基于PSO-BP神经网络的气溶胶质量浓度测量系统湿度补偿

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.

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

Aerosol mass concentration is a key indicator of air quality.The influence of the changes in ambient humidity on aerosol mass concentration measurement based on single particle light scattering is large.Especially,physical 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 considerable error.Therefore,considering the above reasons,a compensation model is presented by particle swarm optimization 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 humidity decreases from the original approximately-10%~45% to-5%~30%,and the overall average relative error is reduced 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.