A method for controlling thyristors of a wind turbine blade electric pulse de-icing device

Abstract

The present application relates to the technical field of wind turbine control, and more particularly to a thyristor control method for a wind turbine blade electric pulse deicing device, which comprises the following steps: obtaining multi-dimensional data and fusing the data into blade working condition perception data sets to provide a rich and comprehensive information base; obtaining multi-time sequence data to train a model and combining calibration data to obtain blade surface icing thickness distribution data; dividing pulse energy demand levels and outputting a distribution matrix to realize differentiated division of energy demand; calculating a thyristor conduction angle adjustment amount and a pulse width modulation parameter to improve the accuracy of control parameters; performing low-temperature compensation correction on the thyristor to ensure the reliability of the trigger pulse sequence in a low-temperature environment; performing differentiated electric pulse deicing operations to improve deicing efficiency and targeting; and obtaining deicing completion degree in real time and dynamically updating regional energy density distribution matrix to realize adaptive optimization control of the deicing process.

Description

Technical Field

[0001] This invention relates to the field of wind turbine control technology, and in particular to a thyristor control method for an electrical pulse de-icing device for wind turbine blades. Background Technology

[0002] When wind turbines operate in cold and humid environments, icing easily forms on the blade surface, altering the blade's aerodynamic shape, increasing rotational loads, and reducing power generation efficiency. In severe cases, it can even lead to structural fatigue or safety accidents. Electrical pulse de-icing technology, as an active de-icing solution, applies high-voltage, short-duration electrical pulses to the icing area by pre-embedding electrodes inside or on the blade. Utilizing Joule heating and potential electromagnetic force effects, it achieves rapid ice removal, offering advantages such as low energy consumption, fast response, and no reliance on chemical media.

[0003] However, in real-world wind power operation scenarios, blade icing exhibits spatial non-uniformity and temporal dynamics, and is influenced by a combination of factors such as wind speed, rotational speed, and ambient temperature and humidity. Existing thyristor control methods often apply uniform energy allocation to the entire blade or a few designated areas, failing to differentiate energy distribution based on the actual icing thickness in different regions of the blade. This results in energy waste in lightly iced areas and incomplete de-icing in heavily iced areas. Furthermore, existing methods lack real-time compensation mechanisms for wind speed fluctuations, aerodynamic cooling effects caused by blade rotation, and thyristor performance degradation at low temperatures. This leads to a mismatch between de-icing energy allocation and actual demand, resulting in unstable de-icing performance and high energy consumption under varying operating conditions. Additionally, trigger failures or delays are common at low temperatures, affecting system reliability and limiting the widespread application of this technology in wind turbines operating in complex meteorological regions. Therefore, this invention proposes a thyristor control method for an electrical pulse de-icing device for wind turbine blades. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the background art, and to propose a thyristor control method for an electric pulse de-icing device for wind turbine blades.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A thyristor control method for an electrical pulse de-icing device for wind turbine blades, comprising: S1. Acquire icing distribution data, real-time wind speed data, blade rotation speed data, and blade surface temperature data on the surface of wind turbine blades, and integrate them into a blade operating condition sensing dataset. S2. Based on the blade condition sensing dataset, ice thickness distribution data is obtained through the ice thickness-region correlation model. S3. Based on the ice thickness distribution data, classify the pulse energy demand level of each grid region and output the regional energy density distribution matrix; S4. Based on the regional energy density distribution matrix, real-time wind speed data and blade rotation speed data, calculate the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region. S5. Based on the blade surface temperature data, perform low-temperature compensation correction on the thyristor and generate a trigger pulse sequence; S6. Input the trigger pulse sequence to the thyristor drive circuit of the corresponding grid area to perform electrical pulse de-icing control; S7. During the de-icing process, the de-icing completion rate of each grid area is obtained in real time, and the regional energy density allocation matrix is ​​dynamically updated based on the de-icing completion rate.

[0006] Furthermore, data on ice thickness distribution is obtained, including: The blade is divided into several grid regions along its spanwise and tangential directions; Based on historical monitoring data and blade condition sensing dataset, temperature time series data, wind speed time series data, and humidity time series data for each grid area are constructed. Temperature time-series data, wind speed time-series data, and humidity time-series data for each grid region are used as training samples, and the actual measured icing thickness is used as the label to train a deep neural network model; wherein, the deep neural network model constitutes an icing thickness-region correlation model. The temperature time series data, wind speed time series data, and humidity time series data of each grid area in the current cycle's blade condition sensing dataset are input into a pre-trained deep neural network model, which outputs the predicted icing thickness value for each grid area. By combining the predicted icing thickness with the calibration data collected by the real-time icing detection sensor, the icing thickness distribution data on the blade surface is obtained.

[0007] Furthermore, the pulse energy demand levels for each grid region are divided, including: Read the ice thickness value for each grid region in the ice thickness distribution data; If the ice thickness is less than or equal to the first thickness threshold, the grid area is classified as the first energy demand level. If the ice thickness is greater than the first thickness threshold and less than or equal to the second thickness threshold, then the grid area is classified as the second energy demand level. If the ice thickness is greater than the second thickness threshold, the grid area is classified as the third energy demand level. The first energy demand level, the second energy demand level, and the third energy demand level constitute the pulse energy demand level.

[0008] Furthermore, the output regional energy density distribution matrix includes: Assign a location weight coefficient to each grid region; Multiply the pulse energy demand level of each grid region by its location weight coefficient to obtain the weighted energy demand value of that grid region; The weighted energy demand values ​​of all grid regions are arranged according to spatial location to generate a regional energy density distribution matrix.

[0009] Further, the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region are calculated, including: Based on the regional energy density distribution matrix, determine the required basic pulse width and basic thyristor conduction angle for each grid region; Real-time wind speed and blade rotation speed data are obtained from the blade condition sensing dataset. The wind speed-rotation speed coupling influence factor is calculated, and the basic thyristor conduction angle is corrected using this factor to obtain the actual conduction angle. The wind speed fluctuation rate is calculated based on real-time wind speed data. When the wind speed fluctuation rate exceeds the set fluctuation threshold, the base pulse width is dynamically adjusted based on the wind speed fluctuation attenuation factor to obtain the actual pulse width.

[0010] Furthermore, the thyristor is subjected to low-temperature compensation correction to generate a trigger pulse sequence, including: The blade surface temperature data in the blade condition sensing dataset is obtained and it is determined whether it is lower than the preset low temperature threshold. If the current blade surface temperature is lower than the preset low temperature threshold, the low temperature compensation coefficient is calculated based on the temperature difference value, and the thyristor trigger pulse parameters, including voltage amplitude, leading edge steepness and pre-trigger time, are adjusted using this coefficient. Based on the actual conduction angle, actual pulse width, and trigger pulse parameters after low-temperature compensation, the thyristor trigger pulse sequence corresponding to each grid region is generated.

[0011] Furthermore, the electrical pulse de-icing control is executed, including: The trigger pulse sequence is input to the thyristor drive circuit of the corresponding grid area to perform differentiated electrical pulse de-icing operations; among them, a multi-pulse triggering mode is used for areas with high ice thickness, and a single-pulse triggering mode is used for areas with low ice thickness.

[0012] Furthermore, the de-icing completion rate of each grid area is obtained, including: Vibration spectrum data of the blades are collected by a vibration acceleration sensor, and temperature distribution data of the blade surface is collected by an infrared thermal imager. Analyze the blade vibration spectrum data and extract the amplitude of the characteristic frequency components caused by ice shedding; Analyze surface temperature distribution data and calculate the rate of temperature change in each grid region before and after the de-icing operation; The amplitude of the characteristic frequency component and the rate of temperature change are input into the completion evaluation function, and the de-icing completion quantification value of each grid region is output. If the quantification value of de-icing completion is lower than the preset completion threshold, the de-icing of that area is determined to be incomplete.

[0013] Furthermore, the regional energy density allocation matrix is ​​updated, including: For grid areas that are determined to be incompletely de-iced, the pulse energy demand level corresponding to that grid area will be increased by one level in the current regional energy density allocation matrix; Set an efficiency saturation threshold; If the quantified value of de-icing completion remains higher than the efficiency saturation threshold for several consecutive control cycles, then in the regional energy density allocation matrix, the pulse energy demand level corresponding to that grid area will be reduced by one level, or de-icing energy will not be allocated temporarily. Using the updated regional energy density allocation matrix as input to step S4 in the next control cycle, the base pulse width and base thyristor conduction angle of each grid region are redefined.

[0014] Furthermore, during the execution of step S6, a thyristor health status monitoring step is also performed simultaneously: Real-time monitoring of the on-state voltage drop and trigger response time of each thyristor; If the on-state voltage drop of any thyristor continuously exceeds the normal range or the trigger response time exceeds the set time tolerance, the thyristor is determined to be in an abnormal state. When a thyristor is determined to be in an abnormal state, the backup trigger channel for that grid area is activated, or de-icing compensation is performed by adjusting the regional energy density distribution matrix of adjacent grid areas.

[0015] Compared with existing technologies, the beneficial effects of the thyristor control method for an electrical pulse de-icing device for wind turbine blades provided by this invention are as follows: 1) By comprehensively acquiring multi-dimensional data on blade surface icing distribution, real-time wind speed, blade rotation speed, and surface temperature, and integrating them into a blade condition sensing dataset, a rich and accurate data foundation is provided for subsequent precise analysis of blade status and formulation of reasonable de-icing strategies, effectively improving the targeting of de-icing control; by constructing an icing thickness-region correlation model to obtain icing thickness distribution data, combining historical and real-time data, using deep neural network models for prediction, and combining real-time detection sensor calibration, the icing situation in each region of the blade can be more accurately grasped, providing a reliable basis for subsequent energy allocation; 2) Based on the icing thickness, the pulse energy demand level is divided, and a regional energy density distribution matrix is ​​output to realize differentiated assessment of the de-icing energy demand of different areas of the blade, so as to make the energy distribution more reasonable, avoid energy waste or insufficient energy, and improve the de-icing efficiency; based on the regional energy density distribution matrix, real-time wind speed and blade speed data, the thyristor conduction angle adjustment and pulse width modulation parameters are calculated, taking into account a variety of real-time operating conditions, so that the thyristor control parameters can dynamically adapt to the actual operating environment and ensure the stability of the electric pulse de-icing effect; 3) Based on blade surface temperature data, low-temperature compensation correction is performed on the thyristor to generate a trigger pulse sequence, effectively solving the problem of thyristor performance being affected by low-temperature environments, ensuring reliable triggering even at low temperatures, and improving the stability of the de-icing system in harsh environments; by inputting the trigger pulse sequence into the thyristor drive circuit to execute electrical pulse de-icing control, differentiated de-icing operations are achieved; different triggering modes are adopted for areas with different ice thicknesses, improving the de-icing effect while reducing unnecessary energy consumption; closed-loop control is formed by acquiring the de-icing completion rate in real time and dynamically updating the regional energy density distribution matrix; the energy distribution strategy is adjusted in a timely manner according to the actual de-icing effect to avoid over-de-icing or incomplete de-icing, further improving the intelligence level and overall performance of the de-icing system. Attached Figure Description

[0016] Figure 1 This is a flowchart of a thyristor control method for an electrical pulse de-icing device for wind turbine blades proposed in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0018] Reference Figure 1 A thyristor control method for an electrical pulse de-icing device for wind turbine blades, comprising: S1. Acquire icing distribution data, real-time wind speed data, blade rotation speed data, and blade surface temperature data on the surface of wind turbine blades, and integrate them into a blade operating condition sensing dataset. S2. Based on the blade condition sensing dataset, the ice thickness distribution data of different grid regions of the blade is obtained through the ice thickness-region correlation model. S3. Based on the ice thickness distribution data, classify the pulse energy demand level of each grid region and output the regional energy density distribution matrix; S4. Based on the regional energy density distribution matrix, real-time wind speed data and blade rotation speed data, calculate the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region. S5. Based on the blade surface temperature data, perform low-temperature compensation correction on the thyristor to generate a trigger pulse sequence that adapts to a wide temperature range. S6. Input the trigger pulse sequence to the thyristor drive circuit of the corresponding grid area to execute differentiated electrical pulse de-icing control; S7. During the de-icing process, the de-icing completion rate of each grid area is obtained in real time, and the regional energy density distribution matrix is ​​dynamically updated according to the de-icing completion rate, thereby adjusting the subsequent trigger pulse sequence.

[0019] It should be further explained that, in the specific implementation process, obtaining icing thickness distribution data includes: The blade is divided into several grid regions along its spanwise and tangential directions; Based on historical monitoring data and blade condition sensing dataset, temperature time series data, wind speed time series data, and humidity time series data for each grid area are constructed. Temperature time-series data, wind speed time-series data, and humidity time-series data for each grid region are used as training samples, and the actual measured icing thickness is used as the label to train a deep neural network model; wherein, the deep neural network model constitutes an icing thickness-region correlation model. The temperature time series data, wind speed time series data, and humidity time series data of each grid area in the current cycle's blade condition sensing dataset are input into a pre-trained deep neural network model, which outputs the predicted icing thickness value for each grid area. By combining the predicted ice thickness with the calibration data collected by the real-time ice detection sensor, the ice thickness distribution data on the blade surface is obtained. Specifically, the spatial grid division of wind turbine blades includes: using the leading edge and trailing edge as references, dividing the blade along the spanwise direction from the blade root to the blade tip into N equal segments (e.g., N=10, where N is the number of spanwise segments), and along the chordwise direction from the leading edge to the trailing edge into M equal segments (e.g., M=5, where M is the number of chordwise segments), thereby dividing the entire blade surface into N×M rectangular grid regions, and assigning a unique spatial coordinate identifier to each grid region; When constructing the input time-series dataset for each grid region, the historical monitoring database stored for a long time is called and the current blade condition sensing dataset is accessed; for each grid region, three sets of time-series data are extracted or reconstructed, including temperature time-series data, wind speed time-series data and humidity time-series data; The construction of a deep neural network model includes: a hybrid structure model comprising an input layer, multiple hidden layers, and an output layer; receiving preprocessed time-series data of temperature, wind speed, and humidity for a specific grid region as input feature vectors; employing an architecture combining convolutional layers and long short-term memory (LSM) network layers: one-dimensional convolutional layers extract local features from the time-series data; LSM network layers capture long-term dependencies in the time-series data; each hidden layer contains multiple neurons (or computational units); in the model, neurons between the input layer and the first hidden layer, as well as between adjacent hidden layers, are fully connected or connected in specific patterns (such as those corresponding to convolution operations). (Local connections); each connection corresponds to an adjustable weight parameter, which determines the degree and importance of the influence of the output value of the previous layer neuron on the input value of the next layer neuron; nonlinear activation functions (e.g., rectified linear units or hyperbolic tangent functions) are introduced in the internal calculations of the convolutional layers and long short-term memory network layers, as well as between the hidden layers and the output layer; the introduction of activation functions enables the neural network to learn and express the complex nonlinear mapping relationship between input features and ice thickness; the last hidden layer is connected to the output layer, which typically contains one or more neurons. In this embodiment, the output layer is a single neuron, and its output value represents the predicted ice thickness value of the grid region; The training of the deep neural network model includes: 1) collecting a large amount of historical working condition data to construct samples for each grid area at different historical moments. Each sample includes: time-series data of temperature, wind speed, and humidity as features, and the average icing thickness value as a label (obtained by simultaneous actual measurement in the grid area using a high-precision contact or optical thickness measuring device); 2) inputting the feature data of the training samples sequentially into the input layer of the deep neural network model. The input layer data is passed to the first hidden layer (convolutional layer), which performs convolution operations with the input data through convolution kernels and applies activation functions to extract primary local features; the output is passed to the subsequent hidden layer (long short-term memory network layer), which processes the time-series features (temperature, wind speed, and humidity time-series data) through its internal gating mechanism to obtain long-term dependencies; finally, the output value of the last hidden layer is passed to the neurons of the output layer, and after linear weighted summation and processing with possible activation functions, the predicted icing thickness value output by the model for the current input sample is obtained; 3 1) Using a preset loss function (e.g., mean squared error function), calculate the difference between the predicted ice thickness output by the deep neural network model and the actual label thickness value corresponding to the sample. The quantified result of this difference is the loss value. 2) Based on the loss value, calculate the partial derivative of the loss value with respect to each trainable weight parameter in the deep neural network model, i.e., the gradient, through the backpropagation algorithm. This process starts from the output layer and proceeds backward layer by layer along the network structure, using the chain rule to distribute the error signal to each weight parameter in each layer. Using the selected optimization algorithm (e.g., stochastic gradient descent or its variants, such as the Adam optimizer), based on the calculated gradient direction and magnitude of each weight parameter, combined with the learning rate hyperparameter, iteratively update all weight parameters in the model to adjust the model in the direction of minimizing the loss function. 3) Repeat steps 2) to 4) and use all training samples for multiple rounds of iteration until the prediction error of the model on the independent validation set converges to the preset range, thus completing the model training. In the real-time application phase, the standardized time-series data of each grid region constructed in the same way within the current control cycle are input into the input layer of the trained deep neural network model. Through forward propagation calculation within the model, the output layer finally outputs the initial predicted value of the ice thickness of each grid region. The data fusion and calibration process to generate the final distribution data includes: installing a small number of direct-measurement icing sensors at key locations on the blade (such as the leading edge of the blade root, the pressure surface in the middle of the blade, and the leading edge of the blade tip); reading the measured icing thickness data collected by the sensors in the current cycle as the calibration benchmark; using a spatial interpolation algorithm (such as Kriging interpolation), fusing the measured icing thickness data from sparse points with the grid prediction data covering the entire blade output by the deep neural network model; for grids with sensors installed, the measured data is used as the primary factor to correct the prediction values ​​of the deep neural network model; for grids without direct sensors, the prediction values ​​calibrated by the spatial correlation of neighboring measured points are used; finally, a digital icing thickness distribution map covering the entire blade surface is generated, in which each grid cell contains a reliable icing thickness value after data fusion.

[0020] It should be further explained that, in the specific implementation process, the pulse energy demand levels of each grid region are divided, and a regional energy density distribution matrix is ​​output, including: Read the ice thickness value for each grid region in the ice thickness distribution data; If the ice thickness value is less than or equal to the first thickness threshold, the grid area is classified as the first energy demand level, corresponding to the low ice thickness grid area. If the ice thickness value is greater than the first thickness threshold and less than or equal to the second thickness threshold, then the grid area is classified as the second energy demand level, corresponding to the medium ice thickness grid area. If the ice thickness value is greater than the second thickness threshold, the grid area is classified as the third energy demand level, corresponding to the high ice thickness grid area. The first and second thickness thresholds are used to classify the icing thickness into three levels, which distinguishes between low, medium and high icing thickness grid regions, thereby matching differentiated pulse energy levels to icing regions of different severity. The first energy demand level, the second energy demand level, and the third energy demand level constitute the pulse energy demand level; among them, the pulse energy required by the third energy level is higher than that of the second energy level, and the pulse energy required by the second energy level is higher than that of the first energy level. Each grid region is assigned a position weight coefficient, with the position weight coefficient of the blade root grid region being higher than that of the blade tip grid region. The position weight coefficient reflects the different requirements for de-icing energy at different locations of the blade due to differences in aerodynamic loads, structural stiffness, and ice accumulation characteristics. The weighted energy demand value for each grid region is obtained by multiplying its pulse energy demand level by its location weight coefficient. The pulse energy demand level can be quantified as a level value, for example, the first energy demand level = 1, the second energy demand level = 2, and the third energy demand level = 3. The weighted energy demand values ​​of all grid regions are arranged according to their spatial location to generate a regional energy density allocation matrix. Specifically, according to the spatial topology of the blade grid (i.e., an N-row M-column grid layout), the calculated weighted energy demand values ​​of all grid regions are arranged in order of their actual spatial location to form an N-row M-column two-dimensional matrix, which is the regional energy density allocation matrix. Each element in this matrix corresponds to a grid region, and its value directly determines the basic intensity of the electric pulse energy allocated to that grid region in subsequent steps.

[0021] It should be further explained that, in the specific implementation process, the calculation of the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region includes: Based on the regional energy density allocation matrix, the required basic pulse width and basic thyristor conduction angle for each grid region are determined. Specifically, the basic parameters are determined based on the regional energy density allocation matrix by reading the weighted energy demand value for each grid position in the regional energy density allocation matrix and converting the weighted energy demand value into the required basic pulse width (i.e., the duration of a single electrical pulse) and basic thyristor conduction angle (i.e., the electrical angle at which the thyristor begins to conduct within one power frequency cycle) for that grid region through a preset mapping relationship (e.g., linear mapping or lookup table method). Real-time wind speed and blade rotational speed data are acquired from the blade condition sensing dataset. A wind speed-rotational speed coupling influence factor is calculated, and this factor is used to correct the conduction angle of the basic thyristor to obtain the actual conduction angle. A wind speed-rotational speed coupling function is established with real-time wind speed and blade rotational speed as input variables. This function characterizes the combined effect of wind speed and blade rotational speed on the airflow field and icing adhesion on the blade surface. Real-time wind speed data and blade rotational speed data are input into the wind speed-rotational speed coupling function to calculate the wind speed-rotational speed coupling influence factor. The corrected actual conduction angle is obtained by multiplying the basic thyristor conduction angle by the wind speed-rotational speed coupling influence factor. The wind speed fluctuation rate is calculated based on real-time wind speed data. When the wind speed fluctuation rate exceeds a set fluctuation threshold, the base pulse width is dynamically adjusted based on the wind speed fluctuation attenuation factor to obtain the actual pulse width. The absolute value of the difference between the wind speed at the current sampling time and the wind speed at the previous sampling time is calculated to obtain the wind speed fluctuation rate. When the wind speed fluctuation rate exceeds the preset fluctuation threshold, the attenuation adjustment mechanism is activated to prevent overshoot or undershoot of de-icing energy due to sudden changes in wind speed. The fluctuation threshold is used to determine the severity of wind speed changes. The wind speed fluctuation attenuation factor is calculated based on the wind speed fluctuation rate. The calculation method of the wind speed fluctuation attenuation factor is: wind speed fluctuation attenuation factor = calibration value / (1 + gain coefficient × wind speed fluctuation rate), where the calibration value and gain coefficient are preset constants. The wind speed fluctuation attenuation factor is inversely proportional to the wind speed fluctuation rate. The base pulse width is multiplied by the wind speed fluctuation attenuation factor to obtain the smoothed actual pulse width. It is understandable that the actual conduction angle and the actual pulse width are the specific forms of the calculated thyristor conduction angle adjustment and pulse width modulation parameters.

[0022] It should be further explained that, in the specific implementation process, the thyristor undergoes low-temperature compensation correction to generate a trigger pulse sequence, including: The system acquires blade surface temperature data from the blade condition sensing dataset and determines whether it falls below a preset low-temperature threshold. If the current blade surface temperature is below the preset low-temperature threshold, a low-temperature compensation coefficient is calculated based on the temperature difference. This coefficient is then used to adjust the thyristor trigger pulse parameters, including voltage amplitude, leading-edge steepness, and pre-trigger time. The low-temperature threshold is a preset temperature boundary value used to determine whether the system has entered a low-temperature operating condition. In thyristor trigger control, leading-edge steepness refers to the rate at which the trigger pulse voltage rises from a low level to a high level, measured by the voltage rise amplitude per microsecond. This parameter directly affects the speed and reliability of the thyristor transitioning from the off state to the on state. Insufficient leading-edge steepness may cause thyristor conduction delay or even trigger failure under harsh conditions such as low temperatures. Specifically, the temperature difference between the blade surface temperature data and the preset low-temperature threshold is calculated; based on the temperature difference, a predefined compensation coefficient mapping table is built-in, which establishes the correspondence between the temperature difference range and the low-temperature compensation coefficient (e.g., temperature difference...). The corresponding low-temperature compensation coefficient is 1.1. The corresponding low-temperature compensation coefficient is 1.25. The above corresponds to a low-temperature compensation coefficient of 1.5). The corresponding low-temperature compensation coefficient is obtained from the compensation coefficient mapping table based on the current temperature difference using linear interpolation. The low-temperature compensation coefficient is used to adjust the trigger pulse parameters (voltage amplitude, leading-edge steepness, and pre-trigger time) of the thyristor trigger pulse, including: multiplying the low-temperature compensation coefficient by the default voltage amplitude of the thyristor trigger pulse to obtain the compensated voltage amplitude, ensuring sufficient driving capability for the semiconductor device at low temperatures; multiplying the low-temperature compensation coefficient by the default leading-edge steepness of the thyristor trigger pulse to obtain the compensated leading-edge steepness, thereby accelerating the thyristor conduction speed and overcoming the effect of decreased carrier mobility at low temperatures; and multiplying the low-temperature compensation coefficient by the default pre-trigger time of the thyristor trigger pulse to obtain the compensated pre-trigger time, compensating for potential delays in the entire drive circuit at low temperatures. Based on the actual conduction angle, actual pulse width, and trigger pulse parameters after low-temperature compensation, generate thyristor trigger pulse sequences corresponding to each grid region; Specifically, the actual conduction angle, actual pulse width, and trigger pulse parameters after low-temperature compensation correction of each grid region are integrated and calculated. For grid regions identified as having high icing thickness (corresponding to the third energy demand level), a multi-pulse train triggering mode is adopted: within the de-icing control cycle, a set (e.g., 3-5) of electrical pulses with the same parameters are generated based on the actual pulse width and actual conduction angle. The electrical pulses are alternately turned on at a relatively high first repetition frequency (e.g., 100 Hz) to form a pulse train. The multi-pulse train triggering mode can apply high energy density in a short time, which is beneficial for breaking up thick ice layers. It can be understood that the repetition frequency refers to the number of periodic electrical pulse events occurring within a unit time (usually 1 second); in the de-icing control scenario, it refers to the triggering frequency of electrical pulses applied to the same grid region of the blade. For grid regions identified as having low icing thickness (corresponding to the first energy demand level), a single-pulse triggering mode is adopted: within the de-icing control cycle, only one electrical pulse with the corresponding actual pulse width and actual conduction angle is generated, and the repetition frequency of the electrical pulse (i.e., the reciprocal of the control cycle) adopts a lower second repetition frequency (e.g., 10 Hz); the single-pulse triggering mode has lower energy consumption and is suitable for sustained de-icing; the first repetition frequency is higher than the second repetition frequency; finally, a set of trigger pulse sequence instructions containing timing and intensity information is generated for each grid region, ready to be sent to the drive circuit.

[0023] It should be further explained that, in the specific implementation process, the execution of electrical pulse de-icing control includes: The trigger pulse sequence is input to the thyristor drive circuit of the corresponding grid area to perform differentiated electrical pulse de-icing operations; among them, a multi-pulse triggering mode is used for areas with high ice thickness, and a single-pulse triggering mode is used for areas with low ice thickness. Specifically, the generated trigger pulse sequence is sent to the gate of the thyristor (or thyristor group) connected to the corresponding grid region through an isolation drive circuit; each grid region is associated with an independent or grouped electrical pulse generation unit, which includes an energy storage capacitor and a fast switch composed of thyristors; When the trigger pulse sequence arrives, the drive circuit generates a physical trigger pulse with corresponding characteristics based on the parameters (voltage amplitude, leading edge steepness) in the trigger pulse sequence and applies it to the gate of the thyristor. For high ice thickness regions set to multi-pulse trigger mode, the associated thyristor will rapidly and continuously turn on and off several times within a short time window according to the first repetition frequency specified by the trigger pulse sequence. Each turn-on will cause the energy storage capacitor to release a high-energy electrical pulse to the de-icing conductor (such as conductive coating or embedded electrode) attached to the blade in that region, generating instantaneous strong current and Joule heating, and may be accompanied by electromagnetic force effects, thereby causing multiple impacts on the ice layer. For low-ice-thickness areas set to single-pulse triggering mode, the associated thyristors are triggered to conduct only once in each longer control cycle, releasing an electrical pulse. The triggering sequence of all grid areas is coordinated by the central controller, and strategies such as sequential triggering, group triggering, or symmetrical triggering are programmed to avoid excessive instantaneous current surges on the grid side or generator end caused by the simultaneous operation of all thyristors, thus ensuring the electrical stability of the entire wind turbine generator set. It can be understood that the differentiated energy strategy calculated in the early stage is transformed into non-uniform de-icing action in the blade space through the hardware execution process. During the execution of step S6, the thyristor health status monitoring step is also performed simultaneously: Real-time monitoring of the on-state voltage drop and trigger response time of each thyristor; If the on-state voltage drop of any thyristor continuously exceeds the normal range or the trigger response time exceeds the set time tolerance, the thyristor is determined to be in an abnormal state. When a thyristor is determined to be in an abnormal state, the backup trigger channel of that grid area is activated, or de-icing compensation is performed by adjusting the regional energy density distribution matrix of adjacent grid areas. Specifically, each thyristor drive unit in the grid region is equipped with an online monitoring circuit. The monitoring targets two parameters: 1) the thyristor's on-state voltage drop. During each triggering and current flow period when the thyristor is turned on, the monitoring circuit samples the voltage between its anode and cathode. Under normal operating conditions and rated current, a healthy thyristor's on-state voltage drop should be stable within a low range (e.g., 1.5V to 2.0V); 2) the thyristor's trigger response time. The monitoring circuit records the time interval from the moment the trigger pulse arrives at the gate to the moment the thyristor's anode-cathode voltage drops below the holding voltage. This time interval is the trigger response time. Under normal circumstances, this time should be short and stable (e.g., less than 10 microseconds). Record the on-state voltage drop and response time for each action in real time; set the upper limit of the normal range for the on-state voltage drop (e.g., 2.5V); set the time tolerance for the trigger response time (e.g., 20 microseconds); if the monitoring finds that the on-state voltage drop of any thyristor exceeds the upper limit of the normal range multiple times (e.g., 5 consecutive times), or its trigger response time exceeds the time tolerance multiple times, then the thyristor is diagnosed as being in an abnormal state; the abnormality may be due to device aging, insufficient drive capability, poor heat dissipation, or connection failure; When a thyristor is determined to be abnormal, a fault-tolerant strategy is immediately activated. If the grid area is equipped with redundant backup trigger channels (such as parallel backup thyristors), the system automatically switches to the backup channel to take over the de-icing task. If there is no backup channel, compensation is made through software adjustments: when updating the regional energy density distribution matrix in the next control cycle, the energy demand level of adjacent grids (such as upwind, downwind, or spanwise adjacent grids) in the grid area to which the abnormal thyristor belongs is appropriately increased, and their pulse energy is increased. By enhancing the de-icing intensity of the surrounding grid areas, it is hoped that the resulting thermal or vibration effects can partially cover the fault area, thereby compensating to some extent for the decrease in local de-icing capacity caused by the failure of the thyristor at this point and ensuring the continuity of the overall de-icing effect. At the same time, the fault information is reported, prompting maintenance.

[0024] It should be further explained that, in the specific implementation process, the de-icing completion rate of each grid area is obtained, and the regional energy density allocation matrix is ​​dynamically updated based on the de-icing completion rate, including: Vibration spectrum data of the blades are collected by a vibration acceleration sensor, and temperature distribution data of the blade surface is collected by an infrared thermal imager. Analyze the blade vibration spectrum data and extract the amplitude of the characteristic frequency components caused by ice shedding; Analyze surface temperature distribution data and calculate the rate of temperature change in each grid region before and after the de-icing operation; The amplitude of the characteristic frequency component and the rate of temperature change are input into the completion evaluation function, and the de-icing completion quantification value of each grid region is output. If the quantification value of de-icing completion is lower than the preset completion threshold, the de-icing of the area is determined to be incomplete; the completion threshold is used as the minimum standard for evaluating the effectiveness of a single or phased de-icing operation. Specifically, during the execution of the electrical pulse de-icing operation or after the end of a control cycle, the de-icing effect sensing process is initiated: 1) The blade vibration signal generated during the de-icing process is collected by a high-frequency vibration acceleration sensor (sampling frequency not less than 1 kHz) pre-attached to the main structure of the blade (such as the beam cap or web); the collected time-domain vibration signal is subjected to a fast Fourier transform to obtain the blade vibration spectrum data; the blade vibration spectrum data is analyzed, and the characteristic frequency components excited by events such as ice layer breakage and shedding are extracted (these frequencies can be obtained through previous ice layer shedding tests, such as sidebands near a certain modal frequency); the amplitude of all characteristic frequency components is calculated as a vibration characteristic index A reflecting the intensity of the de-icing activity. 2) Scan the blade surface using an infrared thermal imager installed on the nacelle or tower; the infrared thermal imager acquires temperature distribution images of the blade surface before and after (or during) the de-icing operation; using image processing technology, register the temperature distribution images with the grid areas of the blade; for each grid area, calculate the difference between its average temperature over a period of time (e.g., 10 seconds) after the application of the electric pulse and the initial average temperature of the grid area before the application of the electric pulse, and then divide by time to obtain the temperature change rate B of the area; where the temperature change rate indirectly reflects the Joule heating effect generated by the electric pulse and the change in surface thermal resistance after the ice layer falls off;

[0025] The vibration characteristic index A and temperature change rate B corresponding to each grid region are input into a predefined completion evaluation function. This function is a weighted summation form, for example: the quantified value of de-icing completion Q = α × (A / A0) + β × (B / B0), where α and β are weighting coefficients (α + β = 1), and A0 and B0 are the normalized reference values ​​of vibration characteristic index A and temperature change rate B, respectively. The completion evaluation function outputs a quantized value Q between 0 and 1, which is used to characterize the effect of this de-icing operation in the region. The higher the value, the more thorough the de-icing. A preset completion threshold is set (e.g., 0.7). If the Q value of a grid region is lower than the completion threshold, it is determined that the de-icing in that grid region is incomplete, and energy input needs to be increased in subsequent control. For grid areas that are determined to be incompletely de-iced, the pulse energy demand level corresponding to the grid area is increased by one level in the current regional energy density allocation matrix. That is, if it was originally the first energy demand level, it will be increased to the second energy demand level; if it was originally the second energy demand level, it will be increased to the third energy demand level; and if it was originally the third energy demand level, it will remain unchanged. The purpose is to increase the subsequent pulse energy for grid areas with poor de-icing effect. Set an efficiency saturation threshold; whereby the efficiency saturation threshold is used to define the critical point at which energy input may become redundant after the de-icing effect has reached its optimal level. If the quantified value of de-icing completion remains above the efficiency saturation threshold for several consecutive control cycles (e.g., 3 cycles), it indicates that the currently allocated energy may have exceeded the optimal de-icing requirement, resulting in redundancy. In the regional energy density allocation matrix, the pulse energy demand level corresponding to this grid area is reduced by one level (e.g., the third energy demand level is reduced to the second energy demand level), or de-icing energy is temporarily not allocated (marked as not requiring active de-icing), and only monitoring is performed. Using the updated regional energy density allocation matrix as input to step S4 in the next control cycle, the base pulse width and base thyristor conduction angle of each grid region are redefined. Example 1:

[0026] To verify the effectiveness and superiority of the thyristor control method for the electric pulse de-icing device for wind turbine blades proposed in this invention in a real complex environment, this method was applied to a 1.5 MW wind turbine unit in a mountainous wind farm in northern my country. The wind farm has a harsh winter climate with wet snow, freezing rain, and supercooled water mist all year round, resulting in prominent blade icing problems that seriously restrict power generation efficiency and operational safety. The electric pulse de-icing devices previously used for this unit, which were based on timed triggering or simple temperature difference threshold control, had problems such as uneven de-icing, high energy consumption, and unstable performance under drastic temperature changes or strong wind conditions. In this embodiment, the method described above is used to control the electric pulse de-icing device of the unit. Based on the aerodynamic shape of the blade, a single blade is divided into 8 segments along the spanwise direction and 4 segments along the chordwise direction, forming a total of 32 independent and controllable rectangular grid areas. An independent thyristor drive unit is configured for each area. Direct icing sensors are installed at key locations on the blade (blade root leading edge, blade mid-suction surface, and blade tip leading edge). At the same time, the anemometer, speed encoder in the nacelle, and temperature sensor network distributed on the blade surface are integrated to collect data in real time and fuse them into a blade operating condition sensing dataset. In a typical winter mixed icing event (ambient temperature) to (With intermittent wet snow and fluctuating wind speeds of 4-12 m / s), the control process of this method is fully executed, including: control cycle initiation, ice thickness-region correlation model (predictive fusion model based on deep neural network) starting to work; the model receives real-time sensing data and outputs ice thickness distribution data covering 32 grids. The results show that the ice thickness is thicker in the leaf root and the middle of the leading edge region (maximum predicted thickness 4.2 mm), while it is thinner in the leaf tip and trailing edge region (about 0.5-1 mm). Based on icing thickness distribution data, the grid is automatically divided into three energy demand levels, and a regional energy density allocation matrix is ​​generated by combining the principle of higher weighting for the blade root region. Combining real-time wind speed (8 m / s, fluctuation ±2 m / s) and blade rotation speed (e.g., 12 revolutions per minute), the basic conduction angle is corrected through a wind speed-rotation speed coupling influence function, and the basic pulse width is smoothed through a wind speed fluctuation attenuation factor. Simultaneously, considering the blade surface temperature (e.g., ... ) below the preset low temperature threshold (e.g. It automatically activates the low-temperature compensation algorithm, which increases the voltage amplitude and leading-edge steepness of the trigger pulse by approximately 30%. The generated trigger pulse sequence is input to the thyristor drive circuit of the corresponding grid area; for the 5 grids with ice thickness exceeding 3 mm (mainly located at the leading edge of the blade root), a high-frequency multi-pulse train mode is used for concentrated energy impact; for the 10 grids with moderate ice, a single pulse mode with medium parameters is used; for the remaining grids with light ice, a low-frequency, low-energy sustaining pulse is applied; the entire de-icing process lasts about 90 seconds. Feedback on the de-icing process was obtained through deployed blade vibration sensors and infrared thermal imagers. Data analysis showed that within 20 seconds of the start of the de-icing operation, the characteristic vibration spectrum energy of the grid area with high ice thickness increased, indicating that the ice layer began to crack and fall off. After de-icing, the infrared image showed that the temperature distribution on the blade surface tended to be uniform, and the temperature of the original low-temperature ice area rose significantly. According to the evaluation function calculation, after de-icing was completed, the quantified value of de-icing completion for all grids exceeded 0.85, meeting the threshold requirement. Based on the de-icing completion rate, the regional energy density distribution matrix was dynamically updated to adjust the trigger parameters of subsequent control cycles. In a two-week comparative test, the de-icing device using this control method was operated alternately with the traditional timed control method. Statistical data showed that under similar meteorological conditions, the control method of this invention reduced the average energy consumption of a single effective de-icing by about 35%, shortened the time for the power generation to recover to the rated value after de-icing by about 40%, and completely avoided the local ice residue or overheating phenomenon sometimes seen in traditional methods. The feedback optimization mechanism began to take effect after several days of operation, further improving the response speed and energy distribution accuracy to recurring icing patterns. This embodiment demonstrates that the control method proposed in this invention, by integrating multi-source sensing, neural network prediction, zoned differentiated energy allocation, multi-condition adaptive compensation, and closed-loop feedback optimization, achieves precise and adaptive control of the thyristor of the electric pulse de-icing device for wind turbine blades, effectively improving the reliability, economy, and safety of wind turbine de-icing in complex real-world environments.

[0027] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. The focus of each embodiment is on its differences from other embodiments. In particular, the apparatus embodiments are described simply because they are fundamentally based on the method embodiments; relevant details can be found in the descriptions of the method embodiments.

[0028] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0029] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0030] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0031] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0032] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0033] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0034] In conclusion, the above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A thyristor control method for an electrical pulse de-icing device for wind turbine blades, characterized in that: S1. Acquire icing distribution data, real-time wind speed data, blade rotation speed data, and blade surface temperature data on the surface of wind turbine blades, and integrate them into a blade operating condition sensing dataset. S2. Based on the blade condition sensing dataset, ice thickness distribution data is obtained through the ice thickness-region correlation model. S3. Based on the ice thickness distribution data, classify the pulse energy demand level of each grid region and output the regional energy density distribution matrix; S4. Based on the regional energy density distribution matrix, real-time wind speed data, and blade rotation speed data, calculate the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region; wherein, the calculation of the thyristor conduction angle adjustment and pulse width modulation parameters corresponding to each grid region includes: Based on the regional energy density distribution matrix, the required basic pulse width and basic thyristor conduction angle for each grid region are determined; real-time wind speed data and blade rotation speed data are acquired from the blade condition sensing dataset, the wind speed-rotation speed coupling influence factor is calculated, and the basic thyristor conduction angle is corrected using this factor to obtain the actual conduction angle; the wind speed fluctuation rate is calculated based on the real-time wind speed data, and when the wind speed fluctuation rate exceeds the set fluctuation threshold, the basic pulse width is dynamically adjusted based on the wind speed fluctuation attenuation factor to obtain the actual pulse width; S5. Based on the blade surface temperature data, perform low-temperature compensation correction on the thyristor to generate a trigger pulse sequence; wherein, performing low-temperature compensation correction on the thyristor to generate a trigger pulse sequence includes: The blade surface temperature data in the blade condition sensing dataset is acquired to determine whether it is below a preset low temperature threshold. If the current blade surface temperature is below the preset low temperature threshold, a low temperature compensation coefficient is calculated based on the temperature difference, and the thyristor trigger pulse parameters, including voltage amplitude, leading edge steepness, and pre-trigger time, are adjusted using this coefficient. Based on the actual conduction angle, actual pulse width, and trigger pulse parameters after low temperature compensation, a thyristor trigger pulse sequence corresponding to each grid region is generated. S6. Input the trigger pulse sequence to the thyristor drive circuit of the corresponding grid area to perform electrical pulse de-icing control; S7. During the de-icing process, the de-icing completion rate of each grid area is obtained in real time, and the regional energy density allocation matrix is ​​dynamically updated based on the de-icing completion rate.

2. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, Obtain ice thickness distribution data, including: The blade is divided into several grid regions along its spanwise and tangential directions; Based on historical monitoring data and blade condition sensing dataset, temperature time series data, wind speed time series data, and humidity time series data for each grid area are constructed. Temperature time-series data, wind speed time-series data, and humidity time-series data for each grid region are used as training samples, and the actual measured icing thickness is used as the label to train a deep neural network model; wherein, the deep neural network model constitutes an icing thickness-region correlation model. The temperature time series data, wind speed time series data, and humidity time series data of each grid area in the current cycle's blade condition sensing dataset are input into a pre-trained deep neural network model, which outputs the predicted icing thickness value for each grid area. By combining the predicted icing thickness with the calibration data collected by the real-time icing detection sensor, the icing thickness distribution data on the blade surface is obtained.

3. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, The pulse energy demand levels for each grid region are divided, including: Read the ice thickness value for each grid region in the ice thickness distribution data; If the ice thickness is less than or equal to the first thickness threshold, the grid area is classified as the first energy demand level. If the ice thickness is greater than the first thickness threshold and less than or equal to the second thickness threshold, then the grid area is classified as the second energy demand level. If the ice thickness is greater than the second thickness threshold, the grid area is classified as the third energy demand level. The first energy demand level, the second energy demand level, and the third energy demand level constitute the pulse energy demand level.

4. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, Output the regional energy density distribution matrix, including: Assign a location weight coefficient to each grid region; Multiply the pulse energy demand level of each grid region by its location weight coefficient to obtain the weighted energy demand value of that grid region; The weighted energy demand values ​​of all grid regions are arranged according to spatial location to generate a regional energy density distribution matrix.

5. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, Performing electrical pulse de-icing control includes: The trigger pulse sequence is input to the thyristor drive circuit of the corresponding grid area to perform differentiated electrical pulse de-icing operations; among them, a multi-pulse triggering mode is used for areas with high ice thickness, and a single-pulse triggering mode is used for areas with low ice thickness.

6. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, Obtain the de-icing completion rate for each grid area, including: Vibration spectrum data of the blades are collected by a vibration acceleration sensor, and temperature distribution data of the blade surface is collected by an infrared thermal imager. Analyze the blade vibration spectrum data and extract the amplitude of the characteristic frequency components caused by ice shedding; Analyze surface temperature distribution data and calculate the rate of temperature change in each grid region before and after the de-icing operation; The amplitude of the characteristic frequency component and the rate of temperature change are input into the completion evaluation function, and the de-icing completion quantification value of each grid region is output. If the quantification value of de-icing completion is lower than the preset completion threshold, the de-icing of that area is determined to be incomplete.

7. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 6, characterized in that, Update the regional energy density distribution matrix, including: For grid areas that are determined to be incompletely de-iced, the pulse energy demand level corresponding to that grid area will be increased by one level in the current regional energy density allocation matrix; Set an efficiency saturation threshold; If the quantified value of de-icing completion remains higher than the efficiency saturation threshold for several consecutive control cycles, then in the regional energy density allocation matrix, the pulse energy demand level corresponding to that grid area will be reduced by one level, or de-icing energy will not be allocated temporarily. Using the updated regional energy density allocation matrix as input to step S4 in the next control cycle, the base pulse width and base thyristor conduction angle of each grid region are redefined.

8. The thyristor control method for an electrical pulse de-icing device for wind turbine blades according to claim 1, characterized in that, During the execution of step S6, the thyristor health status monitoring step is also performed simultaneously: Real-time monitoring of the on-state voltage drop and trigger response time of each thyristor; If the on-state voltage drop of any thyristor continuously exceeds the normal range or the trigger response time exceeds the set time tolerance, the thyristor is determined to be in an abnormal state. When a thyristor is determined to be in an abnormal state, the backup trigger channel for that grid area is activated, or de-icing compensation is performed by adjusting the regional energy density distribution matrix of adjacent grid areas.

Images