Solar Energy Prediction Using Complementary EEMD and Optimum Least Square-Support Vector Machine

Accurate solar energy prediction is adequate for the robust combination of solar power into the energy grid, optimizing energy management, and reducing reliance on nonrenewable resources. This work integrates the Complementary Ensemble Empirical Mode Decomposition (CEEMD) and optimum Least Square-Support Vector Machine (OLS-SVM) for accurate solar energy prediction. Initially, the dataset is pre-processed by the cubic spline interpolation (CSI). The CEEMD is presented for extracting the non-stationary and non-linear solar energy time-series data into intrinsic mode functions (IMFs), isolating different frequency components and reducing noise. These IMFs are fed as input features for the OLS-SVM, a robust machine learning (ML) model that can manage computational efficiency and capacity to handle nonlinear relationships. Here, the optimal process is done by the rat optimizer (RO) The LS-SVM is trained and optimized to predict future solar energy values based on the extracted features. The suggested CEEMD-OLS-SVM approach provided better MSE (0.2) and RMSE (0.3) in solar energy prediction compared to conventional ML models. By utilizing the strengths of both CEEMD and OLS-SVM, this method provides a reliable tool to improve solar energy prediction and enhance the robustness of solar energy utilization.