基于RF-PSO-SVM的油页岩干馏工艺粉尘爆炸风险评估
Dust Explosion Risk Assessment of Oil Shale Retorting Process Based on RF-PSO-SVM
于立富 1李航 1王天枢 1孙威2
作者信息
- 1. 沈阳化工大学 环境与安全工程学院,辽宁 沈阳 110142
- 2. 沈阳化工大学 化学工程学院,辽宁 沈阳 110142
- 折叠
摘要
为准确预测油页岩干馏工艺过程粉尘爆炸风险等级,以加强油页岩粉尘爆炸事故防范能力,提出了一种快速精准的风险评估模型.按照 4M 分类原则将评价指标分为人、物、管理和环境 4 大类和 30 小项,采用随机森林(RF)对 30 项特征指标进行属性约简,进而提取关键指标;使用粒子群算法(PSO)对支持向量机(SVM)进行更新全局寻优,合理优化SVM的参数.通过随机选择 30 组评价数据进行测试,进行了RF-PSO-SVM模型与SVM模型、RF-SVM模型以及PSO-SVM模型对比.结果表明:该模型风险预测结果正确率最高且运行时间较短,识别准确率达 93.33%,体现出该模型对油页岩干馏工艺粉尘爆炸风险预测的精准性和及时性.
Abstract
In order to accurately predict the dust explosion risk class of oil shale retorting process and strengthen the prevention ability of oil shale dust explosion accident,a fast and accurate risk assessment model was proposed.According to the 4M classification principle,the evaluation indexes were divided into 4 categories including human,material,management and environment and 30 items.The random forest(RF)was used to reduce the attributes of the 30 feature indexes,and then the key indexes were extracted.The particle swarm optimization algorithm(PSO)was used to update the global optimization of support vector machine(SVM)and optimize the parameters of SVM reasonably.By randomly selecting 30 groups of evaluation data for testing,the RF-PSO-SVM model was compared with SVM model,RF-SVM model and PSO-SVM model.The results showed that,the risk prediction result of this model had the highest accuracy and short running time,with the recognition accuracy of 93.33%,which reflected the accuracy and timeliness of this model in predicting dust explosion risk of oil shale retorting process.
关键词
油页岩干馏/随机森林/支持向量机/风险评估Key words
Oil shale retorting/Random forest/Support vector machine/Risk assessment引用本文复制引用
基金项目
辽宁省教育厅科学研究经费资助项目(LQ2020024)
出版年
2024
NETL
NSTL
万方数据