A Particle Swarm Optimization-Long-Short Term Memory (PSO-LSTM) Hybrid Model for Forecasting Global Horizontal Solar Radiation

Electrical energy demands require clean and sustainable energy production. Solar power output as a clean energy depends on solar radiation and other meteorological features. The variability of solar radiation and other meteorological features impaired solar energy output, significantly affecting the economics of intelligent grids and microgrid operation reliability. This study proposed a hybrid system that comprises a particle swarm optimization (PSO) algorithm and Long-Short Term Memory (LSTM). The PSO algorithm was employed to conduct the adaptive parameter adjustment of the LSTM models. The study benchmarked the proposed hybrid PSO-LSTM model with LSTM-based Genetic Algorithm (GA), Grey Wolf Optimization (GWO), and Ant Colony Optimization (ACO) models. Four performance evaluation metrics were utilized to compare forecasting accuracy models: R-Square, RMSE, MAE, and MAPE. Comparing the developed model, PSO-LSTM, R² (0.950643), RMSE (60.304873), and MAE (25.442901) with the benchmarked model, it was discovered that the PSO-LSTM is superior. Its full implementation will assist in integrating solar systems into the national grids and enable the stakeholders and policymakers to make effective decisions.