A Machine Learning Approach for Anaerobic Reactor Performance Prediction Using Long Short-Term Memory Recurrent Neural Network

• DOI: 10.21741/9781644901311-8
• Publication Date: 2021-02-16T14:27:34Z
• Authors: Thomas Kuen; L.R.F. Rose; Weilong Kong; Benjamin Steven Vien
• Keywords: Predictive modelling; Feature (linguistics)

Abstract. Predictive models are important to help manage high-value assets and ensure optimal safe operations. Recently, advanced machine learning algorithms have been applied solve practical complex problems, of significant interest due their ability adaptively ‘learn’ in response changing environments. This paper reports on the data preparation strategies development predictive capability a Long Short-Term Memory recurrent neural network model for anaerobic reactors employed at Melbourne Water’s Western Treatment Plant sewage treatment that includes biogas harvesting. The results show rapid training higher accuracy predicting production when historical data, which include outliers, preprocessed with z-score standardisation comparison those max-min normalisation. Furthermore, trained reduced number input variables via feature selection technique based Pearson’s correlation coefficient is found yield good performance given sufficient dataset training. It shown overall best comprises processed standardisation. initial study provides useful guide implementation techniques develop smarter structures management towards Industry 4.0 concepts.