Objective This study aims to overcome the limitations of traditional static Bayesian networks (BN) and reveal the nonlinear, time-varying dynamic causal relationships among internal and external factors affecting China's grain production security, monitor key risk sources influencing China's food security, shift the focus of grain production security risk management from "post-event response" to "pre-event warning", and provide decision-making references for further safeguarding national food security.

Method Static and dynamic factors affecting China's grain production security, such as labor force, agricultural infrastructure, water resource, land resource, extreme climate, material input, finance, and frequent major emergencies, were systematically integrated into a unified multi-dimensional dynamic Bayesian network (DBN) model. Eight first-level indicators and twenty secondary indicators were established to construct a risk monitoring index system for China's grain production security. Corresponding risk indicators were used as DBN nodes to develop a DBN-based risk monitoring and early warning model for China's grain production security. Using inter-provincial panel data from 2000 to 2024, the GeNIe 2.0 visualization software was employed for empirical test and optimization of the DBN model. The risk probability of grain production security in China was measured, key risk-inducing factors and sensitive points were diagnosed, and early warning levels for various risk sources were precisely defined.

Result Empirical test verified that risk sources such as quantity of labor force, agricultural machinery level, effective irrigation index, grain sown area, pesticide input, and fertilizer price fluctuation are key risk factors and important sensitive nodes for China's grain production security. In the risk warning level classification, five indicators—water infrastructure, agricultural water security level, cultivated land area, grain sown area, and financial support for agriculture funds—achieved risk values of 0.368 or higher, corresponding to the "black light" risk level. Six indicators—quantity of labor force, power infrastructure, per capital share of water resource, affected area of crops, fertilizer input, and plastic film input—showed risk values between 0.330 and 0.356, falling under the "red light" risk level. Reverse reasoning results indicated that financial factor, land resource factor, and water resource factor are the most significant risk-inducing contributors to overall risk. Sensitivity analysis revealed that frequent major emergencies and quantity of labor force have particularly significant impacts on the risk system. Empirical test demonstrated the feasibility and effectiveness of the model and algorithm.

Conclusion Through empirical test of the model and algorithm, key risk factors affecting China's grain production security were accurately identified, and potential hazard sources were revealed. This provides a decision-making basis for promoting the high-quality development of China's grain industry and effectively preventing grain production security risks.