An Integrated Convolutional and Long Short Term Memory Network Approach for Predicting Electricity from Microbial Fuel Cells

Consumption of fossil fuels causes an increase in pollution which is also a major cause of global warming. There is a growing demand for cleaner and renewable energy sources due to world's pollution growth. Researchers are seeking for alternate sources of electricity for industrial and home uses since electric motors and other fuel-free alternatives have grown in popularity. Treatment of waste water is one of the major concerns in India. Microbial fuel cells (MFC's), which directly convert organic matter in wastewater to electric energy without the need for intermediary stages, have recently been the subject of substantial research. The biodegradation efficiency of organic matter and the electron transfer efficiency have a significant impact on the generation of bioelectric energy in MFC's. Many empirical and mathematical models are used to predict the electricity generation from MFC's. This paper describes the development of a unique deep learning model to relate the voltage generation in MFC with input chemical oxygen demand (COD) of wastewater. A model to predict voltage and current corresponding to changes in COD, pH, total suspended solids (TSS) has been proposed. The model developed by employing convolutional layers and long short-term memory layers has been able to predict voltage and current with considerably low error values resulting from an exhaustive training process. The model has proved to be a good fit on experimental data making it reliable for predicting electricity from small-scale microbial fuel cell setups in educational institutions and laboratories.