The Research On Methods Of Soft Sensor, Prediction And Numerical Simulation For Wind Speed In Wind-farm

Wind power technology is more and more in depth study.The research of wind speed in wind farm is an important part of wind power genaration's study.Based on results of published papers,the main content of this thesis is as follow:1.Analysis of wind speed features is done,and three questions about wind speed in wind farm are epurated aimed at wind power generation technology.The three questions are:wind speed soft sonsor for wind turbin controls,wind speed prediction in wind farm and short term wind speed simulation.2.The effective wind speed for wind turbine can not be measured directly.Soft sensor modeling for effective wind speed was proposed based on support vector regression(SVR),and the effective wind speed was estimated by using the SVR-based model that relates the corresponding variable with other measurements.A max power point track method integrated with wind speed soft sensor is proposed.The problems about wind speed sampling,SVM trainning and the fusion of wind speed infomation and fuzzy logic controller are solved.The proposed method is able to capture more wind energy.3.The primary issue of wind speed forecasting error compensation is the forecasting error estimation.Forecasting error estimation is converted into a samples classification problem in this paper.Firstly,support vector regression model is trained by wind speed time serials;and the analytics of forecasting errors is done with test samples. Secondly,according to the corresponding margin of error of different samples divided into several classes,to facilitate the training of confidence machine.Finally,confidence machine estimates the margin of forecasting errors.Experimental results show that the accuracy and reliability of the classification can be used to reduce the risk of wrong-compensation,and the proposed approach can achieve higher quality of mean hourly wind speed forecasting.4.An approach of wind speed tolerance intervals prediction is proposed.In support vector regression prediction basis,wind speed tolerance intervals are predicted using inductive confidence machine.Wind speed tolerance interval's width and confidence reflect the accuracy and reliability of the prediction.Compared to pure wind speed forecasting,the accuracy and reliability of the prediction can be used to reduce the risk of decision-making. 5.Short-term wind speed simulation is an important part of wind turbine generation simulation.A wind speed spectrum fitting approach using wavelet transform is proposed. Firstly,Gaussian white noise is decomposed using wavelet transform.Secondly,the wavelet coefficients of every frequency band are shrunk according to the wind speed's power spectrum.Finally,the wind speed signal is reconstructed using the shrunk wavelet coefficients,and a wind speed time series is proposed.By comparing with other method and real wind speed,it was shown that the proposed approach can approximate the spectral density characteristics of wind speeds more accurate.