Ultra Short Term Wind Power Prediction Based On ISSA-LSTM

Due to the strong pollution and non renewable nature of traditional fossil energy,energy transformation is imperative.As a renewable and clean energy,wind energy has been widely concerned because of its great development potential and application space.However,due to the intermittence and volatility of wind energy,the grid connection of wind power brings challenges to the safe and stable operation of the whole power system.The effective and accurate prediction of wind power is of great significance for power grid dispatching.Start with the widely used artificial intelligence methods,aiming at the problem that it is difficult to determine the super parameters of long short term memory neural network(LSTM),this paper proposes an ultra short term wind power prediction model based on improved sparrow search algorithm(ISSA).Firstly,aiming at the shortcomings of the traditional method for determining the model’s hyperparameters,the sparrow search algorithm(SSA)is used to optimize the hyperparameters setting process of the model.The number of neurons in the first layer of hidden layer,the number of neurons in the second layer of hidden layer,the number of iterations and learning rate of LSTM with four layer structure are optimized by SSA.The average absolute error is taken as the objective function,the SSA-LSTM model is built,Compared with LSTM,the results show that SSA-LSTM has high prediction accuracy.Compared with LSTM model,the root mean square error,average absolute error and average absolute percentage error are reduced by 25.06%,14.37%and 0.49%respectively,and the Nash coefficient is increased by 2.62%.Secondly,aiming at the shortcomings of SSA,an improved sparrow search algorithm is proposed.The specific improvement measures are as follows:tent chaotic map is used to improve the population initialization process and replace the pseudo-random number generator in SSA to make the initial distribution of the population more uniform;The dynamic step weight formula is introduced to replace the random number of the control step in SSA,so that the algorithm searches with a longer step in the early stage and a smaller step in the later stage,which improves its global search ability,and makes the discoverers in the population affected by the global optimization of the previous generation;Through Gaussian perturbation,the randomness and diversity of the optimal feasible solution are enhanced,and the greedy rule is used to judge whether to replace the most feasible solution with the optimal solution after Gaussian perturbation.Five basic test functions are used to test the performance of IAAS.Compared with whale optimization algorithm(WOA),Grey Wolf algorithm(GWO)and sparrow search algorithm,the results show that ISSA has strong anti-interference ability and global search ability,faster convergence speed and smoother convergence curve.Finally,the ISSA-LSTM ultra short-term prediction model is simulated and analyzed with the data of a wind farm.Compared with LSTM,SSA-LSTM,WOA-LSTM and GWO-LSTM,the results show that the proposed model ISSA-LSTM can significantly reduce the prediction error,improve the stability and quality of the model,and its root mean square error,average absolute error and average absolute percentage error are reduced to 0.40922,0.30985 and 2.8336%respectively,Compared with SSA-LSTM,it decreased by 1.53%,13.56%and 0.32%respectively,and the Nash coefficient increased to 0.96852,which increased by 0.1%compared with SSA-LSTM.