EnWin Utilities Reduces Water Main Breaks by 21 Percent

Copyright © 2014 Rockwell Automation, Inc. All rights reserved.

Tweet About It Using #PSUG

EnWin Utilities Reduces Water Main Breaks by 21 Percent

with Model Predictive Control

Garry Rossi, Director, Water Production – EnWin Utilities Edward Scott Canadian Optimization Manager – Rockwell Automation

November 2014


#PSUG

Agenda

Summary

The Results

The Opportunity

System Overview and Fast Facts

Mission Statement


#PSUG

Windsor Utilities Commission Mission Statement


#PSUG

Drinking Water System Overview

• Total daily supply capacity: 349 ML (92MGD)

• Reservoir storage capacity: 118 ML (31MG)

• Number of treatment plants: 2

• Number of pumping stations: 3

• Number of Elevated Storage Tanks: 2

• Length of water main: 1,100 km (690 Miles)

• The Albert H. Weeks Water Treatment Plant supplies an average of 140 ML (37 MG) of water to City residents per day.


#PSUG

System Fast Facts

• 17% Non-Revenue Water

• 66 psi average operating pressure (at pump station)

• Single pressure zone

• 17 Pressure stations

• Fully redundant SCADA system architecture (2010)

• Fibre communication ring throughout the service area


#PSUG

The Opportunity

• 238 Main Breaks per year (average)

• 44 average age of distribution water main (one of the oldest in Ontario/Canada)

• Increasing Electricity costs

• Inconsistent system pressures during peak/low demand periods


#PSUG

The Opportunity

Discovered predictive control

technology applied in

manufacturing

What is it and can the technology

be applied in the drinking water

sector?

What can Model Predictive Control

(MPC) do for WUC?

Entered into a collaborative project

with Rockwell Automation to

leverage this technology


#PSUG

The Opportunity

Process Variation = inefficiencies = increased customer costs


#PSUG

The Opportunity

Why the Variation?

Inconsistency in operating

Too many parameters to monitor simultaneously

Changing one variable affects many variables

Previous Controls

Train staff more?

More procedures?

Simple PID Control?


#PSUG

The Solution

How do we Implement this solution effectively?

Review current control strategy for processes.

Narrow the focus to the highest ROI for the operation.

Implement over several phases to reduce risk.

Advanced Process Control

Real-time Optimization

Model Predictive Control

Nonlinear Multivariable Control

Linear Multivariable Control

Inferential Sensors

Advanced Regulatory Control

Regulatory Control

• Scalable to meet a wide range of advanced process control requirements

• Improve yield, increase throughput, reduce costs, and improve quality

• Economic optimization

• Maximize throughput

• Increase yield

• Reduce variability

• Improve product transition

• Virtual lab readings

• Gain scheduling

• Rule-based control

• Loops with long lag times

• Basic loop PID control Incre

asin

g E

ffo

rt &

In

cre

as

ing

Valu

e

Single variable in & single variable out • Set up a target and control process variable to the target.

What would be a best SP is often unknown

PID

So What Makes MPC Different?

12

MPC

Multivariable in & multivariable out • Control strategy based on a cross correlated matrix

of key process variables

Feedback control • The controller will take no action unless the errors show in PV

Predictive control • Dynamic models developed through process step tests

• Future process behavior prediction and feed forward control

Indirect property variable control • Control property variables through temperature, pressure or flow rate,

etc.

Direct property variable control • Property targets could be set and control directly

Poor control quality when there are large time delays or complex dynamic interactions

PID setpoints are not necessarily optimal with respect to plant wide objectives

Optimal PID targets are continuously calculated, set, & modified

Explicit dynamic models leverage computer capabilities to calculate impact to the product

Model Predictive Control (MPC)

MV setpoint

MV measured value

MV prediction

CV prediction

CV measured value Controller optimizes future trajectories

using an internal simulation

Use first step of horizon

MV desired value

CV desired value

MV desired level specified manually

or by steady-state optimizer

Current time indicator

Past Future


PlantPAx MPC – Solution Development

14

Engineering Workstation

AOI (.L5X)

Programming in the familiar Logix5000 environment

Identification

Control Task

Specs

Monitoring/Tuning Online change monitoring

Logix5000 program download

Logix5000 MPC Builder

Development cycle

1. Identify Process Parameters

2. Specify Objective Function & Constraints

3. Generate MPC AOI

4. Import and instantiate AOI

5. Download project to Logix

6. Monitor and tune

How MPC Generates Benefits

104

102

100

98

96

94

92

90

Process/Specification Limit

Before

During

After

Variability

reduction with

Pavilion MPC

Variability under

operator control

Benefits

Time

• Reduces Variability

• Achieves “Plant Obedience”

• Manages the process within constraints

• Achieves uplift – operate closer to specifications and

performance limits while maintaining safety margins

Push process

toward limits

Maintain quality

just within

specification


#PSUG

The Results

Shows pump start/stop to meet flow demand

Fixed demand operator dependant Note the step changes Large variation in system pressure


#PSUG

The Results

Pump start/stop pressure amplitude reduced

Less operator dependant Reduced step changes Reduction in system pressure variation


#PSUG

The Results

Pump start/stop improved via FCV’s

Eliminated step changes Reduction in system pressure variation


#PSUG

The Results


#PSUG

The Results


#PSUG

The Results

Pump start/stop cycles


#PSUG

The Results

When were the pumps started?


#PSUG

The Results

0

5

10

15

20

25

30

35

40

45

50

January February March April May June July August September October November December

2012-2014 Main Breaks

Pre-MPC Post-MPC Avg Breaks


#PSUG

The Results

-12

-10

-8

-6

-4

-2

0

2

0

10

20

30

40

50

60

January February March

2014 vs Coldest Average Mean Temperature

Pre-MPC

Post-MPC

Pre MPC Avg Temp

Post MPC Avg Temp


PlantPAx MPC Implementation

25

Header Pressure Logix-Based

MPC Controller

VFD Speed SP

VFD Speed SP

FCV SPs

M1 M2 M3 M4 M5 M6 M7 M8

Pressure Meters

Meter Pressure Logix-Based

MPC Controller

Pressure SP

Legend

MPC Controls

Meter Pressure Logix-Based

MPC Controller

Pressure SP

Header Pressure Logix-Based

MPC Controller


#PSUG

Summary

Lessons Learned

• Pressure station data affected

pressure control model

• Additional VFD’s improve overall

control

Copyright © 2014 Rockwell Automation, Inc. All Rights Reserved. PUBLIC INFORMATION

Logix Soft Sensor®

A Range of Advance Control Solutions

REGULATE

VALUE ADD

Single Loop

Multiple Loops

Process Unit

Multiple Units

Plant-wide Area

PREDICT CONTROL OPTIMIZE

27

PlantPAx & Pavilion8® MPC Pavilion™ RTO


#PSUG

Summary

Next steps

• Add VFD control at the George

high lift station

• Incorporate this new VFD into the

model

• Evaluate opportunities for MPC

control in other areas


#PSUG

Summary

Operational Improvement Summary

• Reduced average system mean

pressure by 2.8% resulting in $125K

savings

• Reduced standard deviation by 29%

• Added multi-zone control for outlying

areas

• Eliminated pump start/stop variations

• Reduced main breaks by 21%

resulting in $125k savings

• ROI less than one year


Tweet About It Using #PSUG

Questions

Garry Rossi, Director, Water Production – EnWin Utilities Edward Scott Canadian Optimization Manager – Rockwell Automation

November 2014

Technology

EnWin Utilities Reduces Water Main Breaks by 21 Percent