PERFORMANCE UNDER STRESS
Calibration of Redcliffe Water Supply Network
K Goraya, J Xu
Publication Date (Web): 2 May 2017
In general, water hydraulic models in Queensland are validated only for seasonal max hour demand conditions for long term planning. Under these conditions, the observed head loss is small in comparison with other potential errors. This gives the false impression that the model is properly calibrated for all conditions. This may result in service deficiencies when the network is stressed, resulting in poor capital expenditure decisions.
Good calibration requires stress testing the network with high flows such as hydrant flow tests. This generates friction losses substantially higher than other sources of error, allowing pipe friction factors to be calculated with greater confidence. Model calibration is important to ensure the model reliably reflects the performance of the operating network. This is especially true in older parts of the network where the tuberculation of unlined cast iron fittings has the potential to severely compromise network performance.
The Redcliffe Water Supply Scheme is one of the oldest networks in Unitywater’s service area with 60% of the network constructed prior to 1975. During the Millennium Drought, the Redcliffe network was reconfigured by dividing it into 17 District Management Areas (DMA), reducing much of the network’s interconnectivity.
Unitywater conducted water network calibration of six Demand Management Areas (DMA) in the Redcliffe Peninsula using Genetic Algorithm (GA) with Extended Period Simulation. These areas were selected to get a better understanding of the performance of the water network which had experienced fire flow performance issues after the division of the network to establish Demand Management Areas. In addition to this, historical as-constructed records and asset information were not always reliably captured in the network model. These factors created uncertainty in predicting network performance such as fire flows and long term network planning.
Flow tests were carried out at hydrants located in the extremities of the DMA and at critical facilities, causing pipe velocity to increase on average 7.5 times for day background flow with pressure loggers installed at strategic locations to detect pressure drops from the hydrant flow tests.
Calibration involved grouping pipes based on internal diameter, pipe material and age. Genetic Algorithm (GA) was used with Extended Period Simulation to find the best roughness match for each pipe group.
Simulated pressure results from the calibrations were then compared with field pressure readings to evaluate if the DMAs were satisfactorily calibrated. Engineering judgement was used to improve the GA predictions where a poor match was observed.
GA was found to be a time efficient alternative to the manual adjustment of friction factors. However the performance of GA was highly dependent on the input parameters such as the provided range of friction factors for each pipe group. GA provided an answer out of many possibilities that fulfilled the tolerance requirements. However, when the provided friction ranges were too wide, unrealistic solutions were considered and GA stopped iterating once the calibration tolerance was met. Therefore, engineering judgement was required to ensure only realistic friction factor ranges for pipe groups are used. If the calibration results were unsatisfactory with those constraints, it was necessary to manually identify local network restrictions and adjust the model accordingly.
The calibrated model will be used to locate underperforming sections of the network for field investigations prior to augmenting or renewing assets. It will also guide the refinement of the capital works program and provide confidence in lowering DMA pressures to reduce leakage while ensuring the desired level of service will still be met.
Click here to read the full paper
Please note: You need to login to access this member-only content.
Not a member? You can purchase the paper in our Online Document Library.