An energy-autonomous chemical oxygen demand sensor using a microbial fuel cell and embedded machine learning

Abstract

The current methods of water quality monitoring tend to be costly, labor-intensive, and off-site. Also, they are not energetically sustainable and often require environmentally damaging power sources such as batteries. Microbial fuel cell (MFC) technology is a promising sustainable alternative to combat these issues due to its low cost, eco-friendly energy generation, and bio-sensing features. Extensive work has been done on using MFCs as bio-sensors or sources of power separately. However, little work has been done toward using MFCs for both applications at the same time. Additionally, previous studies using MFCs for water quality measurement have been mostly limited to laboratory conditions due to the biochemical complexity of the real-world. Another limitation of MFCs is how little power they can generate, requiring the MFC-based systems to have minimal power consumption. This work addresses these challenges and presents an energy-autonomous water quality sensing unit that uses a single MFC both as its sensory input and the sole source of power for computing the chemical oxygen demand (COD). In the proposed unit, geometric features of the voltage profile of the MFC (e.g., peak heights) are used as the inputs to a machine learning algorithm (support vector regression (SVR)). The electrical power generated by the MFC is used to drive a low-power microcontroller which logs the MFC voltage and runs the machine learning algorithm. Experimental evaluation showed that the device is capable of detecting the COD of natural pond water samples accurately (coefficient of determination (R2) = 0.94). This work is the first demonstration of energy autonomy in an MFC-based sensing unit for measuring water quality and represents a step forward in the development of energy-autonomous sensors for environmental monitoring applications.