Including snowmelt in influent generation for cold climate WRRFs: comparison of data-driven and phenomenological approaches

Abstract

Influent generation models are developed to provide the influent disturbance at the inlet of a WRRF. A reliable influent model is important for WRRF design, upgrade and different digital twin studies. In this work, a data-driven methodology is proposed to create an influent generator (IG) model, which describes the influent flow and water temperature dynamics under the impact of snowmelt under cold climate conditions. The model structure applied was the long short-term memory (LSTM) artificial neural network with residual connection. The final result of influent generation for a Canadian case study is compared with a previously proposed phenomenological model. The performance is evaluated by different performance criteria and the results revealed that the LSTM approach has a better performance than the phenomenological model in terms of accuracy. In conclusion, the proposed model can successfully reproduce the influent dynamics of a combined sewer system's wastewater generation with snowmelt infiltration impacts.