Estimation of Biochemical Oxygen Demand with Artificial Neural Networks Optimized by Hybrid BBO-GA: Application at Bouira Wastewater Treatment Plants

Abstract Addressing the global water crisis requires efficient wastewater treatment and reuse, where accurate estimation of Biochemical Oxygen Demand (BOD) is a key factor. Traditionally, BOD is measured as BOD₅, requiring a five-day incubation period. This long delay means that calculations of purification parameters rely on outdated values from the previous five days, often leading to imprecise monitoring and control. Previous studies using Artificial Neural Networks (ANNs) for BOD estimation have mainly focused on optimizing the weights and biases, while the number of hidden layers and neurons within each layer is usually determined experimentally—a process that is both time-consuming and often imprecise. In this work, we propose a hybrid artificial intelligence (AI) approach that addresses this limitation. The optimal ANN architecture is automatically determined using a Genetic Algorithm (GA), a heuristic method inspired by natural selection, while the weights and biases are further fine-tuned using Biogeography-Based Optimization (BBO). This dual optimization strategy significantly improves model robustness and accuracy. By modeling the relationship between BOD and easily measurable parameters such as Chemical Oxygen Demand (COD) and Total Suspended Solids (TSS), and applying it to data from the Bouira wastewater treatment plant, our method enables rapid and precise same-day BOD prediction. This AI-driven approach eliminates the five-day lag of conventional methods, enhances the accuracy of purification parameter calculations, and supports more effective real-time monitoring for sustainable water management.