基于随机森林的LNG场站泄漏风险评价模型研究

Research on the Evaluation Model on the Risk of Leakage in LNG Stations Based on Random Forest Classification

通过对国内外LNG场站泄漏事故进行统计和分析,结合相关行业专家研究,总结出一套LNG场站泄漏风险评价体系。该风险评价体系主要包括物质状态、设备设施、作业现场、安全管理和应急救援5个一阶评价指标和20个二阶评价指标。本研究以多个LNG储运企业的多位专家评估结果为数据样本,结合随机森林算法对泄漏风险评价体系进行机器学习,分析得出一套针对LNG场站的泄漏风险评价方法。该方法可以运用于LNG场站进行风险管控自查,对LNG储运产业提高自身安全管理,提高智能化管理有极大的推动作用。

This research summarizes an evaluation system on the risk of leakage for LNG stations,based on the statistics of the relevant accidents and relevant industries researches at home and board.The system mainly contains 5 first-order indexes,including material condition,equipment and facilities,workplace,safety management and emergency rescue and 20 second-order indexes.This research also summarizes an evaluation method,using random forest classification to machine learning based on multi-expert evaluation results of multiple LNG storage and transportation enterprises.This method can be applied to LNG stations to conduct risk self-assessment,and can promote the improvement of the safety management and intelligent management of the LNG storage and transportation industries.