Solar Power Time Series Prediction Using Wavelet Analysis

Following the major revolution in the sun energy field, the forecast of this one is getting increasingly relevant in competitive electricity markets. Since the solar energy produced is acutely dependent on the weather conditions, the unanticipated variation in the solar power would increase operating cost. Furthermore, the unpredictability of the output power will be a significant hindrance in incorporating this variable energy resource into the network. The object of this paper is to afford an accurate solar power forecasting technique, that may reduce the inaccuracies involved with predicting the near future generation at different weather conditions. This technique is based on the association of the wavelet transform (WT) with ANN and SARIMA models. In this proposed model the WT is used to ensure a considerable effect on the ill-behaved solar power time-series. The ANN and the SARIMA models are applied to both approximation and detail components of the WT in order to capture the nonlinear behavior of the solar power data. The second aim of this study is to evaluate the impact of the detail components of the WT, by comparing the predicted outcomes with and without the detailed elements. The reliability of the suggested models has been demonstrated by comparing statistical performance measures in terms of MAPE and RMSE.