Under the background of increasing volatility of new energy and complex trading behavior, the formation of electricity price in power market has shown significant dynamic coupling characteristics. Aiming at the problem of insufficient description of the linkage between price generation process and agent behavior in traditional methods, this paper constructs a dynamic deduction method of market behavior of electricity price formation mechanism for electricity market simulation system, which integrates market data collection, scenario construction, feature representation, state modeling, multi-agent game and intelligent deduction mechanism into a unified computing framework. Based on Python 3.11 and PyTorch 2.2 platform, the experiment was carried out with continuous samples of 180 d and 15 min granularity. The results show that the root mean square error of the electricity price in this paper is 4.87 yuan /MWh, the average absolute percentage error is 3.96%, the price turning point capture rate is 88.4%, the market clearing rate is 94.6%, and the supply-demand deviation rate is 2.3%. The results show that the proposed method can better reveal the internal relationship between electricity price evolution and market behavior adjustment, and provide computational support for electricity market simulation analysis and rule optimization.
