Roughness Sensitivity Analysis For Water Distribution Model Calibration

Parameter estimation in water distribution system (WDS) modeling plays an important role, which includes nodal demand and pipe roughness estimation. Now nodal demand can be abtained by statistic method, but roughness values cannot be measured directly. Usually, the roughnesses are backcalculated from flow test results, which is conducted by opening a hydrant to induce flow. But it's difficult to carry out this process in real network because the number of unknowns is too large. In this thesis, the effect of pipe roughness change on the state of the theoretical WDS is analysed. The conclusion is drawn that the nodal pressure is more sensitive to only a few pipes' roughness, and is mostly determined by the roughness of these pipes, while the roughness of the other pipes have little impact. We also get the conclusion that the sensitive coefficients of the nodal pressure to the pipe roughness are power functions of the flow velocities in the pipes. Therefore the new approach is proposed, which estimates those pipes' roughness that has the large sensitivity coefficients, and sets the other pipes' roughness as those determined by the experience. Since Genetic Algorithms is flexible and convenient to find a global solution, it can be used for parameter estimated in water distribution network. A real WDS model calibration based on Genetic Algorithms is used to validate the approach and shown to perform extremely well, matching the optimal solutions produced by complete enumeration in a series of validation tests.