Hybrid Optimization Techniques for Energy Management in Temperature Stressed Power Grids

Abstract

Temperature-stressed power grids, particularly during prolonged heatwaves driven by climate change, face simultaneous surges in electricity demand for cooling and widespread derating of generation, transmission, and distribution assets. Traditional single-method optimization approaches often struggle with the scale, nonlinearity, uncertainty, and multi objective nature of energy management under these conditions. Hybrid optimization techniques that combine mathematical programming, metaheuristics, machine learning, and decomposition methods offer superior performance by leveraging the strengths of multiple approaches. This research paper presents a comprehensive framework for hybrid optimization in temperature stressed power grids, integrating mixed-integer linear programming (MILP) for unit commitment and dispatch, Lagrangian relaxation or Benders decomposition for large-scale coupled problems, reinforcement learning for adaptive real-time control, and physics-informed neural networks for fast surrogate modeling of nonlinear heat balance and power flow constraints.