Water Journal : Water Journal November 2013
NOVEMBER 2013 WATER 51 Feature Article The relatively high ow rates and friction losses in the Tamworth System translate into a faster dynamic system response, i.e. due to higher friction, the ow rates in the Tamworth System settle relatively fast. The relatively low ow rates and friction losses in the Dandalup System translate to a slower, more oscillatory dynamic system response, i.e. due to low friction, pressure waves travel up and down the Dandalup System more easily before damping out. Therefore, the ow rates in the Dandalup System only settle after a relatively long time. A strong interaction between the two ow control loops would be much more likely if the Tamworth System and the Dandalup System were more similar in dynamic behaviour. Instability would also be more likely if the Dandalup ow control loop were to be tuned to respond as fast as the Tamworth ow control loop. CONTROL SYSTEM DESIGN OF RAVENSWOOD PUMP STATION After testing three control strategies, a control system consisting of a master ow controller and slave delivery pressure controller was selected for both the Tamworth Bank and the Dandalup Bank (Figure 4). In the selected control strategy, the ow rates of pump banks are controlled indirectly by the master ow controllers by controlling the delivery pressures of the pump banks. The ow setpoints for the Tamworth Bank and Dandalup master ow controllers are set by the remote operator, based on the target daily transfers to, respectively, the Tamworth System and the Dandalup System (GHD, 2009). The control strategy selected for the two pump banks has the bene t that the control system can cope with a loss of ow meter signal. In this situation, the control of a bank would revert to delivery pressure control only, using the last known delivery pressure setpoint. The control strategy selected was shown to perform well in controlling the ow rates of pump banks, both in simulations and at other variable speed pump stations in Perth's Integrated Water Supply Scheme (IWSS). In addition to the ow control feedback loop, the control system also includes a suction pressure override on the Tamworth Bank that prevents the pressure on the suction side of Ravenswood Pump Station from dropping too low. AUTOMATICALLY STARTING AND STOPPING PUMPS The simulations in the detailed design stage showed that the control logic for automatically starting and stopping pumps could be kept relatively simple, while still being effective in maximising the pump ef ciency. When the ow setpoint of a bank is increased, the pump ow will increase past the best ef ciency point (BEP) ow of the pumps. An additional pump is started if the average pump ow for a bank exceeds 110 per cent of BEP. When the ow rate drops, a pump is stopped if the average pump ow for a bank falls below 105 per cent of BEP. The last pump running in a bank is only stopped when the pump ow falls below 50 per cent of BEP. This control logic ensures the pumps operate as closely as possible to their BEP. When the control logic triggers an additional pump to start, the starting pump simply ramps up at a constant rate until it reaches the speed of the pumps already running (Figure 5). When the ow controller starts noticing the effect of the starting pump, it will automatically adjust the speed of all pumps until they are running at the same slightly reduced pump speed. MODEL VALIDATION The WANDA model for the control system of Ravenswood Pump Station was calibrated and validated using SCADA data recorded during commissioning tests. For the model validation, the initial conditions in the hydraulic model were set up using SCADA data and the bank start-ups were simulated using the control parameters established in the model calibration. The validation results for start- up of the Dandalup Bank are presented in Figures 6, 7 and 8. The results of the WANDA model validation (Figures 6 and 7) show that the delivery and suction pressures captured by SCADA during start-up are replicated accurately by the WANDA model. The SCADA data for pump speed and ow rate are replicated accurately during the rst part of the start-up. There are small discrepancies between the ow rate and pump speed captured by SCADA and the values calculated by the model for the second part of the start-up, but the overall control response is replicated well. Figure 8 shows the operating ranges for one, two and three Dandalup Bank pumps running in parallel. The limits of the operating ranges are the pump curves at maximum and minimum speed, and the curves for 50 per cent and 110 per cent of BEP pump ow. The gure also shows plots of the bank ow rate versus the pump head, referred to here as the "transient duty points", for the SCADA data and the simulation results. The transient duty point shows where the bank is operating in respect to the pump operating ranges. The gure shows that the transient duty point from SCADA and the transient duty point calculated with the WANDA model follow a similar path, with only some minor discrepancies. In reviewing the SCADA data of the commissioning test, it was noted that the transient duty point travels outside of the one-pump operating range. This would normally cause the second pump to Figure 6. Model validation -- Dandalup Bank pump speed and ow rate. Figure 5. Simulation results -- Tamworth Bank automatic pump start.
Water Journal September 2013
Water Journal December 2013