Online Scenario Simulation and Customized Generation Method for Disaster Emergency Plans

Objectives: Rapidly evolving and multi-hazard natural disasters pose increasing challenges to traditional emergency planning systems, which are typically static and weakly coupled with real-time situational awareness. Existing approaches lack a unified mechanism to transform dynamic disaster states into structured and executable emergency plans. This study aims to develop an online scenario simulation and emergency plan generation framework that integrates disaster evolution modeling, knowledge graph reasoning, and multi-source data fusion. Methods: The proposed framework consists of four components. (1) A hierarchical disaster scenario simulation model is constructed, including disaster event, scenario, scenario element, and elemental state, which discretizes continuous disaster evolution into computable state transitions. (2) A natural disaster emergency knowledge graph integrates heterogeneous data such as emergency plans, historical cases, spatial data, and model outputs to enable semantic reasoning and knowledge fusion. (3) A scenario-driven emergency plan generation mechanism maps simulation outputs to standardized emergency response levels and dynamically fills structured plan templates. (4) An online system architecture is implemented using a microservice design, Spring Boot backend, Vue.js and CesiumJS frontend, supported by PostgreSQL, Neo4j databases, enabling continuous scenario updates and intermediate-state initialization. Results: A case study of the “3.30” Xichang forest fire was conducted. The system successfully reconstructed multi-stage disaster evolution, including ignition, initial disposal, full suppression, and final containment. Scenario outputs were dynamically aligned with emergency response levels to trigger corresponding plan templates. The generated plans include structured components such as command organization, risk zoning, evacuation strategies, key target protection, and rescue task allocation, with embedded spatial visualization. The end-to-end processing time from data input to plan generation is approximately 80 seconds, demonstrating near real-time decision support capability. Conclusions: This study proposes an integrated online framework for disaster scenario simulation and emergency plan generation. By coupling scenario simulation, knowledge graph reasoning, and template-based plan synthesis, it enables structured transformation from dynamic disaster states to actionable emergency responses. Although the current system is primarily designed for decision support and does not yet fully implement autonomous learning, it provides a scalable foundation for intelligent emergency management. Future work will focus on real-time sensor integration, adaptive learning mechanisms, and cross-disaster generalization.