Assessment of forecasting techniques for solar power production with no exogenous inputs

► We assess forecasting techniques for PV power production without exogenous inputs. ► We test five different methods: Persistent, ARIMA, kNN, ANN and GA optimized ANN. ► ANN and ANN optimized models perform better than persistent, kNN and ARIMA. ► Forecasting error depends strongly on the season. ► GA method efficiently determines ANN inputs and general topology.