Study On Wind Speed Prediction Of Boundary Layer

With the increasing severity of energy and environment problems,renewable energy plays a vital role in the global energy supply.Wind energy is an emerging renewable energy source,as a viable alternative to chemical fuels in the past decades,the development and utilization of wind energy has attracted the attention and attention of countries around the world.Wind speed usually exhibit strong fluctuating and nonstationary feature,which can affect the safe and stable operation of the power grid.This problem can be effectively solved by predicting wind speed,which is of great importance to the development and utilization of wind energy.In this paper,according to the current research status in the field of wind speed prediction,combined with field measurement data,the problem of improving the accuracy of point prediction and quantifying the uncertainty of prediction results is studied.The main research content is as follows:(1)The time correlation,non-stationary and chaotic characteristics of wind speed time series are analyzed,autocorrelation function and power spectrum are used to analyze the characteristics of wind speed.Aiming at the chaotic characteristics of the wind speed sequence,the phase space reconstruction and neural network are combined to calculate the delay operator and the embedded dimension,we build a prediction model for the measured sample.The results show that using the data after phase space reconstruction can reduce the prediction error.(2)Based on signal decomposition technology,a combined prediction model based on variational mode decomposition and neural network is developed.First,wind speed series is decomposed into several intrinsic mode functions(IMFs)based on variational mode decomposition.After obtaining the IMFs,the sample entropy of each IMF is calculated to investigate the complexity of the signal.The neural network combination forecast model is proposed based on the sample entropy.Finally,the final predicted wind speed series are fitted by the analysis results of each component using BP neural network.The feasibility of the proposed prediction model is verified by modeling and predicting the wind measurement samples of 325 m wind measurement tower in Beijing.(3)In order to address the problem of quantifying the uncertainty of prediction results,a wind speed probability interval prediction model based on variational modal decomposition-sample entropy and long-short-term memory neural network quantile regression is proposed.Based on the measured typhoon data at 30 m height in the suburbs of Wenzhou,a prediction model was established to construct a prediction interval at different confidence levels.The simulation results show that the interval coverage of the proposed model can meet the requirements of the confidence level when the confidence level is the same,and the average width of the interval is the smallest.The prediction results under different quantile are estimated by the Gaussian kernel density,and the probability density function of wind speed is obtained.