Research On Combined Forecasting Of Short-term Wind Power

Wind power is unstable,and as its penetration rate in large power grids increases year by year,wind power has become one of the uncontrollable factors in power systems.Accurate short-term prediction of wind power can optimize grid dispatching,which is an effective way to solve the impact of wind power instability on the grid,and thus improve the grid’s ability to absorb wind power.This paper combines numerical weather prediction(NWP)and historical power data of a small wind farm in Yunnan Province in January,April,July and October to conduct short-term wind power combination forecasting research.Firstly,data processing is performed,and there is a mismatch between meteorological and power data for the original data.On the basis of rationality test,the data were removed according to the standard wind power curve of the unit;For the phenomenon of missing data records,Mean Method,Linear Interpolation Method and Near-field Fill Method were used to process the missing data;According to the influence of multi-dimensional NWP on the efficiency of the algorithm,the Pearson correlation analysis method is used to calculate the correlation between different NWP factors and output power,so as to determine the main influencing factors and reduce the input dimension of the algorithm reasonably.Secondly,BP neural network,ANFIS and wavelet neural network are used to predict each month.By analyzing the prediction effect diagram and the forecast trend indicator(CC)of each algorithm in different months,it is found that the prediction curve is similar to the real value fluctuation trend,which means that the above algorithms have general applicability to the wind field.However,by analyzing the absolute value mean error(MAE)and root mean square error(RMSE)of the prediction results,it is found that none of the above algorithms can always maintain the best prediction effect in different months,that is,a single algorithm cannot adapt to the whole Different forecast scenarios for the year.Finally,for the problem that a single algorithm can’t adapt to the forecasting scenarios of different months in the whole year,this paper adopts the method of weighting the three kindsof single algorithms to make combined forecasting.Firstly,the combination of average weight method and covariance optimization method is used to predict the combination.The results show that the combined prediction effectively reduces the impact of different scenarios on the prediction effect,and reduces the monthly average MAE and RMSE.Aiming at the weight assignment problem of the average weight method and the covariance optimization combination method,the optimal weight determination based on the small world optimization algorithm is further adopted.The prediction results show that the method further improves the prediction accuracy.