A wind turbine blade partition ice melting system and method based on multi-modal perception

Abstract

The application belongs to the technical field of wind power generation equipment safety, and particularly relates to a wind turbine blade partition ice melting system and method based on multi-modal perception, which comprises a multi-modal perception module, an intelligent control module and an execution and energy saving module; the multi-modal perception module comprises a plurality of infrared array sensors arranged in an equidistant array along a blade leading edge, and a vibration accelerometer installed at a blade root, which are used for monitoring blade surface temperature distribution and vibration characteristics in real time; the intelligent control module comprises an edge computing unit integrating an AI algorithm based on a long short-term memory neural network (LSTM), and a partition control strategy, which are used for realizing real-time prediction of ice layer thickness distribution and partition control according to data collected by the multi-modal perception module. The application can effectively solve the problems of high energy consumption and high misjudgment rate of traditional ice melting technology, has the advantages of high detection precision, fast response speed, long service life and the like, and is suitable for anti-icing protection of various large wind turbine generators in a high-humidity low-temperature environment.

Description

Technical Field

[0001] This invention belongs to the field of wind power equipment safety technology, specifically relating to a wind turbine blade zoned de-icing system and method based on multimodal perception. Background Technology

[0002] Against the backdrop of the global energy structure accelerating its transformation towards cleaner and lower-carbon energy, wind power, with its renewable and pollution-free characteristics, has become an important energy pillar for achieving the "dual carbon" goal. Statistics show that in 2023, the global installed capacity of wind power reached 140GW, bringing the cumulative installed capacity to over 1,000GW. Among these, horizontal axis wind turbines, due to their high aerodynamic efficiency and mature technology, account for over 98% of the market and are widely used in wind-rich areas such as coastal regions and high-altitude areas.

[0003] Problems with existing technology:

[0004] However, in low-temperature and high-humidity environments (such as winters in Northeast and Northwest my country, where ambient temperatures often fall below -10°C and humidity exceeds 85%), wind turbine blades are highly susceptible to icing. Ice buildup causes aerodynamic distortion of the blades, reducing aerodynamic efficiency by 30%-50% and resulting in an average annual power generation loss of 15%-25%. Simultaneously, the added mass of the ice layer (approximately 10 kg per square meter of 1 cm thick ice layer) exacerbates blade fatigue loads. Industry statistics indicate that blade fractures caused by icing account for 42% of wind turbine structural failures, seriously threatening the stable operation of the power grid.

[0005] Currently, wind turbine icing detection technology has significant shortcomings: traditional detection devices rely on temperature changes in the natural environment to simulate icing conditions, failing to reproduce complex icing forms such as rime and hoarfrost; laboratory simulations often use single-variable control, making it difficult to replicate the icing process under the coupled effects of temperature, humidity, and wind speed in actual operation. These limitations prevent accurate understanding of the power characteristics of wind turbines under different icing conditions, making it difficult to adapt unit control strategies to complex operating conditions, resulting in energy waste and equipment safety hazards. The deficiencies of existing technology urgently necessitate innovative detection methods to achieve accurate identification and efficient response to icing conditions. Summary of the Invention

[0006] The purpose of this invention is to provide a multimodal sensing-based wind turbine blade zonal de-icing system and method, which can effectively solve the problems of high energy consumption and high false judgment rate of traditional de-icing technology. It has the advantages of high detection accuracy, fast response speed and long service life, and is suitable for anti-icing protection of various large wind turbine generators in high humidity and low temperature environments.

[0007] The specific technical solution adopted by this invention is as follows:

[0008] A wind turbine blade zonal de-icing system based on multimodal sensing includes a multimodal sensing module, an intelligent control module, and an execution and energy-saving module;

[0009] The multimodal sensing module includes multiple infrared array sensors arranged at equal intervals along the leading edge of the blade, and a vibration accelerometer installed at the root of the blade, for real-time monitoring of the temperature distribution and vibration characteristics of the blade surface.

[0010] The intelligent control module includes an edge computing unit that integrates an AI algorithm based on a long short-term memory neural network (LSTM), and a partition control strategy, which is used to realize real-time prediction and partition control of ice thickness distribution based on data collected by the multimodal perception module.

[0011] The execution and energy-saving module includes shape memory alloy wires for fixing carbon fiber heating strips to the surface of the blades, and an energy management system integrating a supercapacitor energy storage device for performing zoned ice melting operations and storing energy through off-peak electricity.

[0012] The interval between two adjacent infrared array sensors is 10 cm.

[0013] Its characteristic is that the edge computing unit takes infrared temperature field data and vibration signal as input and predicts the ice layer thickness distribution through LSTM model.

[0014] The zoning control strategy is to automatically trigger the carbon fiber heating belt in the corresponding area when the system detects that the ice layer thickness is ≥2mm.

[0015] The carbon fiber heating belt has a power density of 500W / m², a response time of <30s, and adopts a gradient heating mode with power distribution decreasing from the center to the edge. The temperature of the central heating area is set to 5℃, and the temperature of the edge area is set to 3℃.

[0016] The phase transition temperature of the shape memory alloy wire is 5°C. When the temperature exceeds 5°C, it automatically elongates to release the mechanical constraint of the heating band.

[0017] The energy management system is configured to store energy during periods of low electricity prices on the grid at night and use the stored energy to power the ice-melting operation during the day.

[0018] The infrared array sensors are arranged along the leading edge of the blade to form a high-density temperature monitoring network, which is used to acquire real-time temperature distribution data on the blade surface; the vibration accelerometer identifies abnormal aerodynamic performance of the blade by detecting changes in vibration characteristics caused by the added mass of the ice layer.

[0019] In the execution and energy-saving module, the carbon fiber heating belt is dynamically connected to the blade through shape memory alloy wire.

[0020] A multimodal sensing-based method for zonal de-icing of wind turbine blades includes the following steps:

[0021] S1, Start detection. After system initialization, determine whether the ambient temperature is below 0℃ and the humidity is >85%. If the conditions are not met, enter sleep mode. If the conditions are met, start multi-sensor data acquisition.

[0022] S2, Data processing: Perform Kalman filtering noise reduction on the data collected by the infrared sensor and vibration sensor, and extract the characteristic parameters of the temperature anomaly region ΔT < -2℃ and the vibration frequency offset Δf > 5Hz.

[0023] S3, Ice layer prediction: The processed data is input into an AI model based on a long short-term memory neural network (LSTM) to predict the ice layer thickness distribution. When the predicted thickness is ≥2mm, a zone heating command is triggered; otherwise, the process returns to the data acquisition stage.

[0024] S4, heating control, triggers the corresponding area of ​​the carbon fiber heating belt to start gradient heating, monitors data in real time, if the ice layer thickness is <1mm, the heating belt is turned off, otherwise the heating power is dynamically adjusted, and data acquisition and ice layer prediction are performed cyclically.

[0025] The technical effects achieved by this invention are as follows:

[0026] This invention integrates an infrared array sensor and a vibration accelerometer to monitor the surface temperature distribution and vibration characteristics of blades in real time, and uses an LSTM neural network algorithm to accurately predict the ice layer thickness.

[0027] This invention employs a zoned heating strategy, activating the carbon fiber heating belt only in the icing area, and uses shape memory alloy wires to achieve dynamic connection of the heating belt.

[0028] This invention effectively solves the problems of high energy consumption and high false alarm rate of traditional de-icing technology. It has the advantages of high detection accuracy, fast response speed and long service life. It is suitable for anti-icing protection of various large wind turbine generators in high humidity and low temperature environments. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall system architecture of the present invention.

[0030] Figure 2 This is a flowchart of the AI ​​model training process of the present invention. Detailed Implementation

[0031] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0032] like Figure 1 As shown, a multimodal sensing-based wind turbine blade zonal de-icing system includes a multimodal sensing module, an intelligent control module, and an execution and energy-saving module.

[0033] The multimodal sensing module includes multiple infrared array sensors arranged equidistantly along the leading edge of the blade, with a spacing of 10 cm between adjacent infrared array sensors, and a vibration accelerometer installed at the root of the blade for real-time monitoring of the blade surface temperature distribution and vibration characteristics. The infrared array sensors are arranged along the leading edge of the blade to form a high-density temperature monitoring network for real-time acquisition of blade surface temperature distribution data. The infrared array sensors use the MLX90640 model and have a temperature resolution of 0.1℃ and a measurement accuracy of ±0.5℃. The vibration accelerometer identifies abnormal aerodynamic performance of the blade by detecting changes in vibration characteristics caused by the added mass of ice. The vibration accelerometer uses the ADXL345 model sensor, and the monitoring range covers vibration frequency and amplitude changes from 0-200Hz.

[0034] The intelligent control module includes an edge computing unit integrating an AI algorithm based on a Long Short-Term Memory (LSTM) neural network. This edge computing unit takes infrared temperature field data and vibration signals as input, uses an LSTM model to predict ice thickness distribution, and implements a zoned control strategy. This strategy enables real-time prediction and zoned control of ice thickness distribution based on data collected by the multimodal sensing module. The zoned control strategy automatically triggers the corresponding carbon fiber heating belt when the system detects an ice thickness ≥2mm. The carbon fiber heating belt has a power density of 500W / m² and a response time <30s, achieving rapid heating. A gradient heating mode with decreasing power distribution from the center to the edge is adopted, with the temperature set at 5℃ in the central heating area and 3℃ in the edge area, effectively reducing damage to the blade material caused by thermal stress due to the temperature gradient.

[0035] The execution and energy-saving module includes shape memory alloy wires for fixing carbon fiber heating strips to the blade surface. The carbon fiber heating strips are dynamically connected to the blades through the shape memory alloy wires. The phase change temperature of the shape memory alloy wires is 5°C. When the temperature exceeds 5°C, the wires automatically elongate to release the mechanical constraints of the heating strips, avoiding material aging caused by long-term pressure. The module also includes an energy management system with an integrated supercapacitor energy storage device for performing zoned de-icing operations and storing energy during off-peak hours. The energy management system is configured to store energy during low-price periods on the power grid at night and use the stored energy to power the de-icing operations during the day, achieving efficient energy utilization.

[0036] Based on the above structure, a zoned heating strategy is adopted, which only heats the icing area, making it more energy-efficient than the traditional whole-blade heating method. Multimodal data fusion technology effectively reduces the icing misjudgment rate and significantly improves system reliability. The heating belt is dynamically connected through shape memory alloy wires, avoiding damage to the blade coating caused by long-term mechanical stress and improving the service life of key system components. The entire process from detecting icing to completing de-icing is relatively short, meeting the real-time protection requirements of wind turbine generators.

[0037] like Figure 2 As shown, a multimodal sensing-based method for zonal de-icing of wind turbine blades includes the following steps:

[0038] S1, Start detection. After system initialization, determine whether the ambient temperature is below 0℃ and the humidity is >85%. If the conditions are not met, enter sleep mode. If the conditions are met, start multi-sensor data acquisition.

[0039] S2, Data processing: Perform Kalman filtering noise reduction on the data collected by the infrared sensor and vibration sensor, and extract the characteristic parameters of the temperature anomaly region ΔT < -2℃ and the vibration frequency offset Δf > 5Hz.

[0040] S3, Ice layer prediction: The processed data is input into an AI model based on a long short-term memory neural network (LSTM) to predict the ice layer thickness distribution. When the predicted thickness is ≥2mm, a zone heating command is triggered; otherwise, the process returns to the data acquisition stage.

[0041] S4, heating control, triggers the corresponding area of ​​the carbon fiber heating belt to start gradient heating, monitors data in real time, if the ice layer thickness is <1mm, the heating belt is turned off, otherwise the heating power is dynamically adjusted, and data acquisition and ice layer prediction are performed cyclically.

[0042] Example 1

[0043] Experimental conditions: ambient temperature -5℃, humidity 88%, wind speed 10m / s, with some areas of the blades showing signs of icing.

[0044] System working process:

[0045] Upon system startup and initialization, the system checks whether the ambient temperature is below 0°C and the humidity is above 85%. If the conditions are not met, the system enters sleep mode. If the conditions are met, the system starts multi-sensor data acquisition.

[0046] Infrared array sensors monitor the temperature distribution on the blade surface in real time and detect an abnormal temperature in the 30-50cm area at the leading edge of the blade, ΔT < -2℃; vibration accelerometers detect a vibration frequency offset Δf = 6Hz, indicating that there is a possibility of icing in this area.

[0047] The edge computing unit inputs temperature and vibration data into an AI model based on a long short-term memory neural network (LSTM), predicts that the ice thickness in the area is 2.5 mm, and triggers a zoned heating command.

[0048] The carbon fiber heating belt in the corresponding area is activated, using a gradient heating mode with a center temperature of 5°C and an edge temperature of 3°C. The shape memory alloy wire elongates when the temperature exceeds 5°C, releasing the mechanical constraint. The energy management system uses nighttime energy storage to power the heating belt.

[0049] Ice melting effect: It took 12 minutes from the detection of ice formation to the completion of ice melting. The heating belt was turned off when the ice thickness dropped to 0.8mm. Compared with traditional full-blade heating, this ice melting process saved 43.2% energy. No signs of coating damage were found on the key components of the system.

[0050] Example 2

[0051] Experimental conditions: ambient temperature -10℃, humidity 90%, wind speed 15m / s, large area of ​​ice formation on the leading edge of the blades.

[0052] System working process:

[0053] Upon system startup and initialization, the system checks whether the ambient temperature is below 0°C and the humidity is above 85%. If the conditions are not met, the system enters sleep mode. If the conditions are met, the system starts multi-sensor data acquisition.

[0054] Infrared array sensors showed that the temperature in the 0-80cm area at the leading edge of the blade was generally below -2℃, and vibration accelerometers detected a vibration frequency shift of Δf=8Hz, confirming large-area icing.

[0055] An AI model based on a long short-term memory neural network (LSTM) predicts an ice layer thickness of up to 3 mm, triggering the simultaneous activation of heating zones in multiple regions, employing a gradient heating mode.

[0056] Multi-zone carbon fiber heating belts operate synchronously with a power density of 500W / m² and a response time of 25s; shape memory alloy wires dynamically adjust constraints; and a supercapacitor energy storage device provides continuous power.

[0057] Ice melting effect: The entire process takes 15 minutes, the ice layer is completely melted, and the heating belt is turned off when the thickness is less than 1mm; the test results show that the false judgment rate is 4.5%, which is 23.5% lower than that of traditional single sensors, and the lifespan of key system components is expected to be extended by 32%.

[0058] Example 3

[0059] Experimental settings: ambient temperature -3℃, humidity 95%, wind speed 20m / s, thin ice appeared on the blade surface and continued to condense.

[0060] System working process:

[0061] Upon system startup and initialization, the system checks whether the ambient temperature is below 0°C and the humidity is above 85%. If the conditions are not met, the system enters sleep mode. If the conditions are met, the system starts multi-sensor data acquisition.

[0062] Infrared sensors monitored the dynamic changes in abnormal temperature areas on the blade surface in real time, and vibration accelerometers detected frequent shifts in vibration frequency, Δf > 5Hz.

[0063] An AI model based on a long short-term memory neural network (LSTM) predicts that the ice thickness fluctuates between 2 and 2.5 mm based on dynamic data, continuously triggering the heating zone in the corresponding area and dynamically adjusting the heating power.

[0064] The heating element adjusts its power based on real-time data to avoid overheating; the shape memory alloy wire releases its constraint in real time as the temperature changes; and the energy management system optimizes energy storage distribution to ensure continuous power supply.

[0065] De-icing effect: It takes 13 minutes from the detection of ice formation to the completion of de-icing, effectively preventing the ice layer from continuing to condense; energy consumption is reduced by 41.8% compared with traditional methods. The reliability of the system has been verified in high humidity and high wind speed environments, and no damage to the blade material caused by temperature gradient has occurred.

[0066] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A wind turbine blade zonal de-icing system based on multimodal sensing, characterized in that... It includes a multimodal sensing module, an intelligent control module, and an execution and energy-saving module; The multimodal sensing module includes multiple infrared array sensors arranged at equal intervals along the leading edge of the blade, and a vibration accelerometer installed at the root of the blade, for real-time monitoring of the temperature distribution and vibration characteristics of the blade surface. The intelligent control module includes an edge computing unit that integrates an AI algorithm based on a long short-term memory neural network (LSTM), and a partition control strategy, which is used to realize real-time prediction and partition control of ice thickness distribution based on data collected by the multimodal perception module. The execution and energy-saving module includes shape memory alloy wires for fixing carbon fiber heating strips to the surface of the blades, and an energy management system integrating a supercapacitor energy storage device for performing zoned ice melting operations and storing energy through off-peak electricity. The zoning control strategy is to automatically trigger the carbon fiber heating belt in the corresponding area when the system detects that the ice layer thickness is ≥2mm. The carbon fiber heating belt has a power density of 500W / m², a response time of <30s, and adopts a gradient heating mode with power decreasing from the center to the edge. The temperature of the central heating area is set to 5℃, and the temperature of the edge area is set to 3℃. The phase transition temperature of the shape memory alloy wire is 5°C. When the temperature exceeds 5°C, it automatically elongates to release the mechanical constraint of the heating band. The infrared array sensors are arranged along the leading edge of the blade to form a high-density temperature monitoring network, which is used to acquire real-time temperature distribution data on the blade surface; the vibration accelerometer identifies abnormal aerodynamic performance of the blade by detecting changes in vibration characteristics caused by the added mass of the ice layer.

2. The wind turbine blade zonal de-icing system based on multimodal sensing according to claim 1, characterized in that: The interval between two adjacent infrared array sensors is 10 cm.

3. The wind turbine blade zonal de-icing system based on multimodal sensing according to claim 1, characterized in that: The edge computing unit takes infrared temperature field data and vibration signals as input and uses an LSTM model to predict the ice layer thickness distribution.

4. A multimodal sensing-based zoned de-icing system for wind turbine blades according to claim 1, characterized in that: The energy management system is configured to store energy during periods of low electricity prices on the grid at night and use the stored energy to power the ice-melting operation during the day.

5. A multimodal sensing-based zoned de-icing system for wind turbine blades according to claim 1, characterized in that: In the execution and energy-saving module, the carbon fiber heating belt is dynamically connected to the blade through shape memory alloy wire.

6. A method for zonal de-icing of wind turbine blades based on multimodal sensing, based on the zonal de-icing system for wind turbine blades based on multimodal sensing according to any one of claims 1-5, characterized in that, Includes the following steps: S1, Start detection. After system initialization, determine whether the ambient temperature is below 0℃ and the humidity is >85%. If the conditions are not met, enter sleep mode. If the conditions are met, start multi-sensor data acquisition. S2, Data processing: Perform Kalman filtering noise reduction on the data collected by the infrared sensor and vibration sensor, and extract the feature parameters of the temperature anomaly region ΔT < -2℃ and the vibration frequency offset Δf > 5Hz. S3, Ice layer prediction: Input the processed data into an AI model based on a long short-term memory neural network (LSTM) to predict the ice layer thickness distribution. When the predicted thickness is ≥2mm, trigger the zone heating command; otherwise, return to the data acquisition stage. S4, heating control, triggers the corresponding area of ​​the carbon fiber heating belt to start gradient heating, monitors data in real time, if the ice layer thickness is <1mm, the heating belt is turned off, otherwise the heating power is dynamically adjusted, and data acquisition and ice layer prediction are performed cyclically.

Images