flow balance for a large water treatment works
TRANSCRIPT
CONDUCTING A FLOW BALANCE ACROSS FRANKLEY WATER TREATMENT WORKS
Claire Ashton, DNV GL, and Mikal Willmott, Severn Trent
SWIG Modelling Workshop, IWM Duxford
28th September 2016
A Brief History Lesson
2
EVA facts
3
The Importance of the EVA
4
EVA Challenges
• Now 112 years old – an ageing asset
• Difficult to carry out repair and maintenance to the EVA
• Single point of failure
5
Severn Siphon
River Severn
Frankley WTW
Severn Aqueduct
Meriden reservoir (from Strategic Grid)
Trimpley River Intake
118km
Elan Valley Reservoirs
Edgbaston boreholes
New pump station
New pipeline
Asset can be maintained when alternative supply in use
Elan Valley Aqueduct(West Free flow)
Elan Valley Aqueduct(East Free flow)
Norton & Beechtreeboreholes
South Staffs Water Barr Beacon to
Perry Barr
B’ham Ground Water boreholes
20Ml/d
Birmingham Resilience – Overview
New Lickhill Quarry River Intake
Birmingham
Critical exports
294 Ml/d
11Ml/d
Process Loss130 Ml/d
250 Ml/d * (raw inlet) River upgrade,
Dual stream & recirculation
120 Ml/d
20 Ml/d
5 Ml/d
0 Ml/d 0 Ml/d
55 Ml/d
“To provide for a continuous water supply to our customers in Birmingham in the event of either planned maintenance or an emergency shutdown on the EVA and the removal of specified single points of failure from the Frankley WTW process”
“The alternative supply will allow us to undertake planned EVA maintenance events (up to 50 days every other year), allowing a 30 day ‘dry possession’ of the aqueduct. It
also provides a critical back-up supply in the event of an unplanned shutdown”
6
Metering Challenges
Inlets into 42” gravity mains
Inlets into 60” gravity main
Velocity Profile for 18” Hollymoor Flow
Velocity Profile for 24” Northfield Flow
7
Technical outline
• Water balance across Frankley WTW,
based on four stages:
Inlet meters (2 meters)
DAF outlets (2 meters)
RGF outlets (5 meters)
Distribution input meters (12 meters)
• Total of 21 major flow meters analysed
8
• Using analogue escada / site scada and digital logged data
• Additional process meters, and estimates of other additions /losses also included
• Comparison of:
different data sources for individual meters
different groups of meters
Methodology overview
9
Statistical analysis (comparison of
individual meter data and groups of meters)
Individual meter analysis (short-term
scada and logged data)
Bias and uncertainty
analysis (groups of meters)
Meter sizing and accuracy
review (individual meters)
Long-term escada/scada analysis
(groups of meters)
Short-term logged data analysis
(groups of meters)
Long-term escada and scada analysis
• Two years’ worth of flow data (15-minute or 1-minute intervals)• Analysis of flow rate at each treatment stage • Corrections made for process additions/losses
10
Two-year flow Inlet DAF outlet RGF outlet Distribution input
Total volume (Ml) 256737 250770 260118 256540
Mean daily flow (Mld) 352.2 344.3 356.8 351.9
Volume difference (Mld) -7.9 4.6 -0.3
% of inlet flow 97.8% 101.3% 99.9%
% difference to inlet -2.2% 1.3% -0.1%
Long-term escada and scada analysis
11
Short-term logged data analysis
• Digital dataloggers installed on 21 bulk flow meters
• Ten days’ concurrent data from loggers
• Data from escada/scada used for process additions/losses
12
10-day flow Inlet DAF outlet RGF outlet Distribution input
Total volume (Ml) 3447 3422 3460 3449
Mean daily volume (Mld) 344.7 342.3 346.0 344.9
Volume difference to inlet (Mld) -2.4 1.4 0.3
% of inlet flow 99.3% 100.4% 100.1%
% difference to inlet -0.7% 0.4% 0.1%
Typical daily flows
13
Individual meter analysis
• Comparison of escada/scada, logged, manual integrator and, where applicable, ultrasonic data for each meter
14
• Identify escada/scada scaling and offset issues
• Flag potential meter under- or over-recording
Statistical analysis• Determine whether pairs of data are statistically similar – individual
meters and groups of meters• T-test: to compare the significance of differences• Scatter plot: to identify numerical patterns• Regression analysis: to provide a best fit equation if the above two tests
identify a suitable relationship
• Single meter regression analysis can give an offset value and scaling factor to allow the escada/scada to be reset
15
Paired T for FT3200 Scada - FT3200 Logged
N Mean StDev SE MeanFT3200 Scada 3692 61.514 9.998 0.165FT3200 Logged 3692 59.620 9.969 0.164Difference 3692 1.8938 1.0965 0.0180
Regression Equation
FT3200 Logged = -1.348 + 0.99112 FT3200 Scada
Bias and uncertainty analysis• Detailed analysis of groups of meters using 1-minute data
• Comparison of Inlet meters – DAF meters – RGF meters• DI meters not included as no hydraulic correlation with upstream works
• Spreadsheet with optimisation algorithm used to compare pairs of meter groups• Bias calculated as average difference, with variance• Uncertainty calculated using root mean square methodology of calculated
variances
16
Meter Bias (%) Bias Uncertainty (%)
Combined Inlet 0.25 0.74
Combined DAF -1.08 1.48
Combined RGF 0.28 2.45
Note: bias and uncertainty values above are much lower than those associated with ultrasonic or insertion probe analysis (7-10%)
Meter sizing and accuracy review• Uncertainty calculation based on manufacturers’ headline error for meter
model; flow performance data; installation conditions; age; output type (analogue/digital)• WRc look-up tables used to determine uncertainty values for flow bands.
• Large gravity mains assumed to be in poor performance band but analysis of data showed that flow rates recorded were in the mid range (see FM9 below).
17
Conclusions• The inlet meters were working effectively and gave a good basis for
calculating the flow through the works.
18
Long term scada analysis gave a potential under-recording of 0.3Mld (0.1%)
Short term logged data gave a potential over-recording of 0.3Mld (0.1%)
• The use of multiple complimentary analytical techniques, along with careful analysis of all the process losses and additions, can result in very precise assessments of the flow through the works and into supply.
• Severn Trent Water were able to confirm the efficacy of all 12 distribution input meters and that the data was fit for purpose for the Birmingham Resilience Scheme.