Research On Combination Model Of Wind Speed Forecasting For Wind Farms

With renewable, non-polluting characteristics, wind energy is increasingly being attention. However, wind speed presents strong randomness and fluctuation which will lead to unstable wind power output. With the increasing proportion of wind power in the electricity grid, wind power will bring some challenges and difficulties for the operation if electric power system. Accurate wind speed prediction can provide necessary technical support and guidance for dispatch of the power system incorporated with wind power, thus to lower or eliminate the adverse effect brought by incorporating wind power. Therefore, the wind speed prediction research is very meaningful.For non-linear characteristics of wind speed data, in this paper, a combination forecasting model of short-term wind speed based on ARIMA and improved Elman neural network is presented. First use ARIMA model to predict the wind speed, the linear rule information is contained in the time series of prediction results, the nonlinear rule information contained in the prediction error of ARIMA. The Elman neural network input variables contains the prediction error of ARIMA model and first-order sequences of historical wind data. Finally, superimpose the Elman neural network prediction error on the ARIMA model prediction results.To prove the effectiveness of the method, compared with ARIMA, ARIMA-BP neural network model. Through verification, on the prediction error, it can reduce the prediction lag, improve forecast accuracy and reduce forecast error than the ARIMA model; on the training speed, the speed is improved by more than30%.Through what has been discussed above, use the combination model to explore in greater depth, and data processing and simulation can be found ARIMA-Elman neural network model has more advantages than a single model to solve practical engineering problems provided some reference.From what has been discussed above, the combination model is discussed in-depth by using the combination model, and data processing and numerical simulation are carried out. So, it can porove that the ARIMA-Elman neural network combination model has more advantages than a single model, provides a method for solving practical problems.