Wind power forecasting under nonlinear conditions using fuzzy time series and long-short term memory models

Abstract

In the present day, there has been an increased focus on sources of clean energy, especially wind, since the depletion of fossil fuel reserves. Using the possibility of wind speed presents obstacles and complexities due to its non-linear characteristics. Therefore, a precise and effective wind energy forecast will greatly assist in resolving the system's operating and planning issues. The Forecasting of wind power is done through three forecasting techniques, the fuzzy time series (FTS) method, the long-short term memory (LSTM) and auto-regressive integrated moving average (ARIMA) method of forecasting. A comparative evaluation utilizing root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) indicates that the proposed FTS approach demonstrates superior performance of RMSE of 3.053 and a MAPE of 17.892% relative to both LSTM of RMSE of 5.378 and a MAPE of 32.844% and ARIMA of RMSE of 3.901 and a MAPE of 25.105%, attaining the minimal prediction errors. The outcomes show that FTS is effective well for datasets with seasonal changes and small training sizes.