Deep-Economic Dispatch: A Deep Reinforcement Learning Framework for Optimal Power Generation Scheduling in Real-Time Electricity Markets

The increasing complexity of real-time electricity markets and fluctuating load demands challenges traditional economic dispatch methods, which often fail to balance cost efficiency and system reliability. This study presents a Deep Reinforcement Learning (DRL) framework for real-time power dispatch, enabling adaptive and intelligent scheduling of generation units. By leveraging state-value and action-value functions, the agent learns optimal dispatch policies directly from system interactions without requiring explicit system models. The framework integrates real-time reward signals, constraint penalties, and gradient-based policy updates to minimize total generation costs while maintaining generator limits and demand-supply balance. A six-unit case study demonstrates the model’s performance, achieving peak cost reductions of $23,780 during high-demand periods and overall economic improvements of 10–15% compared to traditional Classical Economic Dispatch (CED). The DRL agent consistently maintains reliability at 100% across varying load conditions (700–2010 MW), while respecting operational boundaries and avoiding violations. Policy probabilities adapt flexibly to changing generation requirements, ensuring both efficiency and robustness. These results confirm that DRL not only reduces operational costs but also provides a scalable, intelligent approach to real-time electricity market management, outperforming conventional static methods. The study highlights the potential of integrating learning-based techniques into modern smart grids for adaptive, cost-effective, and reliable power generation scheduling.