The use of PI and SigmaFine in the Water Industry

1
The use of PI and SigmaFine in the Water Industry

• B. D. Neve

Rex
2
Introduction

• In the last three decades water has increased
  dramatically in value and cost in both the clean
  water and waste utilities.
• This is due to a number of factors, in particular
  environmental issues, population growth,
  urbanization, and in some areas climate change.
• Water will never again be a free commodity, and
  indeed the cost is likely to go on rising at an
  even faster rate.
• The Water and Waste utilities now realize that
  the implementation of industry standard control,
  SCADA, plant information and flow accounting are
  now fully justifiable and that data quality is
  vital to the stewardship of their valuable water
  assets.
• The paper covers the real benefits of
  implementing PI and SigmaFine to cover flow
  balancing and accounting and data quality
  improvements which can be over 2 million per
  year for a 5 million household utility.

3
Overview

• The PI implementation data in this paper comes
  from the experiences of Instem Beaver Valley, the
  distributor of PI in the UK
• The SigmaFine data comes from the experiences of
  the author on 1 pilot and 2 commercial projects
  implementing SigmaFine in the Water industry to
  improve data quality.
• In addition to work on the projects, the author
  conducted a study involving 6 large water
  companies in the UK on the benefits of
  improving data quality via data reconciliation

4
Typical 80,90 and 2000 operational IT structure
5
Problems with the old Architecture

• To much customised hardware and software
• Supplier locks user into high costs and low
  performance compared to new solutions
• Difficult and expensive to expand
• knowledge of the system disappears which leads to
  misuse of data and degradation of data quality
• Result is low data quality to the user
• Bad business decisions,
• E.g.money spent on wrong meters, costly manual
  studies to give robust water balances and on
  leakage detection

6
The Solution

• Change OMS level to PI based standard systems
• Install SigmaFine to audit, test improve data
  quality
• Migrate SCADA to standard PC and Fieldbus based
  systems over time
• This paper majors on improving data quality since
  such a project can have a significant hard ROI
  which can help pay for the other two enabling
  technologies

7
The Final Goal
reconciled, accurate, consistent and auditable
information stakeholders
Overall reconciliation
Distribution Input
Flow balance reconciliation around storage and
distribution
Supply and plumbing losses
Water taken legally unbilled
Distribution losses
Water taken illegally unbilled
Flow balance around treatment plants
Water delivered billed un-measured
Water delivered billed measured
Leakage monitoring and reduction systems
8
The use of PIin the Water Industry
9
PI In Use at Southern Water
Slide courtesy of Instem Beaver Valley
Telemetry Archive -------------------------- 100,0
00 PI points MTDB, PIQA Tag Group
Database WADIS Report Manager PI-SCOPE
Interface PI-CMS Interface
Derived Values Archive ---------------------------
------ 5,000 PI points Derived Values
Calculations Task Scheduler Exception Notification
Servelec Regional Telemetry / SCADA SCOPE-X 4,000
Remote Sites
10
Telemetry Archive System
Slide courtesy of Instem Beaver Valley

• Installed in late 1998 to replace P.ARCH and to
  provide
• Easy access to telemetry data
• Larger SW audience
• More accurate and complete data
• Process data in a time frame meaningful to
  business

11
Telemetry Archive System
Slide courtesy of Instem Beaver Valley

• Server
• 100,000 Point PI Data Archive
• Oracle
• PI Quality Archive
• PI SCOPE Interface
• PI CMS Interface
• Master Translation Database
• Tag Group Database
• Report Database

Client PI-ProcessBook PI-DataLink PI-Manual
Logger Archive Edit PIQA Viewer End-to-End Test
Logging Report Manager Data View Tag Manager Tag
Group manager
12
Derived Values Archive
Slide courtesy of Instem Beaver Valley

• Installed in summer 2000 to provide
• Process Management Water Resources
  Information System
• Integration of data from a number of sources
• PI-UDS, Operational Database, WAACS, ISIS, QXP,
  
• Derivation of meaningful performance indicators
  of treatment processes (PM)
• Maintenance of customer supplies using current
  hydrometric and antecedent conditions (WRIS)

13
DVA Calculations
Slide courtesy of Instem Beaver Valley
Telemetry PI Archive (PI-API)
DVA PI Archive PI-API
Microsoft Excel 25 Calculation Functions Multiple
function calls Multiple sheets Exception Report
Operational Database (ODBC)
Remote Task Scheduler Manager
Microsoft Task Scheduler
14
Thames Water PI System
Slide courtesy of Instem Beaver Valley

• Server
• 10,000 Point PI Data Archive
• ABB Aqua Master Interface
• ABB Gateway
• Radcom Interface
• Radcom Gateway
• DMA Function Sets
• MeterMan

Client PI-ProcessBook PI-DataLink Archive
Edit PIQA Viewer Report Manager Data View Tag
Manager Tag Group manager
15
PI in the water industry

• PI is being used in the UK to supplement and
  replace the existing OMS systems
• As existing OMS systems become more expensive and
  difficult to maintain more systems will transfer
  to cross industry standard systems such as PI
• Meanwhile PI becomes an enabler for SigmaFine
  data reconciliation for improved data quality

16
PI in the water industry

• Lessons Learnt
• PI copes well with the requirements for
• Flexibility
• Expandibility
• PI is a cost effective purchase for water
  companies in terms of
• Initial capital cost
• Whole life cost

17
Improving Data Qualityin the Water Industry
18
The Problems that arise from inadequate data (1)

• Unreliable leak detection and estimation
• Water balancing non closed balances, too much
  guess work, and lack of consistent history
• Investment decisions based on inaccurate and
  inconsistent data
• Difficulties in describing the networks

19
Problems that arise from inadequate data (2)

• Little knowledge of meter accuracy, drift or bias
• Different data in different parts of the company
• Unknown operational/process performance
• Arguments over shared asset agreements

20
The Problems that arise from inadequate data (3)

• Unaccounted for flows
• Difficult and resource consuming reporting to
  Water Regulator
• Difficulties supporting arguments during billing
  disagreements
• Difficulties in justifying increased monitoring
  or improved measurements

21
UK Regulator Ofwat reporting requirements

• Section 2 Chapter 10 of July Return Reporting
  requirements definitions manual
• Water delivered forms the majority of the water
  balance. A company's approach to Table 10 can
  validate any assumptions used to estimate water
  delivered components. Ofwat encourages companies
  to estimate each component of distribution input
  and compare the sum of these with measured
  distribution input. Where there is a small
  discrepancy (say less than I or 2) this can be
  allocated to those components with the greatest
  uncertainty. A large discrepancy suggests that a
  review of a company's estimating process is
  required, as it is clearly not satisfactory for a
  company to be unable to account fully for its
  major product.
• The company should give an explicit explanation
  of any reconciliation adjustment, indicating
  which water balance components have received the
  adjustment using the Maximum Likelihood
  Estimation method. Where the company's
  estimating process has been reviewed the company
  should provide a full briefing outlining the
  degree of the discrepancy, which components were
  reviewed, what assumptions were altered, and is
  so why, and which water balance components needed
  improvement.

22
Ofwat reporting requirements

• To estimate distribution losses (Mld) companies
  should use the Integrated Flow Method. The
  resultant leakage level should then be checked
  against monitored night flows. Companies should
  therefore use the Integrated Flow Method and the
  Minimum Night Flow Method in conjunction, as a
  means to substantiate their estimation of
  leakage.
• Ofwat would also encourage companies to support
  estimates with effective data monitoring systems
  an example would be a domestic consumption
  monitor used by Severn Trent Water to support
  their estimate of unmeasured household per capita
  consumption.
• Ofwat would also expect to see the impact of
  metering on some water delivered components

23
Ofwat reporting requirements

• Distribution input (Mld
• Reliability Grade A The sum of the separately
  estimated water balance components reconcile with
  the measured volume of distribution input to
  within 1-2. There has been no adjustment made
  to measured distribution input other than as a
  result of the aforementioned reconciliation that
  is, the sum of the water balance components with
  measured distribution input. Measured
  distribution input has been estimated from
  water-into-supply meters which record 95 of the
  volume of distribution input, and the meters have
  been used and regularly recalibrated in
  accordance with the manufacturers
  recommendations.
• Reliability Grade B The sum of the separately
  estimated water balance components reconcile with
  the measured volume of distribution input to
  within 5 but not to within 2. There has been
  no adjustment made to measured distribution
  input, other than as a result of the
  aforementioned reconciliation that is, the sum
  of the water balance components with measured
  distribution input. Measured distribution input
  has been estimated from water-into-supply meters
  which record 90 of the volume of distribution
  input, and the meters have been used and
  regularly recalibrated in accordance with the
  manufacturers recommendations.

24
Ofwat reporting requirements

• Overall water balance
• Reliability Grade A The water balance components
  reconcile with measured distribution input to
  within 2. An explicit explanation for any
  reconciliation adjustment is given and an
  adjustment has been made to distribution input or
  has been distributed between water balance
  components. Water-into-supply meters have been
  used and recalibrated in accordance with the
  manufacturers recommendation. The water balance
  components have been separately estimated and
  reconcile with the equivalent residual of the
  water balance. 90 of the volume of distribution
  input (not including distribution input) has been
  awarded a reliability band of A or B within the
  separately estimated water balance components.
• Reliability Grade B The water balance components
  do not reconcile with measured distribution input
  to within 5, hence an adjustment has been made
  to distribution input or has been distributed
  between water balance components using the
  Maximum Likelihood Estimation technique.
  Water-into-supply meters have been used and
  recalibrated in accordance with the manufacturers
  recommendation. The water balance components
  have been separately estimated and reconcile with
  the equivalent residual of the water balance. 90
  of the volume of distribution input should have
  been awarded a reliability band of A or B within
  the separately estimated water balance components.

25
Sources of data quality problems

• Measurement/metering errors
• Plant/Network errors
• Hidden flows or leaks
• Un-metered flows
• Un-measured inventory changes
• Dynamic effects
• Data processing errors

26
Sources of Flow MeasurementError

• Installation Effects
• Precision
• Fouling
• Fossilized Bias (buttered toast)

27
Measurement Uncertainty
Daily performance
28
Data Processing Errors

• Manual data entry systems
• Multiple values for single data points
• Incorrect engineering calculations
• Lack of time synchronization measurements
• Different end of period for accounting and
  engineering
• Data is historized and stored in multiple
  locations
• Data is changed and fixed by multiple
  functional areas
• Supply, Distribution, Planning, Engineering
  Accounting
• Control

29
How to improve the data

• Carry out a top down data quality improvement
  project
• Use Data Reconciliation as an integral element
• A proven method from the Oil and Petrochemical
  industries

30
Conventional Wisdom on Data Quality

• Engineering and accounting data are different
• Meter errors balance out in the long run
• Volume balances are the same as mass balances
• Mass balances are simple
• Mass balances are impossible
• Manual estimates are not important
• Accounting data does not matter
• Custody transfer measurements are correct
• Inventory measurements have no variance

31
Measures of Data Quality

• Completeness
• Meters, inventories, Transactions, Composition,
  Densities
• Redundancy
• How many times is the same volume measured
• Precision
• What is the variance of the measurement device
• Accuracy
• How is the measurement compared to a standard

32
Data Quality Analytical Tools

• Expert systems
• Neural networks
• Reconciliation systems

33
Expert Systems

• Rules of Thumb
• Complex to build and maintain relationships
• Useful for gross error detection

34
Neural Networks

• Recognizes patterns
• Model setup is important
• Accurate to a few percent
• Useful for gross error detection

35
Data reconciliation

• Data reconciliation is a systematic way of using
  all the available information about a process or
  system or business to improve consistency and
  accuracy
• Very often some information is overlooked
• This information can be flows, inventories,
  levels, , meter accuracies, loss estimates and
  equations i.e.. mass balances, component
  balances, energy balances
• Sigmafine is an advanced data reconciliation
  package designed for the process and utility
  industries

36
The theory behind data reconciliation
Reconciled value i.e.. best estimate of value
consistent with all information
4.5 ML/day
Flow Meter
Level Meter
Integrated flow reading
Change in inventory
Delta Level
Average Area
Tolerance of flow meter
Tolerance of level measurement
Tolerance of reconciled value
0.0 ML/day
37
The mathematics

• The SigmaFine Data reconciliation algorithm
  distributes all the errors in proportion to the
  confidences on the data (e.g. meter readings) so
  that
• All the balances are precisely satisfied
• the total sum of the perturbations on the data
  is minimised
• The sum is the squared deviation normalised by
  the confidence on each piece of the data.
• This is a large constrained minimum sum of errors
  squared problem and uses a Kalman filtering
  algorithm.

38
The history of data reconciliation

• Data reconciliation has been used for 20 years in
  the Oil and Petrochemical sector
• It produces accurate material, energy and
  component balances
• It helps the accountants track expensive feed,
  intermediates and products and account for losses
  
• Before SigmaFine, data reconciliation was
  expensive and cumbersome to use

39
What is the SigmaFine package
NETWORK MODELS AND CASES
RECONCILEDDATA
OTHER PROGRAMS
APPLICATIONS
SQL
AD HOC REPORTS
PI
DATA RECONCILIATION
SIGMAFINE
The optimised reconciliation algorithm
HISTORICAL
Process Book
REPORTS

• WATER BALANCES
• LEAK ESTIMATES
• INVENTORIES
• ETC.

PUMPS
OTHER INPUTS
RESERVOIRS
FLOW METERS
MANUAL DATA
NIGHT LINE DATA
40
How SigmaFine deals with inventories
For irregularly shaped tanks or reservoirs,
SigmaFine has an automatic built in strapping
feature
The vessel is divided into a number of slices
and each slice has an area associated with it.
The program interpolates linearly between the
slices to calculate the area at any depth and
thus the change in volume for any change in
depth.
41
Typical applications

• Accounting mass/water balance
• Operational water balance
• Leakage estimation and tracking
• Suspect meter reports/meter proving
• Dosing component balance
• Dosing accuracy improvement
• Shared processing/asset agreements

42
Applications continued

• Preparation of data to Regulator
• Water stock monitoring/reporting
• Recovery support after upset
• Network description and documentation
• Training of operational personnel
• Historical performance reporting

43
Applications continued

• Mass, volume and component balances at treatment
  plants
• Mass and volume balances around sewage works
• Improved process knowledge
• Diurnal flow estimation and balancing
• Adverse trend detection, e.g. solids build-up

44
Applications continued

• Identification of problems
• Instrument/meter problems
• Badly installed, faulty, or biased meters
• Faulty calibration or instrument drift
• Missing measurements
• Model or network knowledge errors
• unaccounted or missing flows
• incorrect association of data
• incorrect time stamping

45
Implementation

• Attend training course (3 days) Install on target
  desktop computer
• Develop initial model/network (few days to few
  months depending on size)
• Set up auto transfer of data PI, and accounting
  systems
• Debug model, test data
• Develop and enlarge model in line with business
  needs. Migrate to larger machine or network.

46
SIGMAfine MonthlyOperations
Overall reconciliation
Distribution Input
Flow balance reconciliation around storage and
distribution
Supply and plumbing losses
Water taken legally unbilled
Distribution losses
Water taken illegally unbilled
Water delivered billed un-measured
Flow balance around treatment plants
Water delivered billed measured
47
Using Sigmafine
balance area
balance area
balance area
balance area
balance area
48

• SigmaFine lets you describe a network of
  treatment works, pipes, pumps, reservoirs and
  meters as a live intelligent graphic which you
  can change at any time
• The reconciled balance formula are derived
  automatically from this picture when you run a
  balance

49

• SigmaFine lets you build up your balances from
  small local zones through district to division
  and company wide balances or vice versa
• Everyone can access a standard updated network
  and can alter their own local copy for test runs,
  feasibility studies, investment decisions etc

District balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
balance area
50
Typical SigmaFine Screen
51
Building a Model
C1 zone
C2 zone
Icons
C3 zone
shaft
LD 2
Treatment works
RX1 sr
RX2 sr
Distribution zone
Shaft 1
Pump station
Shaft 2
Shaft 3
C4 zone
52
(No Transcript)
53
(No Transcript)
54
Catchment 1 dem
Rx10 to C1
Catchment 1 zone
C3 to RX1
C3 to RX3
C4to FX2
C5 to RX6
55
Catchment 1 dem
Rx10 to C1
Catchment 1 zone
C3 to RX1
C3 to RX3
C4to FX2
C5 to RX6
56
Experience so Far with SigmaFine in the Water
industry

• Three projects so far
• Large ring main flow balance (circa 90 meters)
• Distribution area flow balance
• Treatment works flow balance

57
Lessons learnt

• Treatment plant have lots of redundancy in meters
  and this can be used to significantly increase
  the quality (i.e.accuracy) of the flows into the
  distribution areas. This can go from a tolerance
  of /-5 or worse to to -.5
• SigmaFine data reconciliation studies should be
  done before planning new (distribution) meter
  projects
• This can reduce number of meters and ensure they
  are in the optimum location. This can save once
  of costs in millions!!
• There are many intangible benefits form
  implementing data reconciliation projects
  including better retention of network knowledge

58
Lessons learnt

• Many meters have biases or slope errors
• These can be soft calibrated using results if
  reconciliation
• Bad meter detection is very useful since all
  water companies have shortages of maintenance
  man-hours
• Large undetected flows or flows in opposite
  direction to anticipated can be present
  especially in old networks.
• Consistent knowledge of network topology is rare

59
Average meter performance
60
Meter with bias and slope error
61
Looks like a bad meter but flow is low
62
Just some calibration errors
63
A sticking meter
64
Total to zones demand in region
Tr w 1 to rm Tre w 2 torlm Trw w 3 to lm C1 to z1
C2 to z1 C3 to z3 C2 to z1 C3 to s1 Sh1 toz1 Sh
1 to z2 Sh 2 to z1 Sh 3 to z3 Sh 4 to z4 sh 5 to
z1 sh 5 th z2 Sh 2 to z 2 Sh 4 to z3 Sh 6 th
z1 Sh7 to z2 Sh 10 to z1 Sh 2 to z2 Sh 5 to z1 Sh
2 to z2 Sh 6 to z1 Sh 3 to z3 Sh 3 to z2 Sh 5 to
z2 C1 to c3
Zone 1 demand Zone 2 demand Zone 3 demand Zone 4
demand Zone 5 demand Zone 6 demand Zone 7
demand Zone 8 demand Zone 9 demand Zone
10demand Zone 11 demand Zone 11 demand Zone 12
demand
Zone 1 z2 Zone 2 to z3 Zone 3 to z4 Zone 6 to
zone2 Rs1 to rx2 Ps2 to ps3 C1 to c3 C4 to c6 C4
to c6 Zone 2 to z 10 X to x1 Des to z3 Rx2 to
rx4 Rx5 to ps2 Ps3 to sx Ps3 to rxy Ps4 to ps4 Xs
rto lf2
65
Cost benefit analysis

• A single undetected leak of treated water can
  cost 50,000 a year
• An investment decision made too soon due to
  inaccurate data can cost many K per month in
  interest alone
• Improved data quality can payback can be very
  fast
• Detailed cost benefit analyses can be provided

66
Benefits
The following benefit calculations are based on
one large municipal water company supplying 5
million households with a cost of 75c per
household per day and a value of water delivered
of 1.5 per M3. Leakage rate is assumed to be at
7 litres per hour per property with a target
leakage rate ( i.e. where the cost of further
reductions balances the cost of repairing leaks)
of 4 Litres per hour per property
67
Conclusions

• Data Quality is an endemic problem in the water
  industry that needs addressing
• Together PI and SigmaFine can help solve this
  problem and significantly improve business
  operations and profitability
• Tangible benefits can be very large and can be
  identified.

68
PI in the Water sector
For More information on existing applications of
PI in the UK Water sector contact

• David Rees
• Instem Beaver Valley
• 2 Watermoor Road
• Cirencester
• Gloucestershire
• CL7 1JN
•  
• Tel [phone]
• Email reesd_at_instem.com

69
Sigmafine in the Water sector
For More information on existing applications of
SigmaFine in the Water sector contact

• Brian Neve
• e-mail bdn_at_rexsoft.com or brian_neve_at_mail.com
• Telephone [phone]
• Direct line 44 (0)2380 745 920
• Direct Fax [phone]
• Mobile [phone]
• Rex Software Limited
• Chesil House
• Shakespeare Road
• Eastleigh
• Southampton SO50 4SY