Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
The use of artificial intelligence techniques to predict the flow in sewers and at CSOs.
Adrian Saul and Will ShepherdPennine Water Group, University of Sheffield
Water Industry Forum 18th May 2011
Contents
• Introduction• Background to data collection in 2
catchments• Application of AI to data and
diagnostic tools• Summary comment
Introduction• In 2005 YW set up a Strategic Partnership
with the University of Sheffield• As part of this partnership two strategic
research catchments were established• Long term monitoring of rainfall, sewer
flow and CSO flow depth• This presentation highlights the
application of Artificial Intelligence techniques that use this data to enhance understanding of system performance
Catchment 1• Population ~ 17,000. Main town + large village + 2
small villages.• Town on steep hill, villages in valley bottom.• Village flows pumped to WwTW in town.
Flow monitoring
• Ultrasonic Doppler velocity sensor.
• Ultrasonic downward looking depth sensor.
• Pressure sensor for surcharged depths and provides redundancy.
• Recorded at 2 minute resolution.
Depth monitoring
• CSOs – two types of ultrasonic depth monitor, one battery powered, one mains.– 5 or 15 minute resolution.
• Tanks use flow meters without velocity sensor.– 2 minute resolution.
Rain gauges
•Tipping bucket rain gauges, 0.2 mm depth per tip.•Tips aggregated into 2 minute rainfall intensities.
Catchment 2• Population ~
15,000. Villages / Suburbs upstream of city.
• Mainly in steep valley with some flatter areas.
• Several off-line storage tanks
Domestic Gully Sensors
Post Mounted Hub
GPRS Internet
Web Based Data Interaction
Web Server /Data Backup
Reports
Direct Alarms
Alarms
Rainfall Radar Data
EquipmentService
Simple, five level Sewer monitors
Business Planning and Development
Sewer Ops Management
Field Engineers -Installation,
Service,Etc.
Depth/Flow Logger
Low Cost, Event Based Sensors
StatusConfiguration
Location
Low cost rain gauge
Direct Reports
Future Vision
Data Data ––Knowledge Knowledge --
ActionAction
Data in real time Data in real time –– rainfall, flow, pollutionrainfall, flow, pollution
Fast simulation of system performanceFast simulation of system performance
Predictive modelling Predictive modelling –– artificial intelligenceartificial intelligence
Proactive management, operations and Proactive management, operations and interventionsinterventions
Sensor optimisation
Daily patterns
0.000
0.002
0.004
0.006
0.008
0.010
0.012
0.014
0.016
0.018
0.020
00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00
Time
Flow
(m3 /s
)
Mon Tues Weds Thurs Fri Sat Sun
CSO level prediction
Data inputData at 5 to 15 minute time steps
Rainfall Hawkeye depth Sewer Flow
Neural NetworkData used to train model (more data better model)
Output data Prediction of depth 15 minutes into future
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000250
300
350
400
450
500
550
Monitoring time steps (5 mins interval)
Cha
mbe
r wat
er le
vel (
mm
)
Model prediction output sample from Carleton Rd Skepton CSO
Prediected valueActaully value
Chamber weir height
Chamber weir
Prediction model output sample
Blockage Detection
0
2
4
6
8
10
12
14
16
18
20
21/05/08 26/05/08 31/05/08 05/06/08 10/06/08 15/06/08 20/06/08 25/06/08Time
Flow
(l/s
)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Velo
city
(m/s
) & D
epth
(m)
Flow (l/s)Depth (m)Velocity (m/s)
A typical data set showing the relationship between rainfall and depth but with system failures Diagnostic fuzzy logic techniques are used to identify irregularperformance in CSO’s with and without screensFailure has now been linked to operational performance and strategies for proactive maintenance identified
Summary CommentAI techniques have been applied to sewer system data and it has been feasible to:1 Predict sewer flow and CSO depth using measured rainfall2 To develop diagnostic tools to identify unexpected changes in system performance – changes in DWF, rapid blockage or gradual build-up of water depth.3 To develop diagnostic tools to identify differences in measured and predicted wet weather performance4 To link the performance of CSO chambers and CSO chambers with screens to maintenance frequency5 To develop strategies for proactive maintenance
Thankyou