Techno-economic feasibility of utilizing electrical load forecasting in microgrid optimization planning

Addressing global energy and environmental crises requires transitioning to a renewable energy system. Most system planning studies rely on historical power loads and some power demand forecasting models are inaccurate. This study demonstrates the feasibility of the proposed hybrid model for long-term hourly electricity forecasting and utilizing the forecasted electricity for hybrid renewable energy system (HRES). The particle swarm optimization (PSO) algorithm performs hyperparameter optimization of long short-term memory (LSTM) and eXtreme gradient boosting (XGBoost) models to improve the combined model forecasting accuracy. The proposed model improves performance by up to 9.77% compared to other models on the same dataset and bridges the gap between the two models for and long-term hourly electricity load forecasting. Then the forecasted data is used for system simulation, optimization and sensitivity analysis. The results indicate that the hybrid system can achieve a renewable fraction of 64.1%, supporting the Panamanian government to meet its climate promise. Additionally, the optimal system configuration generates approximately 14.9 billion kWh annually at a cost of energy of $0.0976/kWh and reduces harmful emissions by approximately 57.4%, is validated using genetic algorithm (GA) and PSO algorithm. The HRES is economically superior to traditional grids, with a payback period of less than seven years. Four inflation rate scenarios are set in the sensitivity analysis. A strong linear relationship is demonstrated between the inflation rate, net present cost, and cost of energy. Considering social stability and residents well-being, Panamanian government should incorporate an energy storage system to enhance energy supply reliability.