Tracking Bounded Rationality in Day-Ahead Electricity Markets: A Dynamic Cognitive-Hierarchy Bidding Method

Day-ahead electricity bidding is increasingly shaped by behavioral uncertainty, as human traders and automated bidding agents coexist in the same market. Current strategic bidding methods typically assume fully rational or behaviorally stationary competitors, and thus fall short in repeated markets where competing bidders may be boundedly rational or shift their strategic depth over time. To address this challenge, this paper proposes a Dynamic Cognitive-Hierarchy (DCH) Bidding method. By integrating cognitive hierarchy theory, we model the heterogeneous bounded rationality of market opponents across varying strategic depths. Moreover, a discounted Bayesian updating mechanism is introduced to continuously distill daily market-clearing outcomes into an adaptive belief over opponents' prevailing cognitive levels. Case studies on a synthetic day-ahead market with an unannounced regime shift show that the proposed method swiftly infers the competitors' post-shift cognitive depth, attains 98.9% of the theoretical upper-bound profit, and reduces cumulative regret by 71.8% and 89.5% relative to Static Bayesian and Static mixture baselines.