基于人工智能预警天然气净化装置非正常工况

Application of artificial intelligence to early warning of abnormal situation on natural gas purification plant

天然气净化装置非正常工况现有预警方式,主要为工业控制系统单参数阈值报警,操作人员的技能经验决定了处置的时效性与适宜性,受人为因素影响较大。为此,寻求一种在非正常工况出现苗头、早于工业控制系统阈值报警的预警技术,降低操作人员技能水平对故障判断的影响,推进安全管理关口前移尤为重要。本文以天然气净化装置脱硫溶液系统发泡为例,结合脱硫溶液发泡的形成原因、参数表征、操作处置等现场经验,结合净化装置工业控制系统实时读取、历史查询生产数据的数据库,开展了基于人工智能建立的脱硫溶液发泡预警模型的研究、设计与应用。运行结果表明,模型预警准确率可达97%,在预警净化装置溶液发泡非正常工况中是可行的,并对其它非正常工况的预警研究具有可复用与迁移的重要意义。

The existing early warning of abnormal situation on natural gas purification plant is mainly industrial control system based on single parameter threshold,which is affected by the skill and experience of operators on problem-solving.Therefore finding a new warning of abnormal situation technology which could set alarm earlier than the existed methods and reducing the influence of human factors to guarantee safety in production are of vital importance.In this paper,we design a model of early warning of abnormal situation on foaming of desulfurization solution of natural gas purification plant based on the practical experience and production database.The experimental results show that the accuracy of our proposed model is up to 97%.The model is practical when facing the abnormal situation of natural gas purification plant and is significant to other abnormal situations.