Agent-Based Modeling for generation bidding strategy Using Policy Gradient Algorithm

With the deregulation of the electricity market in recent years, market participants are actively participating in the power market competition in order to obtain higher profits. Most of the past research models formulate the generation bidding strategy as a mathematical program with equilibrium constraint (MPEC) problem, but the real market environment is often unknown and risky. In this paper, we will construct a power producer who participates in a day-ahead electricity market, which power transmission capacity constraints with risk-averse bidding strategy. Therefore, to solve the above problems, we will use Conditional Value-at-risk (CVaR) to quantify the risk of the electricity market and apply the Policy Gradient (PG) algorithm to solve producers’ optimal bidding strategies.