Uncertainty Analysis Method for Petroleum System Modeling Based on SurrogateModel to Improve Thermal Maturity Evaluation

Quantitative evaluation of source rock thermal maturity is a core component of petroleum system analysis. Traditional deterministic modeling methods are affected by geological parameter uncertainties, making it difficult to meet the demands of shale oil and gas sweet spot prediction. Taking the Permian source rocks of the Junggar Basin as the research object, this paper constructs a thermal evolution surrogate model integrating 3D convolution and spatial attention mechanism, and adopts the Unscented Kalman Inversion method to realize key parameter inversion and uncertainty quantification. Thermal history uncertainty propagation is accomplished via Monte Carlo sampling. Simulation results show that this method can efficiently improve the accuracy of source rock thermal maturity evaluation. The Permian source rocks have generally entered the oil generation stage, with the sag centers reaching the high-maturity gas generation stage, which is consistent with drilling measured data. This method provides a new paradigm for uncertainty modeling of petroliferous basins and has important guiding significance for hydrocarbon exploration.