Time Series Forecasting for Daily to Monthly Temporal Hourly-based Solar PV Output Power

This paper presents the application of the Auto-Regressive Integrated Moving Average Exogenous (ARIMAX) model and compares its performance with Auto-Regressive Integrated Moving Average (ARIMA) and Seasonal Auto-Regressive Integrated Moving Average (SARIMA) in forecasting daily, weekly, and monthly average solar PV output power. This study considers long-term hourly temporal-based solar PV output power for the Java-Bali region of Indonesia, as obtained from the Renewables.ninja solar PV model web-based tool. Using the Dash framework and Python, the study develops a web-based dashboard application that allows users to explore and analyse daily to monthly forecasting using these three methods. The testing results show that the time series methods are best suited for predicting monthly average output power, with the ARIMAX outperforming all other methods when applied to all cities/regencies in Central Java. It achieved the RMSE values of 10.74, 25.36, and 60.27 for daily, weekly, and monthly forecasting, respectively.