Research On Short-term Power Combination Forecasting Model Of Wind Farm Based On Intelligent Optimization Algorithm

Wind energy is an inexhaustible energy source that does not contain any polluting impurities.The technology for its development has become quite mature so far.One of the most widely used wind resource development methods is wind power generation.my country is currently in a period of rapid development of the wind power industry.As of the end of 2019,about 210 million KW of wind power installed capacity integrated into the grid has become the world's number one.The uncertainty of output power is one of the root causes of a series of grid-connected problems faced by wind power.The most effective way to solve the uncertainty of output power caused by the randomness of wind speed is to implement an accurate prediction mechanism for wind power.This paper combines the numerical weather forecast(NWP)data provided by a wind farm in Gansu Province with the original historical wind power data,and conducts research on short-term wind power forecasting through a combination of algorithms.Firstly,this article analyzes and elaborates on domestic and foreign wind power forecasting technologies,analyzes the impact of changes in meteorological parameters over time on wind power output power,and addresses the original data due to equipment failures,wind abandonment and power rationing,and wind turbine maintenance.Factors that cause abnormal phenomena in power data are eliminated by referring to the wind speed-power standard curve of wind turbines;in view of the impact of input data dimensions on algorithm efficiency,mutual information method and random forest feature importance method are used to determine the relationship between feature parameters and output power Analyze the correlation of the output power and select the parameters with the highest correlation with the output power to reduce the dimensionality of the input variables in the prediction model.Secondly,on the basis of many researches on wind power forecasting,this paper uses three single algorithms,BP neural network,XGBoost and long short-term memory(LSTM)neural network,to predict the output power of wind power.Through simulation analysis of the prediction result curve of each single prediction model,calculate their absolute mean error(MAE),root mean square error(RMSE)and R~2,and find that the predicted value of wind power output power and the true value of the wind power output power are basically the same in time.,Which shows that the above single prediction algorithm can reflect the change law of the output power of the wind farm.However,through the analysis of the MAE,RMSE and R~2 of the prediction results of the three single models,it is found that the prediction accuracy of XGBoost in the above algorithm is higher.Thirdly,in order to further improve the accuracy of short-term wind power forecasting,this paper adopts a combined forecasting method for modeling.According to the advantages of convolutional neural network(CNN)in extracting time series features and the advantages of long-term short-term memory neural network(LSTM)in time series memory,they are combined to perform short-term prediction of wind power power.After simulation analysis,it is found that CNN-The improvement effect of the LSTM prediction model is not as good as expected.Therefore,this article uses the Sparrow Search Algorithm(SSA)to optimize the learning rate and the number of neurons in the LSTM and CNN-LSTM networks to establish SSA-LSTM and The SSA-CNN-LSTM prediction model simulates the historical data of a wind farm in Gansu Province.The results show that the Sparrow Search Algorithm(SSA)has good convergence and optimization capabilities,and can find the optimal parameters faster and more accurately.The prediction accuracy is thus effectively improved,and the accuracy of the SSA-CNN-LSTM prediction model is higher than that of the SSA-LSTM model.