Solar Power Photovoltaic Output Forecasting Using Multiple Methods Approach, Case Study: Cambodia

Abstract

Solar energy is one of the most potential renewable energy sources of sunlight. Due to increase and satisfying demand for energy in developing countries like Cambodia, solar power energy is the main and significant energy to the procedure for supply local to reduce import power energy from neighboring’s countries. In this case, the ability to an accurate solar output forecasting is critical for planning to decide based on forecast conditions, while many forecasting methods have been improved for forecasted values. However, the specific research on solar power PV output forecasting in Cambodia is still lacking to secure better accuracy during the rapidly extending inquiry of energy. This study is conducted to investigate a trial of short-term forecasting of solar power photovoltaic output in Bavet city, Cambodia, using several methods for comparisons such as Neural Network (NN), Linear Regression (LR), and Autoregressive Moving Average (ARMA). This process is based on the daily reality historical data from $\mathrm{I}^{\mathrm{s}\mathrm{t}}$ January 2018 to 1$0^{\mathrm{t}\mathrm{h}}$ January 2019 which were recorded by Nation Control Center (NCC). Weather daily index data is obtained from the Renewable Energy Community of NASA Power Data Access Viewer Website Forecast of Global Energy Resources. The reliability of the forecasting of the three methods was assessed by using Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE). Based on the simulation result of these three models, the Neural Network model showed better accuracy and results that were promising and beneficial for solar forecasting in Cambodia.