Water and wastewater systems 0.2cm in the era of data ... · Conclusion Water and wastewater systems in the era of data explosion ... Final seminar 2012 Water and wastewater systems

OutlineDrinking water network

Sewer networkWastewater treatment plants

A full-scale exampleConclusion

Water and wastewater systems

in the era of data explosion

Michela Mulas

Water and Environmental Engineering

Dept. of Civil and Environmental Engineering

Aalto University

BioRefine and Vesi - Final seminar 2012 Water and wastewater systems in the era of data explosion




MotivationsWater treatment facilities are continuously challenged to satisfy new constraints interms of quality of the effluents for compliance with stringent environmentalregulations, sustainable reuse and cost optimization.

Data explosion

The facilities are becoming modern with a vast amount ofon-line and off-line measurements collected routinelyand the operators and process engineers are becomingmore experienced in instrumentation and automationconcepts.

Data are used for decision support , for early warning ofdisturbances and process changes , for trackinginteresting and relevant parameters as well as for thebasis of control actions .

ObjectiveDiscuss the potential of encapsulating data-enhanced process knowledge

and modelling capability in automation systems.






Data explosion










Data explosion









Outline

Introduction

Drinking water networkSävel research project

Sewer networkEfeSus research project

Wastewater treatment plantsDiamond research project

A full-scale exampleThe Viikinmäki WWTP

Conclusion





Sävel research project

Water mains managementThe effect of aging on the water distribution network are very ha rd to detect , because pipesare buried below the level of soil frost penetration.

Sävel project

◮ Funded by Tekes◮ Research parties

◮ Aalto UniversityRiku Vahala and Kia Aksela

◮ National Institute of Health and WelfareWater and Health UnitIlkka Miettinen

◮ Water companiesCompanies and associations

◮ Project duration : 1 July 2008 - 30 June 2012

Goals◮ Develop a system that detects network malfunctions more effectively and

respond to them with greater precision.◮ Detect disturbances by analyzing continuous measurements, water-use

forecast and network model.

◮ Improve the management of the network asset.





EfeSus research project

Sewer network renovationSewer are valuable assets . There is the need to systematic approach to renovation planningas well as for comprehensive data collection and automated t reatment of data .

EfeSus project

◮ Funded by Tekes◮ Research parties

◮ Aalto University:Riku Vahala and Tuija Laakso

◮ Finnish Meteorological Institute:Jarmo Koistinen

◮ University of Exeter (UK): Dragan Savic◮ Companies

Wastewater utilities◮ Project duration : 1 June 2011 - 30 June 2014

Goals◮ Improve the prioritization of renovation and repair activities - what to

renovate and when.◮ Develop the use and the processing of existing data resources.

◮ Create a software prototype for assessing the different network areas.





Diamond research project

Wastewater treatment plants

The lack of appropriate data management tools is clearly limiting a broader implementation andefficient use of new sensors, monitoring systems and advanced process controllers.

Diamond project

◮ Funded by FP7 Capacities:Research for the benefits of SMEs

◮ RTD performers :◮ CEIT (Spain)◮ Uppsala Universitet (Sweden)◮ IVL Svenska Miljoe Institutet AB (Sweden)◮ Aalto University (Finland)

◮ Small and Medium Enterprises:◮ Mondragón Sistemas de Información (Spain)◮ Cerlic Controls AB (Sweden)◮ Mipro Oy (Finland)

◮ End-Users:◮ Aguas de Gipuzkoa S.A (Spain)◮ Stockholm Vatten AB (Sweden)

◮ Project duration:1 September 2012 - 31 August 2014





Diamond research project

Wastewater treatment plants

DIAMOND

AdvanceD data management and InformAtics for the optimuM operatiON anDcontrol of wastewater treatment plants

◮ Processing, centralising, synthesizing, correcting and completing allthe heterogeneous data available in a WWTP.

◮ Interpreting and extracting the maximum information from theseplant data.

◮ Constructing new information from existing measurable variables.◮ Facilitating the decision-making process of the operators of the

WWTP.

◮ Facilitating the design and implementation of plant-wide operationalstrategies and automatic controllers.

Optimize the operation of wastewater systems by adequately managingand using all the information available in the plant.





The Viikinmäki WWTP

Wastewater treatment plants: A full-scale example

Viikinmäki project

◮ Funded by MVTT◮ Aalto University :

◮ Water and Environmental Engineering◮ Environmental and Industrial Machine Learning

◮ Viikinmäki’s personnel

The wastewater treatment line consists of:◮ bar screening and grit removal◮ pre-aeration and primary sedimentation◮ activated sludge process (8 lines)◮ secondary sedimentation

◮ biological post-filtration (post-denitrification)

The sludge treatment line has mesophilicdigesters and dewatering systems

Retrieved fromwww.flickr.com/photos/sameli/sets/72157607704932823/

Monitoring nitrate concentrations in the denitrifying pos t-filtration unit.


www.flickr.com/photos/sameli/sets/72157607704932823/





Wastewater treatment plants: A full-scale example

Viikinmäki project

◮ Funded by MVTT◮ Aalto University :

◮ Water and Environmental Engineering◮ Environmental and Industrial Machine Learning

◮ Viikinmäki’s personnel

The wastewater treatment line consists of:◮ bar screening and grit removal◮ pre-aeration and primary sedimentation◮ activated sludge process (8 lines)◮ secondary sedimentation

◮ biological post-filtration (post-denitrification)

The sludge treatment line has mesophilicdigesters and dewatering systems


Monitoring nitrate concentrations in the denitrifying pos t-filtration unit.







Process description: The denitrifying post-filtration unit

◮ Overall a total nitrogen removal of 90%

◮ Ten Biostyr R© filters arranged in parallel

◮ The influent wastewater is equally distributed

◮ Before each cell, the incoming flow is split in two

◮ Attached biomass tends to clog the cell

◮ Periodic backwashes with effluent wastewaterand a counter-current air flow

◮ To favor the removal, methanol is dosed with afeedback loop policy

◮ Dosing according to the nitrate concentration inthe filters, measured on-line with analyticalinstruments

◮ Treated wastewater is discharged into a commoneffluent channel





































Proper functioning of the nitrate sensorsis of crucial importance

from an environmental and aneconomical point of view

but harsh environmental conditionsexpose these instruments to malfunction

One big problem is vicinity of the analyzersto the effluent channel

◮ Methanol control is compromised◮ Need for a back-up instrument


Main goalDevelop an array of soft-sensors that estimatein real-time the nitrate concentration in the cells

◮ Accurate and computationally lightmodels are the priority

◮ Starting from easy to measureprocess variables








Proper functioning of the nitrate sensorsis of crucial importance

from an environmental and aneconomical point of view

but harsh environmental conditionsexpose these instruments to malfunction

One big problem is vicinity of the analyzersto the effluent channel

◮ Methanol control is compromised◮ Need for a back-up instrument


Main goalDevelop an array of soft-sensors that estimatein real-time the nitrate concentration in the cells

◮ Accurate and computationally lightmodels are the priority

◮ Starting from easy to measureprocess variables







Soft-sensors based on unstructured modelsBased on a more-or-less accurate description of relationsh ips between data

An unstructured software sensor is often seen as input-output model

◮ the inputs X are easy to measure

◮ the output y are hard to measure

Assuming the existence of a functional relationshipbetween the inputs and the outputs (y = f (X)+ ε),

the model is calibrated to reconstruct it (̂f )

Unstructured models rely on methods for dataanalysis:

◮ Techniques for sample selection◮ Techniques for variable selection

◮ Techniques for regression

KISS: Start with a simple model type






Soft-sensor design

A set of process measurements hasbeen collected:

◮ 3 years of continuous operations(2008 - 2010), hourly averages

Sample selection

The overall number of available processvariable relevant to the task is 142:

◮ 7 for the influent◮ 5 for the effluent

◮ (12+1)×10 for the filters

Variable selection

TAG Description UnitsI-NO3-1 Influent Nitrate-Nitrogen (sensor 1) mg/lI-NO3-2 Influent Nitrate-Nitrogen (sensor 2) mg/lI-SS-1 Influent Suspended Solids (sensor 1) mg/lI-SS-2 Influent Suspended Solids (sensor 2) mg/lI-O2 Influent Dissolved Oxygen mg/lI-PO Influent Phosphate-Phosphorus mg/lI-TP Influent Total Phosphorus mg/lE -NO3 Effluent Nitrate-Nitrogen mg/l

E -TOC Effluent Total Organic Carbon mg/lE -PO Effluent Phosphate-Phosphorus mg/lE -TP Effluent Total Phosphorus mg/lE -T Effluent Temperature ◦C

Fi-QWW i-th Filter Backwashing water flowrate m3/sFi-QWA i-th Filter Backwashing air flowrate m3/s

Fi-QW -1 i-th Filter Wastewater flowrate (line 1) m3/s

Fi-QW -2 i-th Filter Wastewater flowrate (line 2) m3/sFi-QM-1 i-th Filter Methanol flowrate (line 1) m3/h

Fi-QM-2 i-th Filter Methanol flowrate (line 2) m3/hFi-P-1 i-th Filter Pressure at the bottom kPaFi-P-2 i-th Filter Pressure at the top kPaFi-NO3 i-th Filter Nitrate-Nitrogen mg/lFi-HL i-th Filter Head-Loss mFi-CR i-th Filter Clogging rate %Fi-HRU i-th Filter Hour in use −

Fi-ITW i-th Filter Intermediate time of backwash −

Simplicity is on of the main requirements to allow adirect implementation in the plant’s control system. Regression models






Soft-sensor design



Sample selection




Variable selection














Soft-sensor design



Sample selection




Variable selection














Soft-sensor design



Sample selection




Variable selection














Soft-sensor design



Sample selection




Variable selection














Soft-sensor design



Sample selection




Variable selection














Estimation performanceThe hardware sensor in Filter 9 is not returning any measurement.

02.Oct.09 03.Oct.09 04.Oct.09 06.Oct.09 07.Oct.09

0.2

0.4

0.6

0.8

F9−QW


0.2

0.4

0.6

0.8

F3−QW

The missing measurements can be replaced by the soft-sensor estimates








0.2

0.4

0.6

0.8

F9−QW


0.5

1

1.5

2

F9−

NO

3

Measured


0.2

0.4

0.6

0.8

F3−QW


0.5

1

1.5

2

F3−

NO

3

Measured









0.2

0.4

0.6

0.8

F9−QW


0.5

1

1.5

2

F9−

NO

3

Measured


0.5

1

1.5

2

F9−

NO

3

k−NN LLROLSR


0.2

0.4

0.6

0.8

F3−QW


0.5

1

1.5

2

F3−

NO

3

Measured


0.5

1

1.5

2

F3−

NO

3

k−NN LLROLSRMeasured









0.2

0.4

0.6

0.8

F9−QW


0.5

1

1.5

2

F9−

NO

3

Measured


0.5

1

1.5

2

F9−

NO

3

k−NN LLROLSR


0.2

0.4

0.6

0.8

F3−QW


0.5

1

1.5

2

F3−

NO

3

Measured


0.5

1

1.5

2

F3−

NO

3








Estimation performanceErroneous measurements are returned by the hardware sensor in Filter 6.

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F6−

QW

F6−QW

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F3−

QW

F3−QW

The soft-sensor is capable of recovering the nitrate concen tration







16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F6−

QW

F6−QW

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F6−

NO

3

Measured

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F3−

QW

F3−QW

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F3−

NO

3

Measured








16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F6−

QW

F6−QW

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F6−

NO

3

Measured

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F6−

NO

3


16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.10

0.1

0.2

0.3

0.4

F3−

QW

F3−QW

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F3−

NO

3

Measured

16.Mar.10 17.Mar.10 18.Mar.10 19.Mar.100

0.5

1

F3−

NO

3







Conclusion

Viikinmäki project

Potential benefits for the WWTP supervisionand monitoring

◮ Back-up system to conventional analyticalequipment for replacing out-of-order components.

◮ Validation tools for existing field measurements.

We discussed the potential of encapsulating data-enhanced process knowledge and modell ingcapability in automation systems through different project examples:

◮ Sävel : Water supply real-time management◮ Efesus : Effective sewer condition management using online sensor information

◮ Diamond : Advanced data management and informatics for the optimum operation and controlof wastewater treatment plants





Conclusion

Viikinmäki project

Potential benefits for the WWTP supervisionand monitoring

◮ Back-up system to conventional analyticalequipment for replacing out-of-order components.

◮ Validation tools for existing field measurements.

We discussed the potential of encapsulating data-enhanced process knowledge and modell ingcapability in automation systems through different project examples:

◮ Sävel : Water supply real-time management◮ Efesus : Effective sewer condition management using online sensor information

◮ Diamond : Advanced data management and informatics for the optimum operation and controlof wastewater treatment plants


Documents

Water and wastewater systems 0.2cm in the era of data ... · Conclusion Water and wastewater systems in the era of data explosion ... Final seminar 2012 Water and wastewater systems