Study On Short-term Wind Speed And Power Prediction Of Wind Farm

Wind energy is one of the most renewable energy resources with the most development potential at present.Due to the intermittent and random nature of wind energy,the grid-connected operation of wind turbines will adversely affect the operation and stability of the grid.Short-term wind speed and power prediction are the presupposition for large-scale wind power grid,which can conducive to the stable operation of the power system and power grid scheduling.In order to improve the accuracy of prediction,the paper carried on the following research work,and carried on simulation analysis based on the wind farm data.(1)The wind farm short-term wind speed combined forecasting model is established.First of all,the Ensemble Empirical Mode Decomposition algorithm is used to decompose the historical wind speed sequence to reduce the interaction between different feature-scale sequences;Secondly,the Sample Entropy is used to calculate the complexity of each sub-sequence,and the sequences with similar complexity are combined,so as to improve the prediction efficiency;In addition,choose Least Square Support Vector Machine as the basic prediction model,and use the Gravitation Search Algorithm to optimize its parameters,solve the kernel width and regularization parameters to determine the artificial dependency problem;Finally,each new sub-sequence were established GSO-LSSVM model to predict,all the predicted values are superimposed to obtain the final short-term wind speed prediction results.(2)Short-term power forecasting by establishing wind farm power curve.Affected by the external environment,there is a difference between the actual power curve and the standard power curve of wind turbines.In order to establish a power curve that conforms to the characteristics of a wind farm,based on the analysis of three commonly used modeling methods,using Particle Swarm Optimization algorithm to build power curve model.In addition,aiming at the shortcomings of PSO,which is easy to be premature and the algorithm is prone to oscillation near the global optimal solution,using the linear weight decreasing method to improve PSO algorithm,that is,the inertia weight is reduced linearly from the maximum to the minimum.Finally,in combination with short-term wind speed prediction,realize the prediction of short-term power.(3)In view of the fact that a single and certain prediction result can not fully reflect the characteristics of wind power fluctuation and can not meet the needs of risk analysis and control,a forecasting model of short-term power fluctuation interval is established.Firstly,triangular Fuzzy Information Granulation is performed on the historical wind power time series,and the effective component information of each window is extracted according to the needs,get three sequences which consist of the minimum value,trend value and maximum value of each window;Secondly,due to the interval prediction need to handle a large number of samples,in order to improve the prediction efficiency,using Extreme Learning Machine with learning speed as the basic predictive model;In addition,in order to improve thelearning ability and generalization ability of ELM and the prediction accuracy,using the powerful global search ability of Glowworm Swarm Optimization algorithm to optimize the weight of connection between the implicit layer and the output layer in the ELM model;Finally,the GSO-ELM model was established respectively for the three sequences,realize the interval prediction of short-term power.