A Real Time Optimization Based Sequential Convex Program for Pressure Management in Water Distribution Systems

Optimizing pressure management to reduce water leakage in water distribution systems (WDSs) is a top priority for water utilities worldwide. This engineering problem can be mathematically formulated as a nonlinear program (NLP), where the decision variables are valve settings and pump speed settings. By formulating and solving the NLP with given demand parameters, control quantities can be computed and applied to the system for effective control. However, in practice, demand parameters continuously change over time, and solving the NLP typically requires a significant computational time, making it unsuitable for real-time control systems. Therefore, the efficiency of obtaining the NLP solution quickly becomes crucial for improving the performance of the control system. In this study, we propose a new real-time scheme based on a sequential convex program (SCP) to compute approximate control profiles in response to changes in water demand patterns. Instead of solving the NLP with high accuracy for each new demand pattern, the proposed approach solves only one convex NLP within the computation framework of the SCP. This provides an approximate solution with acceptable accuracy, delivered in near real-time. To demonstrate the effectiveness of the real-time optimization scheme, we apply it to determine fast control quantities for a real-world WDS in Vietnam and a WDS benchmark for optimal pressure management. The results demonstrate that by applying this real-time optimization scheme, the obtained control profiles achieve acceptable accuracy and lead to a decrease in excessive pressure, with lower intensity of pressure fluctuations.