Applications of Deep Learning for Smart Water Networks

Abstract

Abstract Deep Learning (DL) is the state-of-art paradigm of Artificial Neural Network (ANN) computing. It is a new breakthrough in machine learning, and differentiates from the conventional or shallow learning algorithms by emulating the six-layer human neocortex, which is unique for human brain containing billions of interconnected neurons. Unlike canonical ANN, DL is capable of self-learning data features by mimicking the self-learning functions layer by layer in human cortex and creating a data-driven model with the given dataset. This paper reports the initial applications of deep learning for simulation, optimization and operation control of water distribution systems. It elaborates the development of efficient deep learning framework with potential applications of facilitating the data fusion, system simulation and predictive analysis, detection of abnormal events from the recorded time series data (pressures, flows and consumptions etc.), water usage prediction, construction of a meta-model as a surrogate to the physics-based models (hydraulic and water quality), and acceleration of the solution search for smart water distribution management, which aims at improving operation efficiency, reducing carbon footprint, and exceling customers’ expectation.