Solar irradiance forecasting using global forecast system (GFS)

Ho, Mun Bock (2020) Solar irradiance forecasting using global forecast system (GFS). Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Preview

Text (Full Text) Solar irradiance forecasting using global forecast system (GFS).pdf - Submitted Version Download (2MB) | Preview

Abstract

Smart grid is the next generation of electrical grid that can be integrated with renewable source to produce electric power. Then, there is a challenge to integrate renewable energy such as solar into smart grid. This is because the output of solar energy is related to the solar irradiance which is lack of stability due to weather variation. Therefore, solar irradiance forecasting is the solution to solve this problem. Then, multiple regression (MR) and neural network (NN) models were built. For the model validation, neural network model has achieved correlation coefficient (R) of 0.9173 and root mean square error (RMSE) of 114.1820 which better than MR model. With weather forecast produced by seasonal ARIMA, one day ahead (inter-day) and an hour (intraday) solar irradiance were forecasted by MR and NN models. In addition, there is Global Forecast System (GFS) was applied then blended model formed by blending MR, NN and GFS models together. As the result, for inter-day forecasting, weighted mean absolute percentage error (WMAPE) of 11.79% achieved by NN model on sunny day. Then, blended model has the lowest WMAPE of 30.50% on cloudy day. On the other hand, NN model has outperformed compared to other models for intraday forecasting.

Actions (login required)

View Item