This study reveals the dynamic relationship between extreme weather and agricultural market fluctuations by constructing an econometric model. It uses the Ordinary Least Squares (OLS) method to conduct linear and nonlinear dimensional analyses, and integrates artificial intelligence technologies to improve prediction accuracy. The study finds that linear models are suitable for describing the cumulative effects of conventional disasters, while nonlinear extensions can capture the abrupt change characteristics in extreme scenarios. Meanwhile, by combining data preprocessing and extreme event identification with artificial intelligence, it proposes an approach to provide quantifiable decision-making tools for agricultural financial risk management and establish a full-chain risk prevention and control system of "monitoring-prediction-hedging".