Short Term Local Prediction Based On Data Processing And Its Application In Wind Speed And Wind Power Forecasting

With the deterioration of the environment and depletion of conventional resources,renewable energy has attracted more and more attention.As a type of renewable energy,wind energy has been vigorously developed in the past decades all around the world.However,the stochastic nature of wind power,with its variability and limited predictability,induces difficulties in operating and managing power system,particularly for balancing electricity consumption and production in the electricity market.The accuracy of wind speed and wind power forecast has a direct impact of the operation in power system and electricity market.Therefore,it is very important to improve the prediction accuracy of wind speed and wind power.Based on data processing,this paper focuses on the study of short term local prediction of wind speed and wind power,and tries its best to improve the accuracy.Based on the analysis of the current research,especially the data processing technologies applied in wind speed and wind power forecast,this paper points out the advantages and disadvantages of the traditional methods about decomposition,data reconstruction,similar segments searching and error feedback,and makes improvement.For decomposition,an empirical mode decomposition based filter(EMDF)and a morphological high-frequency filter(MHF)are proposed.Simulations are conducted,and the results indicate that the mean trend revealed by EMDF fluctuates regularly and MHF reveals the tendency of the historical time series.The decomposition results of EMDF and MHF are better than the traditional methods,such as empirical mode decomposition and wavelet decomposition,which lead to better forecast result.For data reconstruction,non-uniform embedding strategy is proposed,which allows different dimensions to use different time delays.And an optimization problem is formulated to find the parameters.Simulation indicates that reconstructing the time series into phase space with non-uniform embedding strategy reveals their characteristics better than that with uniform embedding.To search the similar segments of high frequency component,a double similarity search(DSS)algorithm is proposed,which considers not only the numerical values of the time series,but also their tendencies.Simulation results found that DSS is not suitable for the mean trend.The reason is that DSS pays more attention on the shape of segments but less on the numerical values.However,the value reveals the sustainability of the mean trend more than the shape.In this way,Euclidean distance criteria is more suitable for the mean trend.Based on the above achievements,a novel wind speed and wind power forecast model(MHF/DSS)is proposed.The MHF is proposed to decompose the time series into two components: the mean trend,which reveals the non-stationary tendency of the time series,and the high frequency component,which depicts the fluctuations.The same strategy is employed to forecast the mean trend and the high frequency component,respectively.The two components are reconstructed in the phase space,respectively,where a non-uniform embedding strategy is proposed to better reveal their information.To select similar segments to be used for local forecast,the novel DSS algorithm is proposed for high frequency component,while the Euclidean distance is used for the mean trend.Finally,the least squares-support vector machine(LS-SVM)model is applied to forecast each component,respectively,and their sum composes the final prediction.Besides,another forecast model is formed where an error feedback mechanism is constructed to find out the relationship between the errors and the original forecast values so as to predict the error feedback and add it to the original forecast value.This model is applied in both uniand multi-variate forecast.In multi-variate forecast,the influence factors are taken into account and the forecast ability is improved.Simulation studies are carried out using wind speed and wind power data obtained from four databases,and the results demonstrate that the above two models provide more accurate and stable forecast compared to the other methods,which verifies the effect of the morphological high-frequency filter,non-uniform embedding strategy,double similarity search algorithm and error feedback mechanism.Last but not least,the economic benefit of accurate wind power forecast for power system dispatch is analyzed through building an optimization problem,which aims to minimize the operation cost.The power system operation cost includes the generation cost of conventional power generator,the invoking compensation of balancer set,penalty cost of wind curtailment and the cost of reserve capacity.Two scenarios are employed to verify the advantage of the dispatch model——a micro-grid and a modified IEEE 30-bus test system.The dispatch results of two scenarios have shown that accurate forecast result decreases the cost of reserve capacity,balancer set invoking capacity and the possibility of wind curtailment,which leads to more economic dispatch of power systems.And it proves the importance of accurate wind power forecast.