In a deregulated power market, participants have to decide their bids under various uncertainties. The decisionmaking on how to bid can be modeled as a fuzzy inference procedure using fuzzy rules that map environmental information to actions. The tuning of fuzzy database is based on a reinforcement learning method called symbiotic evolution learning method where fuzzy sets and fuzzy rules are learned based on reward/punishment signals. Such a learning ability enables our agents to adapt to different situations without being taught beforehand. Experiments show that our model agrees with cournot game and bertrand game solution in game theory method if agents? strategies are restricted to quantity competition or price competition.