PV Power Forecast Based on Artificial Neural Network at Indonesia Shopping Mall PV Rooftop

In this research, forecast of Photovoltaic (PV) rooftop power output has been carried out. Forecast is using machine learning Artificial Neural Network (ANN). ANN is precisely used for forecast by memorizing and learning data patterns that are used as training and targets. The input training data used are irradiance, PV temperature and time per 10 minutes and time per 1 hour for 6 days. While the target is the PV output power data for 6 days. The training algoritm that used is Levenberg-Marquardt (LM) Algorithm. LM trains the network faster and has high rate of convergence. The performance parameter used to determine a good ANN forecast is the Mean Square Error (MSE). After conducting the training data, the data is tested for ANN forecast using input data on the seventh days. The output is the forecasted PV output power. The forecast results will be compared with the PV output power measurement data.