Optimization of water quality sensor placement in water distribution networks based on multi-objective evolutionary algorithm and logistic regression

For the efficient identification of pollution events in water distribution networks using limited sensor monitoring data, we propose the MOEA-LRM algorithm as a method for optimizing the water quality sensor layout of water supply networks by integrating a multi-objective evolutionary algorithm (MOEA) with a logistic regression model (LRM). The effectiveness of this approach is demonstrated through its application to the Anytown and Fosspoly1 pipe network systems. The MOEA-LRM algorithm aims to minimize the number of sensors, as well as the average and worst-case impact risk, by constructing a mathematical model using the MOEA algorithm that achieves Pareto equilibrium within a pipe network system. Based on this premise, the MOEA-LRM algorithm leverages the LRM to efficiently screen and identify the optimal sensor layout, thereby enhancing the accuracy of contamination source identification across the entire network. The results illustrate that this approach consistently identifies an optimal sensor configuration that ensures accurate identification of the source of contamination throughout the pipe network and effectively reduces the impact of exogenous water pollution incidents on users.