复杂水质环境下基于PSO-ELM的BOD快速检测

Rapid Detection of BOD Based on PSO-ELM in Complex Water Quality Environment

针对检测生化需氧量(BOD)的传统五日培养法(BOD5)操作复杂、时效性差的不足,以及水质中复杂环境因素干扰检测过程等问题,提出了基于微生物膜法的快速检测系统,进而以粒子群算法(PSO)优化的极限学习机(ELM)算法来实现BOD检测。检测系统以溶解氧传感器和微生物膜反应器为核心,能够在35 min内完成检测,其中微生物反应器使用功能化的螺旋玻璃管制成,但微生物膜易受水质复杂环境的影响。为此,运用PSO-ELM算法消除水质中浊度(SS)、p H值、氧化还原电位(ORP)对检测结果的干扰,与BP神经网络和ELM算法相比,运行时间分别缩短0.92 s和0.24 s,测试误差分别减小5.3%和4.0%。在实际海水水样的测试结果中,该方法与BOD5法相对误差保持在2.69%~3.86%内。

In order to overcome the shortcoming of complex operation,poor time-efficiency of traditional five-day culture method( BOD5),and the influence of complex environmental factors on the detection process in actual water quality,a detection method of BOD combining the rapid detection system and the extreme learning machine( ELM)algorithm optimized by particle swarm optimization( PSO) was proposed. The detection system was centered on a dissolved oxygen sensor and a microbial membrane reactor,which can perform the test within 35 minutes. The microbial reactor was made of functionalized spiral glass tubes,but the microbial membrane was affected by the environment of water quality. For this reason,PSO-ELM algorithm was used to eliminate the influence of turbidity( SS),p H and REDOX potential( ORP).Compared with BP neural network and ELM algorithm,the running time is shortened by 0. 92 s and 0. 24 s respectively,and the test error is reduced by 5. 3% and 4. 0% respectively. In the test of actual seawater samples,the relative error varies from 2. 69% to 3. 86%.