Calibrating water distribution system model automatically by genetic algorithms

Calibration of a water distribution system model is a complicated task. There are many uncertain parameters that need to be adjusted to reduce the discrepancy between the model predictions and field observations of junction pressure and pipe flow. This paper outlines the genetic algorithms (GA) based calibration framework which facilitates a variety of practical network model calibration tasks including the extended period flow and pressure calibration. The efficient genetic algorithm drives the search process for locating the optimal and a number of near-optimal solutions. It automatically generates and evaluates hundreds of thousands of possible solutions, which is not possible by conventional trial-and-error method. Thus the search process effectively improves the calibration accuracy. The case study demonstrates that the integrated calibration method gives modelers the maximum flexibility to improve the model accuracy and robustness.