作者
Xinke Li,Jianchun Yang,Yu Xiong,Kui Liang,王辉敏,Huarong Peng,W Yang
摘要
A dual-optical-path full-spectrum UV–Vis system was developed for multiparameter water-quality monitoring. The system comprises a dual-optical-path sensor, a spectral detection module, and an intelligent spectral analysis model. It enabled the simultaneous in situ determination of total nitrogen (TN), chemical oxygen demand (COD), and suspended solids (SS). By updating the reference spectrum in real time, the system compensated for light-source drift. With compact dimensions of 62 × 53 × 27 mm, the module operated over a wavelength range of 200–800 nm, with a resolution of 2.3-2.9 nm and a signal-to-noise ratio of 471:1. The baseline drift of the sensor is reduced by approximately 87% compared to a single optical path, and the COD absorbance–concentration calibration (at 254 nm) yielded an R² of 0.9987. The DGAtt-1DCNN spectral intelligent analysis model was constructed, integrating a concentration domain weighting mechanism to improve the prediction accuracy in the low-concentration range. The MAPE values for TN, COD, and SS were 6.88%, 3.04%, and 6.07%, respectively. Laboratory test results show that the zero-point drift, range drift, and repeatability of the system are all less than 2%, and the maximum relative errors for TN, COD, and SS are 8.88%, 4.70%, and 7.93%, respectively. For thermal-power-plant cooling water (n = 6) and simulated circulating-water discharge from a nuclear power plant (n = 4), the maximum relative errors for all three parameters were below 10%.