Deep Learning Forecasting of Photovoltaics Output Using Digital Twin Data

Solar power is an accessible form of renewable energy, especially in South Africa, where solar radiance is concentrated. However, this source of energy is not constant in supply. Its stochastic nature can lead to fluctuations in the voltage and frequency in the energy grid. Accurate forecasting of the availability of solar power is crucial for operational efficiency and distribution of power. Machine learning models provide a powerful forecasting tool for solar photovoltaic power. Due to the inaccessibility of accurate and measured solar irradiance data, this study uses a digital twin to generate meteorological attributes to predict the solar output power of the University of Johannesburg solar photovoltaic plant based in Auckland Park, Johannesburg. The synthetic dataset contains hourly data over a calendar year. The deep learning algorithms that were used for prediction are the Long-Short Term Memory (LSTM) and Recurrent Neural Network. The findings showed that the LSTM is the best predictor using the Root Mean Squared Error (RMSE), Mean Squared Error (MSE) and Mean Absolute Error (MAE) performance metrics. This study is crucial as it supports the Sustainable Development Goals (SDG) aim to reduce carbon emissions, provide clean energy and improve access to electricity.