Solar power forecast with weather classification using self-organized map

This paper presents a practically efficient method for 1h to 3h ahead solar power forecast (SPF) with weather classification using a Self-Organized Map (SOM). The method requires minimal input data which are normally monitored at all sites so that additional expensive weather observatory devices are not necessary. The proposed SOM can classify weather condition into sunny, cloudy and rainy day, using only the current and past irradiance profile data as input. Then, the ANN-based SPF model for each weather condition is constructed, trained and validated, accordingly. Performance of the proposed method has been evaluated using actual field data from the 500-kW solar farm in Mae Hong Son district, Thailand. Accuracy performance obtained indicates that additional spinning reserve of less than 10 %, with respect to the installed solar power capacity, would be required when the proposed SPF method is applied to the energy management system (EMS) for the city.