A Multi-Objective Multi-Constraint Explainable AI-based Approach for Smart Energy Grid Systems

As the global shift toward sustainable energy accelerates, modern electrical grids are increasingly incorporating renewable sources like wind and solar PV. Moving from traditional centralized power to decentralized smart grids demands better optimization and forecasting methods to ensure reliability and efficiency. The inclusion of renewables, energy storage solutions, and bidirectional energy exchanges boosts the flexibility of smart grids and supports sustainable energy practices. Successfully planning and managing hybrid renewable energy systems (HRES) requires accurate and clear forecasting methods in conjunction with robust optimization strategies. In this paper, we present a two-stage Multi-Objective Multi- Constraint Evolutionary computing based on a Genetic Algorithm and Non-dominated Sorting Genetic Algorithm III (GA-NSGA-III) optimization, which enables the procurement of renewable energy sources while concurrently minimizing the Levelized Cost of Energy (LCOE) and the Expected Energy Not Supplied (ENS). We have performed various experiments with real-world data in which, for cost-optimal configurations, NSGA-III decreases LCOE by 26.1% and worst-year ENS by 58.7% relative to NSGA-II. At balanced design, the ENS decreases by 72.7%, and the LCOE falls by 16.3%. This proposed system promotes the utilization of renewable energy sources while also enhancing the efficiency of storage. To enhance the design of hybrid renewable energy systems under uncertain conditions, the proposed method integrates multi-objective optimization, absence-aware reliability modeling, and interpretable load forecasting. These results support more consistent operation and simplify decision-making across multi-year planning horizons. Load forecasting is provided using an Explainable Artificial Intelligence (XAI)-based Interval Type-2 Fuzzy Logic System (IT2FLS). Compared to Type-1 FLS, IT2FLS decreases RMSE by 48.6% and mean absolute error by 50.2% and gains R² from 0.625 to 0.901. These findings show that advanced XAI based multi-objective optimization improves system reliability and resilience.