The Method For Evaluating Wind Turbine Structure State Based On Optical Fiber Sensing Technology

As the most important way for humans to use wind energy,the safety and stability of wind turbine is very important.As we known,most wind farms are located in desert,mountainous areas,pastures,tidal flats or offshore areas.Wind turbines bear complex and diverse loads in harsh natural environments.These factors will inevitably cause fatigue damage or sudden damage to wind turbines.The traditional wind turbine structural damage identification methods most use manual observation,resistance sensing to detect wind turbine damage.Such methods have low observation efficiency,which cannot meet the monitoring requirements of high-power wind turbines at this time.Based on the highperformance of fiber-optic sensing technology,this paper makes full use of the characteristics of distributed high-density measurement and quasi-distributed dynamic measurement to obtain the true strain response of the wind turbine.Research on a new method for structural monitoring and evaluation of wind turbines.The main contents are as follows:A new method for monitoring the full-length strain response of offshore wind turbine pile foundations based on DPP-BOTDA was proposed.This method was applied to a 5MW offshore wind turbine steel pipe pile bearing capacity test,the distributed strain value of the pile along the elevation was obtained,and the ultimate bearing capacity of this pile was determined by the load-strain maximum curve.The average absolute percentage error between the calculated and the measured displacement value of the pile top is only0.03548.The ratio coefficient between the pile top displacement and the load is 0.9136mm/k N.An ice-covering estimation method based on measuring the surface temperature of wind turbine blades was proposed.A 5 k W wind turbine blade model was tested in a low temperature laboratory based on DPP-BOTDA distributed fiber sensing technology.The start time,duration,and end time of icing can be determined from the distributed temperature data.Finally,the duration of the temperature stabilization stage is used as a direct indicator of ice thickness estimation,and the characteristic relationship between ice thickness and duration under different operating conditions is obtained.A method for identifying the ice coating of large wind turbine blades based on quasidistributed fiber grating sensor data was proposed.A monitoring system based on fiber grating sensing was designed and installed on a large 1.5 MW test wind turbine.The basic operating characteristic information of the test wind turbine was obtained.Combined with the actual operation information of the blades during the winter season,the blade icing was finally judged through the strain ratios of different months of the blade's waving and shimmy direction.A wind turbine blade strain prediction method based on LSTM neural network was proposed.Based on the measured data set of 1.5 MW large wind turbine blades,the strain prediction neural network was established with wind speed,wind turbine rotation speed,power generation,yaw position,wind angle,and ambient temperature as parameters.The measured data was used to train this network,and the effectiveness of the model was verified by comparing the predicted strain and the measured strain,and the blade service state was evaluated by the ratio of the predicted strain to the measured strain.