, significantly improving flight safety margins under complex weather conditions.

[0062] In some embodiments, the distributed heating network is constructed using transparent indium tin oxide thin-film electrodes, and its zone control unit implements gradient power output based on the differences in aerodynamic characteristics of the blade regions.

[0063] In some embodiments, the gradient power output method employs high-intensity heating in the leaf tip region to counteract the rotational acceleration effect, maintains medium-power anti-icing in the leaf middle region, and implements basic power anti-frost measures in the leaf root region.

[0064] In some implementations, the system linkage unit interacts with the SCADA (Supervisory Control And Data Acquisition) system through a preset transmission protocol, and sends a triplet command containing the yaw angle correction amount and duration when icing is severe.

[0065] In some implementations, the yaw angle correction is dynamically calculated based on the comprehensive threat index, causing the unit to temporarily deviate from the prevailing wind direction by a specific angle to reduce the icing rate on the windward side.

[0066] Transparent indium tin oxide (ITO) thin-film electrodes can guide the material composition of electric heating elements. Nanoscale ITO thin films are deposited on composite substrates using magnetron sputtering. The aerodynamic characteristics of the blade region refer to the flow field distribution features on the airfoil surface. Based on CFD simulations, pressure coefficient distribution cloud maps at different angles of attack are obtained, and critical regions with significant Mach number gradient changes are delineated, guiding differentiated heating power allocation strategies. The rotational acceleration effect refers to the unique operating conditions at the blade tip. Considering that the centrifugal force generated by the high-speed rotation of the blade increases the water droplet impact velocity by 3-5 times, pulse width modulation technology is used to increase the heating power to twice that of the conventional region, effectively preventing ice crystal accumulation in the critical region. Base power anti-frost refers to the low-energy maintenance mode. A baseline heating amount is applied to the blade root region to maintain the material surface temperature at 273-275K. A PID controller automatically compensates for the ambient temperature drop, preventing frost formation under static conditions without affecting structural strength. Triplet commands refer to the standardized format of control parameters, encoding yaw correction angle, duration, and priority identifiers into 32-bit floating-point arrays. A double-buffering mechanism ensures the timing accuracy of command execution, preventing control command conflicts or loss. Windward icing rate refers to the dynamic indicator of ice growth. Combining millimeter-wave radar liquid water content data and infrared thermography surface temperature gradients, an ice growth prediction model based on heat and mass transfer theory is established to quantitatively evaluate the anti-icing effectiveness of yaw corrections.

[0067] By using a gradient power distribution that matches the transparent heating film with aerodynamic characteristics, intelligent anti-icing protection for key aircraft components is achieved. At the same time, through deep integration with the SCADA system, the flight control strategy can dynamically respond to changes in icing risk, maximizing energy efficiency while ensuring flight safety.

[0068] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this invention. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A dynamic early warning system for icing on wind turbine blades, characterized in that, include: The system comprises an environmental situation awareness module, an intelligent decision-making module, and an execution control module. The environmental situation awareness module acquires airspace water droplet distribution data, blade vibration characteristics, and surface temperature field, and transmits the data to the intelligent decision-making module. The intelligent decision-making module performs fusion calculations on the water droplet distribution data, vibration characteristics, and temperature field to generate a threat assessment result. Based on the assessment result, it outputs an anti-icing strategy containing heating power distribution and unit control commands to the execution control module. The execution control module controls the heating network partitions to operate according to the anti-icing strategy and coordinates the adjustment of the unit's operating status.

2. The system according to claim 1, characterized in that, The environmental situation awareness module includes: The frequency-modulated continuous wave radar array scans the airspace to form a three-dimensional concentration distribution model that includes particle size differentiation, and performs moving target filtering on the radar echo signal to generate the water droplet concentration gradient change rate. An array of orthogonally woven piezoelectric fibers is used to collect vibration signals, which are then processed by a frequency feature extraction algorithm to generate harmonic distortion rate. The infrared temperature measurement unit is used to dynamically calibrate the surface temperature and calculate the rate of change of the temperature difference slope at each measurement point.

3. The system according to claim 1, characterized in that, The intelligent decision-making module is further configured as follows: Improved Kalman filtering is applied to millimeter-wave signals to generate denoised water droplet distribution data; Vibration spectrum is generated by extracting time-frequency features from piezoelectric signals through wavelet transform processing; A dynamic weight allocation model is constructed to process the rate of change of water droplet concentration gradient, harmonic distortion rate, and rate of change of temperature difference slope to generate weighted parameters; Anti-icing strategies are generated by processing weighted parameters using digital twins.

4. The system according to claim 3, characterized in that, The improved Kalman filter uses constant false alarm rate detection to eliminate moving noise, and the wavelet transform performs time-frequency analysis on a specific frequency band of the piezoelectric signal to generate spectral data containing harmonic frequency components and vibration amplitude components.

5. The system according to claim 3, characterized in that, The dynamic weight allocation model uses vibration harmonic distortion rate, supercooled water droplet concentration gradient and surface temperature difference slope as core input parameters, wherein the harmonic distortion rate weight is adaptively adjusted according to wind speed changes to generate frequency band effectiveness weights.

6. The system according to claim 3, characterized in that, When the comprehensive threat index exceeds a threshold, the digital twin initiates strategy simulation to model the impact of different heating schemes on aerodynamic performance and outputs an optimized scheme that includes regional topological intensity distribution and time series.

7. The system according to claim 6, characterized in that, The comprehensive threat index is calculated by integrating the absolute value of the water droplet concentration gradient, the logarithmic value of the harmonic distortion rate, and the rate of change of the temperature difference slope. Its threshold is determined by training with historical icing accident data.

8. The system according to claim 7, characterized in that, The first formula for calculating the comprehensive threat index includes: in, This represents the comprehensive threat index. The rate of change of water droplet concentration gradient in the i-th spatial unit is derived from the three-dimensional concentration distribution model of the millimeter-wave detection unit. The harmonic distortion rate of the j-th time window is derived from the frequency feature extraction of the piezoelectric sensing unit. The slope of the temperature difference at the k-th temperature measurement point is derived from the dynamic calibration data of the infrared temperature measurement unit. The time decay coefficient of the k-th temperature measurement point is derived from historical icing accident data training; N represents the total number of spatial units, M represents the total number of time windows, and P represents the total number of temperature measurement points.

9. The system according to claim 8, characterized in that, The second formula for calculating harmonic distortion rate includes: in, Indicates harmonic distortion rate. The nth vibration harmonic frequency component is derived from the wavelet transform result of the piezoelectric signal. The reference energy value for the m-th frequency band is derived from a historical vibration spectrum database. This represents the p-th vibration amplitude component, derived from the time-domain analysis of the piezoelectric signal; The environmental disturbance coefficient of the p-th vibration component is derived from the unit's operating status parameters; The effective weight of the m-th frequency band is derived from the dynamic weight allocation model; Q represents the total number of frequency bands, R represents the total number of harmonic frequency components, and S represents the total number of vibration amplitude components.

10. The system according to claim 1, characterized in that, The execution control module includes a distributed network and system linkage unit, wherein... The heating network implements zoned energy control based on an anti-icing strategy; The linkage unit interacts with the unit control system to perform operational adjustments.

Images