Laboratory-based research on microbial fuel cells (MFCs) is often costly and limited to a small number of variables, making optimization challenging. However, machine learning (ML) offers a promising solution by enabling efficient multivariate principal component analysis (PCA) and multivariable optimization. These techniques can provide significant insights and optimization opportunities. The goal of this study is to propose an ML-based approach to explore the relationships between bioelectricity generation (in terms of voltage, power density (PD), current density (CD), and coulombic efficiency (CE)) and two key variables, chemical oxygen demand (COD) and pH, as well as to recommend their optimal combinations. Specifically, the objectives are to (1) integrate a laboratory-based MFC study with multivariate data analyses; (2) apply PCA to reduce data complexity by focusing on the principal components that account for the greatest variance, thus improving interpretability; and (3) identify the optimal combinations of COD and pH for maximizing bioelectricity generation. The PCA results demonstrated that COD positively influenced the generated voltage while having an inverse effect on CE. Additionally, both PD and CD increased with higher pH values. The optimal combination of COD and pH improved CD, PD, and CE; however, their optimal combination for generated voltage differed, with higher COD leading to higher voltage. The optimal predicted voltage, CD, PD, and CE of the study were 795.71 (mV), 1451.80 (mA/m2), 57.46 (mW/m2), and 4.85%, respectively. By integrating ML approaches, this study contributed to the optimization of bioelectricity generation from wastewater and offered valuable insights for researchers working in this field.
Khanaum et al. (Tue,) studied this question.