基于EM-FR-C5.0DT耦合模型的输气管道地质灾害风险预测模型

Research on Geological Hazard Prediction of Gas Pipeline Based on EM-FR-C5.0DT Coupling Model

延安气田地处陕北山区,输气管道沿线发生地质灾害的风险较高,管道生产运行存在一定安全隐患,通过加强风险预测研究,可快速准确甄别沿线高后果区,对管道防灾减灾具有重要意义。为此,选取延安气田内部临镇-子长输气干线作为研究对象。首先,通过相关性分析筛选出高程等11个影响因子,依次开展灾点空间分布规律研究;其次,采用加权频率比法将灾点属性值转换为可体现灾害风险贡献率的EM-FR(加权频率值),划分出低、极低风险区,在此范围内选取非灾点,以此构建EM-FR-C5.0DT(加权频率比-C5.0决策树)、EM-FR-BP(加权频率比-BP神经网络)等2种耦合模型,并预测研究区域的风险性;最后,在研究区域内随机选取非灾点,构建单一C5.0DT、BP模型,并与上述2种耦合模型开展精度对比分析。结果显示:耦合模型预测性能优于单一模型,其中EM-FR-C5.0DT模型效果最优。研究成果表明,在低、极低风险区内,选取非灾点构建数据集得到的耦合模型,可明显提升模型预测精度,更适合小样本地质灾害风险性建模,可为延安气田输气管道风险性研究提供一定借鉴。

Yan’an gas field is located in the mountainous area of northern Shaanxi Province and the risk of geological disasters along the gas pipeline is high,which has certain safety risks for the normal transportation of the pipeline.The risk prediction can quickly and accurately identify the areas with high consequences,which is of great significance for disaster prevention and mitigation along the pipeline.Therefore,this paper selects Linzhen-Zichang gas transmission line in Yan'an gas field as the research area,selects 11 influencing factors such as elevation,and aspect to study the spatial distribution of disaster points,converts the disaster point attribute value to the weighted frequency value(EM-FR)that can reflect the contribution rate of disaster risks by the weighted frequency ratio method,divides the low and very low risk areas,and chooses non-disaster sites to construct a coupled model of weighted frequency-ratio C5.0 decision tree(EM-FR-C5.0DT)and weighted frequency-ratio BP neural network(EM-FR-BP),predicts the risk of the study area,builds a single C5.0DT and BP model by selecting non-disaster points in the study area,and carries out a precision comparison study with the above coupled models.The results show that the predictive performance of the coupled model is better than that of the single model,EM-FR-C5.0DT has the best effect.It also shows that the coupling model obtained by selecting non-disaster points to build data sets in low and very low risk areas can significantly improve the prediction accuracy of the model,and is more suitable for the risk modeling of small samples of geological disasters,which can provide some reference for the risk research of gas pipeline in Yan'an gas field.