Optimizing TAN Concentration Prediction in Various Anaerobic Digestion Systems: A Machine Learning Approach

Monitoring Total Ammonia Nitrogen (TAN) is critical for the stability and effectiveness of anaerobic digestion systems. This study concentrates on predicting TAN concentration in AD systems employing multiple machine learning models with a dataset of 2447 units covering various AD operating conditions with various OLR and temperature ranges. Four models were developed and optimized: the Support Vector Machine (SVM) Regressor, Extreme Learning Machine (ELM), Partial Least Squares Regression (PLSR), and Long Short-Term Memory (LSTM) network. The ELM model outperformed other models, with a R2 of 0.93 and RMSE of 0.24 on test data. The results demonstrate the significance of non-linear models such as ELM and LSTM in capturing complex patterns in data, as well as the importance of including various AD reactor settings to improve model resilience and generalization.