Development Of A Forecast Model For The Prediction Of Photovoltaic Power Using Neural Networks And Validating The Model Based On Real Measurement Data Of A Local Photovoltaic System

In order to increase the proportion of solar power used to charge electric vehicles via decentralised photovoltaic (PV) systems, a forecast model is required that predicts the energy generated by the system over the next day. For this purpose, the paper presents a model consisting of a forecast model of solar irradiation and a mathematical model for determining the PV power based on solar irradiation. The model has been developed, validated and tested using real measurement data. The forecast model is supposed to he able to predict the solar irradiation in local areas for the next day on the basis of freely available weather forecasts. A neural network has been developed for this purpose, which has been trained and validated on the historical weather data. In the validation, the forecast model reached an accuracy of 60 W/m(2) and 6.68 % related to full scale over the entire year 2017. The model for the prognosis of generated decentralised PV power was tested on a PV system on empirical measurement data. Accuracies of mean absolute error of 51.16 W and 3.18 % related to full scale were achieved.