Research On Determination Of Water Quality Parameters Based On Voltammetric Sensor Array And Deep Learning Methods

Rapid and accurate detection of surface water parameters is a crucial step in water quality assessment.In our country,various water quality indicators are used to evaluate the quality of surface water,among which Chemical Oxygen Demand(COD_Mn),Nitrate（NO₃^--N）,Dissolved Inorganic Phosphorus（DIP）,and Ammonia Nitrogen（AN）are the four fundamental water quality parameters.Currently,the detection methods for water quality parameters require sample pre-treatment,complex operational procedures,and large-scale equipment.These factors hinder the implementation of high temporal and spatial resolution measurement of water parameters.In contrast,electrochemical analysis offers a convenient and rapid detection method.A voltammetric sensor array composed of silk-screen printed carbon electrodes（SPCE）modified with gold nanoparticles（Au NPs）,silver nano-dendritic structures（Ag NDs）,and reduced graphene oxide（r GO）is constructed in this study.The sensor array exhibits good cross-sensitivity in voltammetric measurements of water quality indicators.To overcome the limitations caused by peak overlap due to similar oxidation-reduction potentials and matrix interference,a Convolutional Transformer model（FCv T）is proposed.FCv T combines convolutional operators and self-attention operators in a loosely coupled manner.The model captures local peak patterns while dynamically fusing feature vectors within a global range using dynamic weights.By integrating local and global complementary features,the model overcomes the constraints of peak overlap and enables the quantification of component concentrations in complex solutions.Subsequently,to meet the requirements of on-site detection for water quality parameter monitoring,a portable multi-channel electrochemical detection device with good interaction and data sharing capabilities is designed.The device can perform cyclic voltammetry or differential pulse voltammetry,and it features a user-friendly graphical interface,four measurement channels,and network accessibility.The device achieves high-resolution measurements at low current levels.A data analysis platform is built around the device,where voltammetric data preprocessing and multivariate analysis models run in the cloud.The final results are visualized on a web page,providing remote access to on-site detection results.In both the self-prepared mixed solution dataset and real water sample tests,the framework combining voltammetric sensor array and deep learning demonstrates excellent performance.The results indicate that the developed multi-voltammetric sensor system,the convolutional Transformer model,and the portable electrochemical detection device have broad application potential for rapid and accurate analysis of surface water quality parameters.