BP神经网络多波长法的COD预测  

COD prediction based on BP neural networks with multi-wavelength method

作　　者：温林强 夏凤毅[1] 沈洲[1] 

机构地区：[1]中国计量大学质量与安全工程学院浙江省家具检测技术重点实验室,浙江杭州310018

出　　处：《中国计量学院学报》2016年第3期306-312,共7页Journal of China Jiliang University

摘　　要：提出了一种紫外多波长与BP神经网络相结合的有机废水COD检测技术.由单片机(single chip microcmputer,SCM)控制GSM(global system for mobile communication)系统,采集的数据以串行通信方式发送到PC机上.由BP算法对超出单波长法检测区间的COD值进行预测.分别从15条光谱扫描曲线中提取8组不同波长对应的吸光度数据作为算法的输入,调整算法的有关参数,并进行一定次数的训练.对最终预测结果进行了误差分析,数据显示相对误差控制在5%以内,预测结果稳定.An automatic organic wastewater COD test method was proposed by using the UV multi-wavelength and the BP neural network technology.The GSM system was controlled by SCM to send the collected data to PC in the way of serial communication.The COD value was predicted by using the BP algorithm.Eight absorbency data extracted from 15 spectral scanning curves were used as the inputs of the BP algorithm to build the prediction model.The error analysis showed that the relative error was less than 5%.

关 键 词：紫外多波长技术 特征单波长法 COD检测 GSM系统 BP神经网络 

分 类 号：X832[环境科学与工程—环境工程]