Abstract
In the pursuit of sustainable industrial operations, efficient energy management has become a critical challenge, particularly under scenarios where the electrical grid is restricted to serving industrial loads. This study addresses the urgent need for intelligent forecasting and scheduling frameworks by proposing a hybrid Gene Expression Programming Adaptive Neuro-Fuzzy Inference System (GEP-ANFIS) for predictive energy management in hybrid renewable energy systems. The model was evaluated using standard forecasting metrics. For solar PV prediction, GEP-ANFIS achieved low short- and long-term error rates, with MAPE values below 6% and 8%, respectively. For industrial load forecasting, the model exhibited high precision, maintaining MAPE values under 2.5% (short-term) and under 3.5% (long-term). These results demonstrate consistent improvements over conventional ANFIS and GEP models. Economic evaluation confirmed significant cost benefits. In a Grid-only configuration, GEP-ANFIS reduced daily energy costs by 7.4% compared to ANFIS. Greater efficiency was observed in PV and Battery-only and Grid-connected PV-Battery setups, where GEP-ANFIS achieved daily cost reductions of 6.5% and 6.3%, respectively. Over a 20-year planning horizon, the system recorded a 6.5% reduction over ANFIS and a 37.7% improvement over HOMER. A sensitivity analysis was also conducted to assess the robustness of the GEP-ANFIS model under varying solar PV power, and battery storage capacity. Results indicated the robustness, efficiency, and scalability of the GEP-ANFIS controller, especially in resource-constrained, PV-dominated microgrids, making it a strategic solution for sustainable industrial energy management while preserving battery longevity by avoiding deep discharge scenarios.