Research On Wind Power Ultra Short Term Forecasting Based On Storm

Wind energy uncertainty and volatility have brought huge obstacles to grid integration of wind power.Meanwhile,with the increase of wind turbines and the increase of the number of wind turbines,the data storage capacity of traditional databases is almost close to the limit.Therefore,real-time online forecasting can be implemented Effectively relieve storage pressure,timely access to useful information.Firstly,this paper discusses the feasibility of Echo State Network algorithm based on distributed optimization algorithm-Alternate Direction Multiplier Method in Storm computing platform,designs the topological structure of Echo State Network on Storm,realizes the distributed Echo State Network algorithm based on Storm,The algorithm is validated by using the actual wind power data.Next,compared with the common Echo State Network algorithm and the leakage integral Echo State Network algorithm,the predictive effect of wind power prediction algorithm is analyzed.Finally,aiming at the disadvantage of large prediction error near peak prediction curve,this paper puts forward a distributed predictive value correction algorithm based on Storm,combining with rough set theory,designs and realizes the parallelization of traditional correction algorithm on storm platform,and verifies the actual data set.The experimental results show that the prediction algorithm based on storm has better prediction accuracy than other predictive algorithms,and the improved parallel correction algorithm improves the accuracy of the whole prediction algorithm while guaranteeing the calculation speed.