Solar Photovoltaic Power Estimation: A One Day Analysis Using Neural Network as a Technique

The ability to estimate the daily power production of any solar photovoltaic (SPV) generator, is the fundamental step required towards the future planning and forecast for any SPV energy delivery. Daily forecast of a SPV generator is quite necessary because, solar insolation (irradiation) is unpredictable and can only be available for about 13hours to 15hours limited period within 24hours of the day; depending on weather conditions. In addition to this, SPV plant infrastructures have underlying design faults that makes SPV generators to have inherent losses. For these reasons, knowing the average daily electrical power production of any SPV generator becomes one of the significant criteria, for private and/or commercial investment decisions in standalone SPV or grid-tied SPV system application. In this study, a one-day electrical power production of a SPV plant is modeled, using time series analysis with feedforward neural network. Inputs to the model are datasets containing SPV module temperature, point of array irradiance relating to the local site and the time the weather parameters are sampled. Output to the model will be the estimated SPV electrical power. The performance of the proposed model is verified using mean square error for the validation datasets and the trained datasets.