Daily wind speed forecasting through hybrid KF-ANN model based on ARIMA

Abstract

The accuracy of wind speed forecasting is important to control, and optimize renewable wind power generation. The nonlinearity in the patterns of wind speed data is the reason of inaccurate wind speed forecasting using a linear autoregressive integrated moving average (ARIMA) model. The inaccurate forecasting of ARIMA model reflects the uncertainty of modelling process. The aim of this study is to improve the accuracy of wind speed forecasting by suggesting a more appropriate approach. An artificial neural network (ANN) and Kalman filter (KF) will be used to handle nonlinearity and uncertainty problems. Based on the ARIMA model, a hybrid KF-ANN model will improve the accuracy of wind speed forecasting. First, the effectiveness of ARIMA will be helped to determine the inputs structure for KF, ANN and their hybrid model. A case study will be carried out using daily wind speed data from Iraq and Malaysia. The hybrid KF-ANN model was the most adequate and provided the most accurate forecasts. In conclusion, the hybrid KF-ANN model will result in better wind speed forecasting accuracy than its separate components, while the KF model and ANN separately will be provide acceptable forecasts compared to ARIMA model that will provide ineffectual wind speed forecasts.