Data Prediction Algorithm for Wind Power Generation Based on Improved LSTM
At this stage,the long short-term memory(LSTM)neural network has some problems in forecasting wind power generation data,such as over fitting,large super parameter selection range,higher time cost and difficulty in improving accura-cy.In view of this,this paper proposes a wind power prediction algorithm which uses sparrow search algorithm(SSA)and at-tention mechanism(AM)to optimize LSTM,and constructs the corresponding SSA-SZLSTM-AM prediction model.On the basis of improving LSTM structure by surprisal-driven zoneout(SZ)regularization,SSA is adopted to solve the problem that the prediction accuracy of LSTM is difficult to improve due to the large randomness of super parameter selection.At the same time,important features are screened out by AM during LSTM prediction,which further improves the prediction accuracy of wind power,and the feasibility of the proposed algorithm is verified by using wind power data of wind farms.The results show that,compared with the 6 comparison algorithms,the root mean square error and mean absolute error of the proposed algo-rithm reduce by 17.93 percentage points and 5.34 percentage points,respectively,while the coefficient of determination(R2)increases by at least 2.14 percentage points,which effectively improves the prediction accuracy of wind power.
wind power predictionsparrow search algorithmattention mechanismlong short-term memory neural network
万方数据
维普
