基于改进型遗传算法的危险气体多泄漏源定位研究

Locating Multiple Hazardous Gas Leak Sources Using Modified Genetic Algorithm

危险气体多源泄漏会造成严重的安全事故,故针对多个危险气体泄漏源进行源强反算,提出改进型遗传算法(MGA),即引入淘汰个体交配机制增加算法的全局搜索能力,并在算法成熟阶段引入最优个体追随策略以加速算法的收敛。采用MGA进行开放空间1~5个泄漏源的定位能力测试,分析泄漏源相对位置和测量随机误差对算法预测能力的影响。结果显示:MGA可以定位处于垂直风向分布且相对距离大于2 m的两个泄漏源;MGA无法准确定位沿风向分布、扩散浓度场存在包含关系的两个泄漏源;测量随机误差的增加会导致算法预测的泄漏源位置偏差增大,而对泄漏源质量流量的影响不明显。

Hazardous gas leaking accidents may cause disastrous consequences.Therefore,the source strength is calculated for multiple dangerous gas leakage sources.The current study proposed a modified genetic algorithm(MGA).The modification includes two aspects:providing crossover chances to eliminated individuals,in order to increase the global searching ability;leading particles following the best individual when the algorithm is about to complete,therefore accelerating the convergence.MGA was used to identify multiple sources in hazardous gas spreading fields containing 1~5 leaking sources.Leaking sources with different relative locations are located using MGA.MGA can accurately identify two sources located vertical to the wind direction when their distance decreases to 2 m.However,when two sources locate along the wind direction resulting in badly overlapped spreading fields,MGA cannot accurately predict the source locating downstream.The location prediction error by MGA increases with increasing measurement noise.While the leaking mass flow rate prediction error bounces randomly with increasing measurement noise.