Research Of Short-term Wind Speed Prediction

With the aggravation of energy crisis,renewable energy has drawn increasingly attention worldwide.As one of promising renewable energy,wind power generation is developing rapidly and the proportion of wind power in the power system is increasing year by year.Now it has become one of the most widely used distributed generation technologies.Therefore,accurate wind speed prediction is crucial for the development of wind energy as well as the management and planning of wind farm.Moreover,it is helpful to ensure the safety,stability and economic efficiency of power system.Because the actual wind speed has the characteristics of volatility and intermittency,which may greatly increase the difficulty of wind speed prediction and thus lead to bad forecasting results.In this paper,a novel model based on a hybrid mode decomposition(HMD)and online sequential outlier robust extreme learning machine(OSORELM)is proposed for multi-step wind speed prediction.In wind speed data pre-processing period,in consideration of the unsteady characteristics of wind speed,the original wind speed time series is deeply decomposed by HMD,which is comprised of variational mode decomposition(VMD),sample entropy(SE)and wavelet packet decomposition(WPD),and VMD is used as the principal decomposition method.First,VMD is applied to decompose the original wind speed into a set of band-limited sub-series;then all sub-series are analyzed by SE method;and part of the sub-series with high SE value is taken into further decomposed by WPD.Finally,all sub-series obtained by HMD have more obvious regularity characteristics and simpler sequence tendency.In model training period,in order to reduce the influence of the randomness of input-weight and hidden layer biases in OSORELM model,the crisscross algorithm(CSO)is used to optimize the initial parameters of the model and thus improve the stability of the model prediction.The model optimized by CSO is denoted as OSORELM-C model.On this basis,all sub-series obtained above are forecasted by OSORELM-C,and the results of all sub-series are generated to obtain final prediction results of the wind speed.In order to validate the effectiveness of the proposed model in 1-step,2-step and 3-step wind speed prediction,the actual wind speed data provided by NREL is used f or experiments.The experiment results show that:(1)The prediction accuracy has been greatly improved with the decomposition by HMD method,and the forecasting results are better than those models based on single decomposition method,which indicates tha t HMD is an effective way for wind speed decomposition and it can capture the potential characteristics of the wind speed more effectively.(2)Compared with other prediction models,OSORELM model has achieved better prediction results in multi-step wind speed prediction.Moreover,OSORELM can maintain stable prediction results with the increase of forecasting steps,which shows the superiority of OSORELM model,and it indicates that the online model has better adaptability and accuracy in practical forecasting.Moreover,the prediction accuracy can be further improved with the optimization by CSO.(3)The proposed HMD-OSORELM-C can improve the wind speed prediction accuracy effectively.