Off-grid Photovoltaic System Power Output Medium-Term Forecasting Using Artificial Neural Network

The use of fossil energy which is always increasing from year to year makes it seem as if the world cannot be separated from these energy source. Therefore, it is necessary to find new energy sources where the energy can be renewable and available for a long time. In this research, a system is designed to predict the power output of PV. This system uses solar irradiation data and the power output of an off-grid solar power plant as the dataset. The dataset obtained from the PV output will be processed using an artificial neural network (ANN) with a backpropagation algorithm. The results of this study are able to predict the medium-term PV power output using the artificial neural network method by looking at the expected error values of Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE). Tests are carried out to predict the PV power output for the next 11 days. The ANN model architecture uses 2 hidden layers with 3 neurons in the layer, 7 neurons in the second layer, and 190 epochs. This model has an error value of 25.837% for MAPE, 0.166 for MAE, 0.043 for MSE, and 0.209 for RMSE which categorizes the model as fairly feasible on predicting the next 11 days of PV power output.