Risk-Aware Clearing Model for Provincial Electricity Markets with Mean Field Game Simulation

Facing price volatility risks and multi-agent strategic interactions in two-level electricity markets, this paper proposes a bi-level decision framework integrating Conditional Value at Risk (CVaR) and Mean Field Game (MFG). The upper level minimizes purchasing costs and CVaR risk for inter-provincial traders, while the lower level models market clearing based on generation cost minimization in sending provinces. An MFG model with homogeneous traders is then introduced, coupling individual optimal control (HJB equation) with population distribution evolution (FPK equation) to capture collective behavior feedback on prices. An adaptive regularization Deep Q-Network algorithm is designed to improve computational efficiency. Multi-scenario simulations using grid data analyze the effects of risk aversion, uncertainty, and network congestion on costs and price volatility. Results show that the model effectively characterizes the interplay between risk attitudes and market equilibrium, offering theoretical and practical support for risk management and market design in inter-provincial trading.