"Kalong Microgrid load data"

"The increasing deployment of renewable-based microgrids requires reliable short-term load forecasting to support energy management under demand uncertainty. This paper proposes a hybrid forecasting framework for day-ahead load prediction in isolated solar microgrids. The approach integrates Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Gated Recurrent Units (GRU), and Particle Swarm Optimization (PSO). CEEMDAN decomposes nonlinear load signals into intrinsic mode components to reduce noise and prevent modal aliasing. Subsequently, the GRU network is employed to capture temporal dependencies within each decomposed series. PSO is utilized to optimize GRU hyperparameters, improving overall forecasting accuracy and generalization. The proposed PSO-CEEMDAN-GRU model is validated using hourly load data from the Kalong community microgrid in Niger State, Nigeria. Results show that the proposed PSO-CEEMDAN-GRU model reduces the root mean square error (RMSE) by 93.69% and 93.97% compared with conventional GRU and bidirectional GRU models, respectively. These findings demonstrate significantly improved forecasting performance for standalone renewable microgrids."