Research

A Toolchain for the Data-driven Decision Support in Waste Water Networks - A Level-based Approach

Read the article

Abstract

This paper aims to enable automated decision-making in combined wastewater and stormwater networks. The proposed concept is based on the deployment of in-sewer water level sensors distributed at critical locations in basins and manholes. With the use of level sensors and weather forecast feeds, we aim to learn how rain infiltrates into the network and build decision support to optimally manage the operation of storage elements, i.e., basins. For that, a data-driven probabilistic framework based on Gaussian Processes is developed. The presented framework enables practitioners to build system knowledge (actuator state, tank dimensions) into the design while leaving the uncertain parts (rain runoff) to be handled by the Gaussian Processes. The paper highlights the practical feasibility of the toolchain through a pilot project with Ishøj Spildevand in Denmark, where the prediction capabilities are tested for five months period, providing a proof of the proposed concept.

IWA decision support machine learning gaussian process data-driven sewer wastewater network control modelling black-box level sensor toolchain utility