Comparison between ARIMA and ANN Models Used in Short-Term Wind Speed Forecasting

Abstract

Wind power is a significant alternate energy in times of energy crisis. In virtue of its intermittency and fluctuation, it poses several operational challenges to grid interfaced wind energy systems. This paper introduced autoregressive integrated moving average (ARIMA) model and artificial neural network (ANN) to forecast the hourly wind speed one to four hours ahead. The models are applied to wind speed records for each month separately from a wind park in Hubei province of China. The experimental results demonstrate that these models are in good agreement with measurement values. The ANN model does a better job than ARIMA model in forecasting short-term hourly wind speed. Besides, different time series with different variance must construct different models. But when the variance is too high, it needs combined models or numerical weather predicting method to pursue better results.