Peter (WP6) – Nonlinear data-driven predictive control design for water distribution networks

In this paper, we present a novel method for controlling Water Distribution Networks (WDNs) using Data-driven Predictive Control (DPC). First, we identify through physical first-principle knowledge that a standard linear predictor is insufficient. However, by mapping the control input as a nonlinear function to a measurable intermediate variable, we can obtain an accurate data-driven predictor. This furthermore allows us to retain the standard cost function and constraints employed for the control of WDNs. The proposed algorithm is implemented and simulated on a small example WDN. The resulting nonlinear data-driven predictive control algorithm performs well on the network, showing the expected response. Read more here: https://doi.org/10.1109/CDC56724.2024.10886363