Short-Term Solar Irradiance Prediction Using Time Series Analysis and Neural Networks for Green Energy Park Photovoltaic Plant.
Abstract
Short-term solar irradiance forecasting is of great importance for optimal operation and grid integration of photovoltaic (PV) plants. Morocco’s Green Energy Park (GEP) of Benguerir is investigating the possibility of providing accurate solar forecasts for its own PV plant. However, to overcome the variability of solar resources under changing weather conditions, a reliable forecasting model is needed. In this context, this work provides a detailed research in short-term solar irradiance forecasting, an implementation and a technical analysis were performed in this study for a suitable and a reliable Global horizontal irradiance (GHI) forecasting model. Comparing the performance of different models such as ARIMA, Multiple linear regression, curve fitting and artificial neural networks, gave valuable information for further development of the final model. This final model is based on NARX ANN. In order to achieve meaningful results when comparing these different approaches as well as for NARX ANN model assessment, a standardized methodology for evaluation is used. NARX model outperformed all the approaches carried out in this study and it gave very satisfying results of 4,57 % MAPE error for Clear-sky and 14,67% MAPE error for Cloudy weather conditions.
Keywords
Ghi, Forecasting, Arima, Mlr, Curve fitting, Ann, Narx