基于遗传算法优化BP神经网络的曝气量预测  

Prediction of Aeration Quantity Based on BP Neural Network Optimized by Genetic Algorithm

作　　者：唐维 陶钰欣 郝启文 庞中华[3] TANG Wei;TAO Yuxin;HAO Qiwen;PANG Zhonghua(Yantai Urban Drainage Service Center,Yantai 264000,China;Beijing KingTrol Data Technology Co.,Ltd.,Beijing 100160,China;Key Laboratory of Fieldbus Technology and Automation of Beijing,North China University of Technology,Beijing 100144)

机构地区：[1]烟台市城市排水服务中心,山东烟台264000 [2]北京金控数据技术股份有限公司,北京100160 [3]北方工业大学现场总线技术及自动化北京市重点实验室,北京100144

出　　处：《控制工程》2024年第10期1746-1752,共7页Control Engineering of China

基　　金：“十三五”水体污染控制与治理科技重大专项项目(2018ZX07105003);国家自然科学基金资助项目(62173002);北京市自然科学基金资助项目(4222045)。

摘　　要：针对城市污水处理厂生化池曝气量的冗余问题,从节能降耗角度出发,提出了一种精确预测生化池曝气量的方法。首先,对从山东某污水厂采集的数据做预处理;其次,经相关性分析,建立以进出水指标和历史曝气量为输入、以当前曝气量为输出的反向传播(back propagation,BP)神经网络模型,并利用遗传算法优化BP神经网络的权值和阈值;最后,为了验证所提方法的有效性,将其与传统BP神经网络模型进行对比。实验结果表明,所提方法的均方根误差减小了39.88%,计算时间缩减了31.00%;且与实际测试数据相比,每月的曝气电费可节约892.80元。因此,所提方法对生化池曝气量的预测是有效和可行的。In order to solve the redundancy problem of aeration quantity of biochemical tank in urban wastewater treatment plant,a method for accurately predicting aeration quantity of biochemical tank is proposed from the viewpoint of saving energy and reducing consumption.Firstly,the data collected by a wastewater treatment plant in Shandong Province are pretreated.Secondly,through correlation analysis,back propagation(BP)neural network model is established,of which the inputs are the inlet and outlet water indices and the historical aeration quantity,and the output is the current aeration quantity.Furthermore,the genetic algorithm is used to optimize the initial weights and thresholds of the BP neural network.Finally,in order to verify the effectiveness of the proposed method,it is compared with the conventional BP neural network model.Experimental results show that the root mean square error of the proposed method is reduced by 39.88%and the calculation time is reduced by 31.00%.Therefore,compared with the actual test data,the monthly aeration electricity cost can be saved 892.80 yuan.Therefore,the proposed method is effective and feasible to predict the aeration quantity of biochemical tank.

关 键 词：污水处理 精确曝气 神经网络 遗传算法 预测 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]