Related Experiment Video
Updated: Jul 11, 2025
Author Spotlight: Understanding and Detecting Environmental Antimicrobial Resistance by Combining Culture-Based Techniques and Genomics
Published on: July 19, 2024
Predicting biochemical oxygen demand in wastewater treatment plant using advance extreme learning machine optimized
Hayat Mekaoussi1,2, Salim Heddam3, Nouri Bouslimanni4
1Institute of veterinary and agronomic sciences, Agronomy Department, Hydraulics Division, University Batna 1-Hadj Lakhdar- Allées 19 mai, Route de Biskra Batna, 05000 Algeria.
A new hybrid machine learning model, ELM-Bat, accurately predicts wastewater effluent biochemical oxygen demand (BOD). This advanced model optimizes extreme learning machine (ELM) with the Bat algorithm, outperforming other methods for effective wastewater treatment plant management.
Area of Science:
- Environmental Engineering
- Water Quality Management
- Machine Learning Applications
Background:
- Accurate wastewater quality modeling is crucial for optimizing wastewater treatment plant (WWTP) operations.
- Predicting effluent biochemical oxygen demand (BOD) is a key challenge in WWTP management.
- Existing models often require complex parameter tuning and may lack predictive accuracy.
Purpose of the Study:
- To develop and evaluate a novel hybrid machine learning model, Extreme Learning Machine optimized by Bat algorithm (ELM-Bat), for predicting five-day effluent BOD.
- To compare the predictive performance of ELM-Bat against established models like MLPNN, RFR, GPR, RVFL, and MLR.
- To identify the optimal combination of input variables for accurate BOD prediction.
Main Methods:
- Developed a hybrid ELM-Bat model integrating the Bat algorithm for parameter optimization with the Extreme Learning Machine.
- Utilized historical wastewater quality data, including Chemical Oxygen Demand (COD), temperature, pH, Total Suspended Solids (TSS), Specific Conductance (SC), and flow (Q).
- Calibrated and validated all models using training and testing datasets, assessing performance via RMSE, MAE, R, and NSE metrics.
Main Results:
- The hybrid ELM-Bat model demonstrated superior predictive accuracy compared to all benchmark models.
- Optimal performance was achieved when all six input variables (COD, temperature, pH, TSS, SC, Q) were included.
- ELM-Bat achieved approximate performance metrics of RMSE=0.885, MAE=0.781, R=2.621, and NSE=1.989.
Conclusions:
- The ELM-Bat model represents a significant advancement in wastewater quality modeling for effluent BOD prediction.
- This hybrid approach offers enhanced accuracy and reliability for WWTP planning and management.
- The findings underscore the potential of advanced machine learning techniques in addressing complex environmental challenges.
