基于PSO-BP神经网络的气溶胶质量浓度测量系统湿度补偿  

The Humidity Compensation for Measurement Systems of Aerosol Mass Concentrations Based on the PSO-BP Neural Network

作　　者：张加宏[1,2,3] 刘毅[1,2,3] 顾芳[4] 沈雷[3] 冒晓莉[2,3] 吴佳伟[3] 汪程[3] 包志伟[3] 

机构地区：[1]南京信息工程大学江苏省大气环境与装备技术协同创新中心,南京210044 [2]南京信息工程大学江苏省气象探测与信息处理重点实验室,南京210044 [3]南京信息工程大学电子与信息工程学院,南京210044 [4]南京信息工程大学物理与光电工程学院,南京210044

出　　处：《传感技术学报》2017年第3期360-367,共8页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(61307113;61306138;41605120);江苏省自然科学基金项目(BK2012460);江苏省高等学校大学生实践创新训练计划项目(201510300034);江苏高校品牌专业建设工程资助项目(TAPP)

摘　　要：气溶胶质量浓度是衡量大气质量的一项关键性指标。在基于单粒子光散射的气溶胶质量浓度测量时,环境湿度变化对其影响较大。尤其在湿度较高时,颗粒物的外貌特征和折射率都会发生相应变化,采用低湿度条件下系统的标定参数来反演气溶胶质量浓度会产生较大误差。考虑以上原因,提出建立基于粒子群优化的BP神经网络补偿模型对不同湿度条件下的测量结果进行数据融合修正。实验结果表明,经过PSO-BP神经网络湿度补偿后,相对湿度较高时引起的测量误差约从原来的-10%^-45%减小为-5%^-30%,整体平均相对误差减小了10%,说明该方法削弱了相对湿度对气溶胶质量浓度测量系统的影响,有效提高了系统的测量精度。Aerosol mass concentration is a key indicator of air quality. The influence of the changes in ambient hu- midity on aerosol mass concentration measurement based on single particle light scattering is large. Especially, phys- ical characteristics and refractive index of particulate matter could change accordingly in high humidity, which uses system calibration parameters under low humidity conditions to invert aerosol mass concentration and leads to con- siderable error. Therefore, considering the above reasons, a compensation model is presented by particle swarm opti- mization BP neural network for data fusion of the measurement data. The experimental measurements show that after humidity compensation based on the PSO-BP neural network, the measurement error caused by high relative humidi- ty decreases from the original approximately -10%-45% to -5%- 30%, and the overall average relative error is re- duced by 10%,which indicates that the PSO-BP method can reduce relative humidity effect on the aerosol mass concentration measurement system and improves system measurement accuracy effectively.

关 键 词：气溶胶质量浓度 光散射 湿度补偿 PSO-BP神经网络 

分 类 号：TN247[电子电信—物理电子学]