Fast Deployable Real-Time Bioelectric Dissolved Oxygen Sensor Based on a Multi-Source Data Fusion Approach

Abstract

Commercially available dissolved oxygen (DO) sensors are hardly suitable for stereoscopic and precise DO monitoring due to their design, high cost and susceptibility to matrix effects. Here, we have developed a DO biosensor based on an integrated chamber-free microbial fuel cell (DOMFC) as the core and a Raspberry Pi microcomputer as the data acquisition system. This biosensor is low cost, readily available and compact in configuration. To this end, stable microbial biofilms with oxygen gradients were established on bioaffinity aluminium foam as anode. The DOMFC sensor has a low internal resistance (9.62 Ω) that can respond to DO changes in less than one minute and produce a reliable voltage signal to record DO (0.15-9.5 mg/L) under challenging conditions. After training the GA-BPNN model with multidimensional data by automatically applying a data fusion strategy from multiple sources, accurate DO predictions (R2 = 0.997, RMSE = 0.0447, MAE = 0.0401) were obtained. The DOMFC sensor and the prediction model showed excellent agreement (R2 = 0.954) in complex natural applications (different pH values, conductivities, water temperatures, etc.), covering a wide range of applications. Since the sensor mini-monitoring system is inexpensive and easy to make and use on a large scale, it is a promising alternative for oxygen measurements in both natural and artificial waters.

Keywords: Dissolved oxygen, biosensor, integrated microbial fuel cell, GA-BPNN, multi-source data fusion