296937
Transcript of 296937
-
8/9/2019 296937
1/8
-
8/9/2019 296937
2/8
Figure 2: Distribution system of Gruszczyn,
including nine new and two existing pressuremeasuring points
The goal of the automation project was to run thewater supply system unmanned, and to optimizethe control of the system. Optimized control shouldresult in a reduction of operational costs byminimizing the energy consumption.
MATERIALS AND METHODS
Production flow controlIn the first stage of the automation and optimizationproject, a relative simple production flow controlloop was programmed in the programmable logiccontroller (PLC). In this control loop the productionflow set-point was directly derived from the level inthe clear water reservoir (increasing level resultedin decreasing set-point, and vice versa). This levelbased production flow control was capable ofcontrolling the production unmanned. However, thiscontrol was not optimal, because many productionflow changes occurred, resulting in a sub optimalwater quality, and water production was not at thelowest possible energy costs.
In the second stage of the project, the computer-based predictive production flow control software(OPIR) was installed. This predictive flow controlmodel forecasts the water demand for the next 48hours with 15-minutes time steps. The self learningforecasting algorithm automatically builds up adatabase with specific water demand curves andfactors of the supply area, and uses these curvesand factors to predict future demand. The functionaldetails and the performance of this forecastingmodel is described by Bakker et al. (2013,submitted). Based on the forecast of the waterdemand, the flow control algorithm calculates set-
points for the production flow. The boundarycondition in this calculation is that the level in theclear water reservoir must stay between a chosenupper and lower level; the optimization is to
produce the water with a minimum of productionflow changes at the lowest possible energy costs.Figure 5 shows trends of the production flow of bothlevel based control as well as model predictive flowcontrol.
Influence of production flow controlThe production flow control influences the variabilityof the production flow, which can be expressed inthe production variation per day (PV d ). Theproduction variation is defined as the sum of the(absolute values of) the difference betweensubsequent hourly average production flow values(F prod,d,h ) divided by the total daily production:
∑ ,, − ,,
∑ ,,∙100% (1)
Production flow control also influences the energy
consumption for the production of the drinkingwater. Because real-time energy measurementswere not available, the energy consumption forabstraction and treatment P prod was estimated with:
+ ∙ + ∙ (2) Where P base [kW] is the constant, flow independent,energy consumption, F prod [m
3 /h] is the production
flow, C 1 is a value representing energyconsumption for static head loss and C 2 fordynamic head loss. The values for P base , C 1 and C 2 were chosen, such that the average estimatedenergy consumption matched the specific energyconsumption based on the energy invoice of 2011(0.456 kWh/m
3 for abstraction and treatment).
Pressure controlThe distribution pumping station of WTP Gruszczynconsists of five identical pumps all equipped withvariable speed drives (VSD). The pumps areoperated as one group at a fixed pump pressure.The clear water is pumped in two directionstowards individual supply areas. Initially the twoareas were separated from each other by a PRV inorder to reduce the pressure in one zone whilekeeping a higher pressure in the other zone. Theoperators chose a relative high pressure set-pointfor the clear water pumping station, because of alack of information about the pressure in the entirenetwork during f low variations.
In the automation and optimization project, ninenew pressure measuring points were installed inthe distribution network (see Figure 2). Themeasured pressures showed that there was noneed to separate the two pressure zones, and thatthe existing PRV could be removed. After removingthe PRV, the pressures in both zones wereequalized.
-
8/9/2019 296937
3/8
For the control of the clear water pumping station,the dynamic pressure control module (DPCM) ofthe OPIR software was installed. Figure 3 showsthe user interface of the DPCM.
Figure 3: Interface of the dynamic pressure controlmodule (DPCM), showing all measured
pressures and their lower bound values, and
highlighting which measuring point is themaster in the control loop
In the conventional control, the pressure set-pointwas a static value chosen by the operator. TheDPCM is a feedback control model, whichdynamically calculates a set-point for the pumpingstation by comparing the measured pressures atthe measuring points with their individual lowerbound values. The measuring point with the lowestpressure in relation to its lower bound value is themaster in the control loop. The DPCM uses aproportional integral derivative (PID) control
mechanism to derive a pressure set-point for thepumping station, based on the desired (lowerbound) and measured pressure value of this masterpressure measuring point.
Using off-line pressure measuring pointsThe nine installed pressure measuring points areequipped with a local logger and GSM modem. Themeasured pressures were buffered locally and sentto the SCADA system of WTP Gruszczyn once perday. This implies that the measured values werenot in real-time available. However, the DPCMestimates the real-time pressure p i for each
pressure measuring point i as a function of the real-time measured pressure at the pumping station p ps and distribution flow to the area F dist :
+ + ∙ (3)
The values for a and b in equation (3) were derivedby the DPCM using data of the previous 72 hours(see Figure 4).
Figure 4: Least squares fit of measured pressuredrop between pumping station and pressure
measuring point as a relation of the flow to thearea
By this functionality, the DPCM is a feedbackcontrol model using a predicted value as input, andcan therefore be considered to be a hybrid form of
a predictive controller and a feedback controller asdescribed by Ulanicki et al. (2000).
Influence of pressure controlPressure control influences the average pressure inthe water distribution network and as a results alsothe leakage in the distribution network. Thebackground leakage q leak can be described with(Gomes et al., 2011; Araujo et al., 2006;Vairavamoorthy and Lumbers, 1998):
∙ (4)
Where K f is a leakage coefficient for the area, p isthe average pressure in the area, and β is pressureexponent. According to Gomes et al. (2011), thepressure component β varies between 0.5 (forleaks with a fixed leaking area, which is the casewith steel pipes or other rigid pipes) and 2.5 (forleaks with a leaking area which is highly sensitivepressure fluctuation, which is the case with HDPEpipes or other flexible pipes). Gomes et al. (2011)use a value of 1.0, where Ulanicki et al. (2000) usea value of 1.5 for background leakages. Asproposed by May (1994) and adapted by Araujo et
al. (2006) we will use 1.18 for β in this paper.
If the pressure in the area changes, the backgroundleakage in the area will change according to:
,,
(5)
A reduction of the leakage will lead to a reduction ofthe amount of water to be pumped. Therefore, alsothe energy consumption will be reduced. Thisreduction dE
loss can be estimated with:
∙ , ℎ (6)
-
8/9/2019 296937
4/8
Where dV loss is the difference in water loss in thewater distribution system and E spec,tot is the totalspecific energy consumption for abstraction,treatment and distribution (0.600 kWh/m
3).
Changing the pressure will also affect the energyconsumption by the clear water pumps. Thedifference in energy consumption dE
pump is
calculated with:
∙ ∙
1000∙3600∙ ℎ (7)
Where ρ is the specific mass of water (1,000kg/m
3), V is the pumped volume of water, dp is the
difference in pump pressure, and η is the totalefficiency of pump + motor of the clear waterpumps (estimated to be constant at 0.60).
RESULTS AND DISCUSSION
Comparison of operational periodsTo evaluate the results of the project, theoperational data (flows, pressures, water levels) ofa period with conventional control were comparedwith a period of optimized automatic control. Theimplementation of the advanced control softwarewas done in several phases, and after initialimplementation a period of tuning followed.Therefore, a contiguous period with a sharptransition from conventional control to advancedcontrol was not available. Therefore, we compared,for both control strategies, a three weeks period inNovember: conventional control in November 2011,and advanced control in November 2012.
Production flow controlFigure 5 shows trends of the total water demand,production flow and reservoir level of both
examined periods.
Figure 5: Difference between level based (upper graph) and model based (lower graph) production flowcontrol
The trend with level based control shows that theproduction flow was switched up and down everyday, and that the minimum and maximum flowvalues were almost equal to the minimum andmaximum distribution flow values. The trend withmodel based control shows a more stableproduction flow, with a smaller difference betweenmaximum and minimum flows.
Table 1 shows that model based control leads to a83% lower value for the production variation,calculated with equation (1). Bakker et al. (2013, inpress) showed that treatment performance was
better at lower values of production variation,resulting in lower values of the turbidity of the clearwater.
Table 1 also shows the minimum and maximumobserved production flows in both periods, and thedifference between the maximum and minimumvalues. The table shows that the differencebetween the maximum and minimum production
flows was 67% lower with model based control.
0.00
0.50
1.00
1.50
2.00
2.50
3.00
0
200
400
600
800
1000
1200
31-Oct-2011 7-Nov-2011 14-Nov-2011 21-Nov-2011
l e v e l [ m ]
f l o w [ m 3 / h ]
Level based production flow control
Production flow Distribution flow Reservoir Level
0.00
0.50
1.00
1.50
2.00
2.50
3.00
0
200
400
600
800
1000
1200
5-Nov-2012 12-Nov-2012 19-Nov-2012 26-Nov-2012
l e v e l [ m ]
f l o w [ m 3 / h ]
Model based production flow control
Pr odu ction flo w Distribu tion flow Reser voir Lev el
-
8/9/2019 296937
5/8
Table 1: Difference between level based and modelbased production flow control
Levelbased
Modelbased
Diffe-rence
PV [%] 9.3 1.6 -83%Production flowMin. flow [m
3 /h]
Max. flow [m
3
/h]Max.-Min. [m3 /h]
52
875823
367
636269
+600%
-27%-67%Energy (est.)Cons. [kWh/m
3]
Cost [ € /1,000 m3]
0.45627.43
0.44726.68
-1.9%-2.7%
A third aspect shown in Table 1 is the energyconsumption and energy cost, which are calculatedwith equation (2) for both periods. The energyconsumption [kWh/m
3] is 1.9%, and relatively more
energy is consumed during low tariff hours resultingin a reduction of the costs of 2.7%. With a total
annual production of WTP Gruszczyn of 5 millionm
3 per year, the implementation of model based
production flow control led to a reduction in energyconsumption of 43,500 kWh per year ( € 3,750 peryear).
Pressure controlFigure 7 and Figure 8 show trends of the waterdemand, the outlet pressure at the pumping stationand the average pressure in the area of bothexamined periods.
In the initial setup of the water supply system, a
PRV was reducing the pressure to one of the two
supply areas (Swarzędz area, see Figure 1). Thepumps were operated at a fixed pressure (330kPa), and the fixed outlet PRV was set to reducethe pressure to the Swarzędz area to 280 kPa. Theinstalled PRV was a medium driven automaticcontrol valve Cla-val NGE9001 (DN250). Prescottand Ulanicki (2003) used this type of valve todevelop their dynamic model of PRVs. According tothis model, the PRV shows a limited flowdependence: during low flows the outlet pressure issomewhat higher than during high flows. Thischaracteristic was also observed in the trends offlow and pressure to the Swarzędz area (Figure 6).
Figure 6: Outlet pressure in the water main toSwarz ędz area with conventional pressure control
including PRV
Figure 7: Difference between static (upper graph) and dynamic (lower graph) pressure control, Pozna ń area
260
270
280
290
300
310
320
330
340
0
100
200
300
400
500
600
700
800
11 -Nov -2 011 12 -Nov -2 011 13-No v-20 11 1 4-Nov -2 011
p r e s s u r e [ k P a ]
f l o w [ m 3 / h ]
PRV, Swarzędz area
Flow Pres. before PRV Outlet pres .
260
265
270
275
280
285
290
0 100 200 300 400 500
o u t l e t p r e s s u r [ k P a ]
Flow [m3 /h]
PRV characteristic, Swarzędz area
100
150
200
250
300
350
400
450
500
0
100
200
300
400
500
600
31-Oct-2011 7-Nov-2011 14-Nov-2011 21-Nov-2011
p r e s s u r e [ k P a ]
f l o w [ m 3 / h ]
Static pressure control, Poznań area
Di str ib ut io n fl ow Pr es su re at PS Pr ess ur e in a re a (a ver age)
100
150
200
250
300
350
400
450
500
0
100
200
300
400
500
600
5-Nov-2012 12-Nov-2012 19-Nov-2012 26-Nov-2012
p r e s s u r e [ k P a ]
f l o w [ m 3 / h ]
Dynamic pressure control, Poznań area
Di str ib ut io n fl ow Pr es su re at PS Pr ess ur e in a re a (a ver age)
-
8/9/2019 296937
6/8
-
8/9/2019 296937
7/8
than in the period with static control. This increaseis caused by a large industrial customer in the area,who increased its water demand. The water flow tothe Swarzędz area shows little difference (2%) inthe two periods.
Table 2: Difference between static and dynamicpressure control
Staticcontrol
Dynamiccontrol
Diffe-rence
Pozna ń areaFlow [m
3 /h]
PS [kPa]MP1 [kPa]MP2 [kPa]MP3 [kPa]Area avg. [kPa]
214330395475462444
306233298373361344
+43%-29%-25%-22%-22%-23%
Swarz ędz area
Flow [m
3
/h]Outlet [kPa]MP1 [kPa]MP2 [kPa]MP3 [kPa]MP4 [kPa]MP5 [kPa]MP6 [kPa]Area avg. [kPa]
261279470516439365329381417
267233422452390318281337367
+2%-16%-10%-12%-11%-13%-14%-12%-12%
A reduction of the pressure in the area resulted in areduction of the water losses in the distributionnetwork. By applying equation (5), a reduction in
background leakage was calculated of 26% for thePoznań area, and 14% for the Swarzędz area(which corresponded to a 20% reduction for theentire Gruszczyn system). The total water losses ofthe water utility Aquanet amounted 5.30 million m
3
or 11.3% (Aquanet, 2012). We assumed that thewater losses were equally distributed over allsupply areas, because no detailed informationabout water losses in different areas was available.With the above, we estimated the water losses inthe Gruszczyn system in 2011 at 565,000 m
3 per
year and in 2012 at 450,000 m3 per year (water
loss reduced from 11.3% in 2011 to 9.0% in 2012,
difference 113,500 m3
).
Girard and Stewart (2007) and Gomes et al. (2011)showed that a reduction in the pressure in a watersupply area also leads to a reduction in the numberof main breaks and service breaks. Based on thereduced pressure in the area, a reduction in thenumber of breaks may be expected in this casestudy as well. However, the number of breaks in theconcerning areas were not registered separately,making it impossible to confirm the expectedreduction in breaks.
Reduction of energy consumptionThe reduction of energy consumption due to theimplementation of advanced control softwareconsists of three elements:
1. Savings due to production flow control;2. Savings due to lower pump pressure of
clear water pumps3. Saving due to reduced water losses.
The energy reduction by production flow controlwas estimated at 43,500 kWh per year ( € 3,750 peryear) earlier in this paper. The energy reductiondue to lower pump pressure was calculated withequation (7). With a pumped volume (V ) of 5million m
3 per year, and a pump head (dp ) of 97
kPa, this results in a dE pump of 225,000 kWh peryear ( € 13,550). The energy reduction due toreduced water losses was calculated with equation(6). Based on a reduced water loss of 113,500 m
3
per year (see above), the reduction in energy
consumption was calculated at 68,500 kWh peryear ( € 4,100).
The reductions in energy consumption are listed inTable 3. The table shows that the lower pumppressure is the main contributor to the reduction ofthe overall energy consumption of 337,000 kWh peryear ( € 21,500 per year). The observed reductioncorresponded to a reduction of 11.5% of the overallenergy consumption of the Gruszczyn water supplysystem.
Table 3: Energy savings due to the implementation
of advanced controlEnergy
kWh/year
Costs
€ /year
1. production flow control2. lower pump pressure3. reduced water loss
43,500225,000
68,500
3,75013,5504,200
Total 337,000 21,500
CONCLUSION
The implementation of the production flow controland the dynamic pressure control modules of theOPIR software at WTP Gruszczyn, resulted in amore constant production flow and a reduction ofthe pump pressure of the clear water pumps. Theproject has led to considerable savings in theoperational costs by the reduced energyconsumption (337,000 kWh per year, or € 21,500per year). The project has shown that extrainformation from the distribution network (from ninenew pressure measuring points) in combinationwith advanced control software led to a moreefficient water supply system.
-
8/9/2019 296937
8/8
ACKNOWLEDGMENTS
The project is financially supported by the Dutchgovernment through the “Partners for Water”programme. The project was implemented by RoyalHaskoningDHV the Netherlands, supported bywater utility Oasen (the Netherlands), water utilityAquanet (Poland) and Royal HaskoningDHVPoland.
REFERENCES
Abdel Meguid, H., Skworcow, P., and Ulanicki, B.
(2011). "Mathematical modelling of ahydraulic controller for PRV flowmodulation". Journal of Hydroinformatics.13 (3): 374-389.
Aquanet. 2012. Annual report 2011.
Araujo, L.S., Ramos, H., and Coelho, S.T. (2006)."Pressure control for leakage minimisationin water distribution systems management".Water Resources Management. 20 (1):133-149.
Bakker, M., Van Schagen, K., and Timmer, J.(2003). "Flow control by prediction of waterdemand". Journal of Water Supply:Research and Technology - AQUA. 52 (6):417-424.
Bakker, M., Vreeburg, J.H.G., Palmen, L.J.,
Sperber, V., Bakker, G., and Rietveld, L.C.
(2013, in press). "Better water quality andhigher energy efficiency by using modelpredictive flow control at water supplysystems". Journal of Water Supply:Research and Technology - AQUA.
Bakker, M., Vreeburg, J.H.G., Rietveld, L.C., andSchagen, K.M. (2013, submitted). "Shortterm water demand forecasting with ageneral adaptive model". Journal ofEnvironmental Modelling and Software.
Bunn, S.(2007). "Closing the loop in water supplyoptimisation", Proceedings IET WaterEvent 2007 , Coventry.
Bunn, S.M., and Reynolds, L. (2009). "The energy-
efficiency benefits of pumpschedulingoptimization for potable water supplies".IBM Journal of Research and
Development. 53 (3).
Colombo, A.F., and Karney, B.W. (2005). "Impactsof leaks on energy consumption in pumpedsystems with storage". Journal of WaterResources Planning and Management. 131(2): 146-155.
Girard, M., and Stewart, R.A. (2007).
"Implementation of pressure and leakagemanagement strategies on the gold coast,Australia: Case study". Journal of WaterResources Planning and Management. 133(3): 210-217.
Gomes, R., Marques, A.S., and Sousa, J. (2011)."Estimation of the benefits yielded bypressure management in water distributionsystems". Urban Water Journal. 8 (2): 65-77.
May, J. (1994). "Pressure dependent leakage".World Water & Environmental Engineer. 17(8): 10.
Prescott, S.L., and Ulanicki, B. (2003). "Dynamicmodeling of pressure reducing valves".Journal of Hydraulic Engineering. 129 (10):
804-812.
Ulanicki, B., Bounds, P.L.M., Rance, J.P., andReynolds, L. (2000). "Open and closedloop pressure control for leakagereduction". Urban Water. 2 (2): 105-114.
Vairavamoorthy, K., and Lumbers, J. (1998)."Leakage reduction in water distributionsystems: Optimal valve control". Journal ofHydraulic Engineering. 124 (11): 1146-1154.