Research on Drainage Network Functional Diagnosis Based on Online Water Quality Monitoring and ANN Correction

At present, the low organic matter concentration in the influent of municipal wastewater treatment plants is a widespread issue in China, significantly impairing nitrogen and phosphorus removal efficiency, particularly pronounced in southern cities. Sewer network defects are identified as one of the critical causes contributing to this problem, making scientific investigation of network operational conditions essential for effective remediation. This study focused on the sewer network in a district of Maoming City, Guangdong Province. Chemical oxygen demand (COD) monitoring equipment and Doppler flowmeters were deployed at 10 key nodes within a 27 km network system to conduct continuous monitoring for 3 days and manual sampling and analysis four times daily. An artificial neural network data correction model based on relevant factor screening was constructed and compared with linear regression, traditional artificial neuron network (ANN), support vector regression, and long short-term memory network models at the methodological level. After correction, the Nash-Sutcliffe efficiency coefficient reached 0.94, the linear correlation coefficient reached 0.96, and the relative error of most monitoring points was controlled within 15%, verifying the applicability and reliability of this technical method in network water quality monitoring. Through temporal longitudinal and spatial horizontal comparative analysis, the COD detection values of the main sewer showed a continuous low-concentration characteristic from upstream to downstream, with small concentration fluctuations across all time periods, and the final influent concentration to the plant was below 80 mg/L. Abnormal flow rates and elevated water levels were observed in the river-crossing pipeline section. From Point 8 onwards, the manholes showed full-pipe and full-load conditions under dry weather. Combined with field conditions and comprehensive analysis, it is indicated that clean water inflow exists in the upstream pipe section, and the river-crossing pipeline may suffer from blockage-induced sewage overflow and river water backflow caused by high head potential in the river channel. Subsequent targeted remediation plans should be developed based on field verification and engineering investigation.