Considering the high volatility and strong nonlinear characteristics of China’s regional electricity spot market

how to effectively manage electricity price risk has become a common concern among electricity market participants.This paper takes the Shanxi electricity spot market as the case study object and focuses on the design of Asian options for the delivery month spot average price.It proposes an improved Monte Carlo pricing method that integrates deep neural network prediction mechanisms.Addressing the inadequacies of traditional stochastic differential equations paths in simulating electricity prices

we propose a path generation approach under the Monte Carlo framework that combines forecasting with perturbation:First

a deep neural network (Transformer) generates the entire benchmark price sequence.Then

mean-reversion

heavy-tailed noise

and sparse jumps are superimposed to preserve spikes and heteroskedasticity

forming price paths that are both realistic and uncertain.Under risk-neutral measures

option prices are derived by discounting the payoffs of numerous simulated paths.Based on empirical data from the Shanxi market

this paper evaluates model inputs and path quality using multi-step forecast error and stochastic path interval metrics.Results demonstrate that this method enhances overall coverage and convergence performance without widening intervals

while making pricing outcomes more consistent with the risk characteristics of options based on the average price of the delivery month.This research provides an actionable technical approach for designing option products and localizing pricing in China’s regional electricity markets.