Adaptive meta-learning extreme learning machine with golden eagle optimization and logistic map for forecasting the incomplete data of solar irradiance

Abstract

Solar energy has become crucial in producing electrical energy because it is inexhaustible and sustainable. However, its uncertain generation causes problems in power system operation. Therefore, solar irradiance forecasting is significant for suitable controlling power system operation, organizing the transmission expansion planning, and dispatching power system generation. Nonetheless, the forecasting performance can be decreased due to the unfitted prediction model and lacked preprocessing. To deal with mentioned issues, this paper proposes Meta-Learning Extreme Learning Machine optimized with Golden Eagle Optimization and Logistic Map (MGEL-ELM) and the Same Datetime Interval Averaged Imputation algorithm (SAME) for improving the forecasting performance of incomplete solar irradiance time series datasets. Thus, the proposed method is not only imputing incomplete forecasting data but also achieving forecasting accuracy. The experimental result of forecasting solar irradiance dataset in Thailand indicates that the proposed method can achieve the highest coefficient of determination value up to 0.9307 compared to state-of-the-art models. Furthermore, the proposed method consumes less forecasting time than the deep learning model.

Keywords

• Data imputation
• Golden eagle optimization
• Logistic maps
• Meta-learning extreme learning machine
• Renewable energy forecasting