Resumen:
|
[EN] Real-time operation of drinking water networks (DWN) through supervisory control and data acquisition (SCADA) system can potentially reduce the amount of energy consumed. This can be achieved by controlling pumps and ...[+]
[EN] Real-time operation of drinking water networks (DWN) through supervisory control and data acquisition (SCADA) system can potentially reduce the amount of energy consumed. This can be achieved by controlling pumps and valves to reduce pumping cost and energy dissipated at valves and improve pressure and water quality throughout the DWN. With the integration of hydraulic models and real-time data sources, optimization of real-time operation of DWNs has become more feasible in recent years. An accurate hydraulic model of a DWN is necessary to realize the benefits of real-time modeling and control, as the model is required for estimating the state of the network for different operating conditions. Pump and valve curve estimation is a key step towards achieving an accurate model. Valves, which act as energy sinks, are vital elements in DWNs and stabilize pressures in certain parts of the network. Accurate flow-head loss relationships for a range of valve openings is important for real-time modeling, for example, to estimate pressures downstream of the valves and optimize the operation of pumps and valves for forecasted demands to reduce the overall energy spent for DWN operation. Head loss across valves is conventionally modeled as proportional to the velocity head across the valve, with the proportionality factor called the minor-loss coefficient. The minor-loss coefficient, in turn, is a non-linear function of the valve opening degree, and valve manufacturers generally provide relationships between the minor-loss coefficient and the valve percent opening. Our experience is that use of these original manufacturer curves can lead to significant prediction errors when compared to operational data, and should be tested and estimated. We show how to use measurements of valve opening, pressure differential, and flow through an individual valve or a valve facility for the estimation of valve curves. The approach can be used as a batch process as part of network calibration, or may also be used in real-time to maintain accurate valve characteristics over time. A method is presented to estimate valve loss curves represented by a simple function (e.g., power-law or polynomial), using measured flow and pressure differential and measurement of valve opening degree. The method uses a gradient-based optimizer to solve a least squares estimation (LSE) problem where the residual error is the difference between predicted and measured flow through the valves. This method is applicable for valve facilities that contain more than one valve in parallel, where flow measurements are available for the valve facility or individual valves. The method developed was applied to estimate 46 valve curves for a large water transmission network in the USA as part of the development of a digital twin for the DWN. Results will summarize the statistical prediction errors for this set of control valves, before and after estimation, and will show the impact on network hydraulic predicti
[-]
|