Predicting wastewater temperatures in sewer pipes using abductive network models

Abstract

A predictive modelling technique was employed to estimate wastewater temperatures in sewer pipes. The simplicity of abductive predictive models attracts large numbers of users due to their minimal computation time and limited number of measurable input parameters. Data measured from five sewer pipes over a period of 12 months provide 33,900 training entries and 39,000 evaluation entries to support the models' development. Two simple predictive models for urban upstream combined sewers and large downstream collector sewers were developed. They delivered good correlation between measured and predicted wastewater temperatures proven by their R2 values of up to 0.98 and root mean square error (RMSE) of the temperature change along the sewer pipe ranging from 0.15 °C to 0.33 °C. Analysis of a number of potential input parameters indicated that upstream wastewater temperature and downstream in-sewer air temperature were the only input parameters that are needed in the developed models to deliver this level of accuracy.