摘要
为了减少化工事故的发生,提高化工生产过程中事故风险预测的准确性,研究了粒子群优化算法与支持向量机(PSO-SVM)模型在事故风险预测中的应用。首先,统计分析近5年化工生产安全事故致因因素,得出化工事故风险因素统计特征,结合层次分析法,建立化工事故风险预测指标体系并确定各指标因素的权重值;然后,基于MATLAB计算生成的化工事故风险程度样本数据,利用PSO算法优选SVM回归预测模型的惩罚因子和核函数参数,建立PSO-SVM相耦合的化工事故风险回归预测模型;最后,将预测指标值样本数据代入模型得到对应预测事故风险值。对比PSO-SVM模型预测风险值和实际计算风险值,可知PSO-SVM模型预测精度良好,预测结果与实际结果较为吻合,表明该模型能有效处理小样本数据回归预测问题,可解决化工生产安全系统各等级风险的异常样本数据稀少问题,模型适用于化工事故风险预测。
In order to reduce the occurrence of chemical accidents and improve the accuracy of accident risk prediction in the chemical production process,the application of particle swarm optimization algorithm and support vector machine(PSO-SVM)model in accident risk prediction is studied.First,statistical analysis of the causative factors of chemical safety production accidents in the past five years is carried out to obtain the statistical characteristics of risk factors.Combined with the analytic hierarchy process,the chemical accident risk prediction index system is established and determine the weight value of each index factor.Then,based on the sample data of chemical accident risk degree generated by MATLAB calculation,the PSO algorithm is used to optimize the penalty factor and kernel function parameters of the SVM regression prediction model,and the PSO-SVM coupled chemical accident risk regression prediction model is established.Finally,the sample data of the predicted index values are substituted into the model to obtain the corresponding predicted accident risk value.Comparing the predicted risk value of the PSO-SVM model with the actual calculated risk value,it can be seen that the prediction accuracy of the PSO-SVM model is good,and the predicted results are more consistent with the actual results.Due to the scarcity of abnormal sample data of various levels of risk,the model is suitable for chemical accident risk prediction.
作者
吴红
王晓明
王斌
顾翩
WU Hong;WANG Xiaoming;WANG Bin;GU Pian(Jiangsu Research Institute of Work Safety,Nanjing 210009)
出处
《工业安全与环保》
2022年第4期70-73,共4页
Industrial Safety and Environmental Protection
基金
江苏省社会发展重大科技示范项目(BE2020729)。
关键词
化工事故
风险预测
支持向量机回归模型
粒子群优化
chemical accident
risk prediction
support vector machine regression(SVR)model
particle swarm optimization
