As power grids transition toward high renewable penetration, traditional capacity expansion planning methods struggle to account for the complexity of solar-dominated systems with integrated storage. Researchers have developed an integrated framework that combines AI-driven demand forecasting with advanced optimization to improve long-term planning decisions.
The framework employs a hybrid grasshopper optimization algorithm paired with random forest machine learning for demand prediction, achieving exceptional accuracy with 98.76% testing performance and minimal error metrics (MAE of 0.1079). This high-fidelity forecasting provides reliable constraints for subsequent capacity planning analysis across multiple scenarios.
To overcome computational limitations in multi-objective optimization, the team developed a novel Differential Evolution-based Harris Hawks optimization (DEHHO) algorithm. Conventional optimization methods often face a critical trade-off: insufficient exploration leaves solutions trapped in local optima, while excessive exploration delays convergence. The DEHHO approach balances these competing demands through integrated mechanisms that improve both solution quality and computational speed.
Testing across 12-year and 28-year planning horizons revealed significant benefits. Under high solar penetration scenarios, system costs decreased by 8.12% to 16.7% compared to baseline planning. Reliability metrics improved substantially, with loss of load probability declining by 79.5 to 86.1% and unmet demand dropping correspondingly. The model simultaneously evaluates trade-offs among capital costs, reliability, emissions reduction, and operational flexibility—critical considerations for decarbonization pathways.
The framework's ability to handle competing objectives reflects real-world planning constraints. Utilities must balance affordability, grid resilience, environmental goals, and system flexibility as they deploy renewable capacity. By quantifying these trade-offs explicitly, planners can make informed decisions aligned with stakeholder priorities and regulatory requirements. The demonstrated improvements in both economic and reliability metrics suggest the approach could significantly enhance planning outcomes across diverse grid contexts.
