摘要
针对污水处理过程中曝气池溶解氧浓度无法精确在线测量的问题,本文采用BP神经网络建立了溶解氧浓度预测的软测量模型。将进水参数氨和铵根离子态的氮Snh、快速可生物降解有机物Ss、异养菌生物量Xbh、颗粒性不可生物降解有机物Xi、慢速可生物降解有机物Xs以及进水流量Q作为BP神经网络软测量模型的输入变量,采用遗传算法对BP神经网络的初始连接权值和阈值进行优化。对预测结果的准确性及遗传算法优化BP神经网络的泛化能力进行了分析,讨论了数据归一化对软测量模型预测结果的影响。仿真结果表明,采用遗传算法优化BP神经网络的权值和阈值以及对训练数据归一化处理,有效地解决了溶解氧浓度BP软测量模型精度差的问题,使溶解氧软测量模型的测量精度明显增强。
In view of the aeration tank in the sewage treatment process of dissolved oxygen concentration is not precise online measurement of the disadvantages, this paper adopts the soft measurement of BP neural network to establish the prediction model of dissolved oxygen concentration. The Snh of the influent nitrogen, ammonia and ammonium ion state parameters rapidly biodegradable organic matter Ss, bacteria biomass of heterotrophic Xbh, particulate unbiodegradable organic compounds Xi, slowly biodegradable organic compounds Xs and flow rate of Q as BP neural network input variables of soft sensor model, were optimized by genetic initial algorithm of BP neural network connection weights and threshold. The accuracy of the prediction results and the genetic algorithm to optimize BP neural network generalization ability is analyzed, the effect of data normalization on soft measurement model to predict the results of the discussion. The simulation results show that, using genetic algorithm to optimize BP neural network weight and threshold value and the training data is normalized, effectively solve the dissolved oxygen concentration in BP soft measurement model accuracy, the measurement accuracy of the dissolved oxygen in the soft measurement model of enhanced obviously.
出处
《计算机与应用化学》
CAS
2016年第1期117-121,共5页
Computers and Applied Chemistry
基金
国家自然科学基金资助项目(61563032
51165024)
甘肃省自然科学基金资助项目(145RJZ024
145RJYA313)
关键词
溶解氧浓度
遗传算法
神经网络
软测量
数据归一化
dissolved oxygen concentration
genetic algorithm
neural network
soft sensor
data normalization
