Methods for Identifying Critical Temperature forControl of Low-Temperature DH Systems
Hjorleifur G. Bergsteinsson1, Henrik Madsen1, Jan K. Moller1, HenrikAa. Nielsen2, Markus Falkvall3, David Edsbacker3
1 DTU Compute2 ENFOR
3 Kraftringen
11/09/2019
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 1 / 19
Outline
Digitalization in District HeatingLoad Forecasting in District HeatingTemperature Optimization in District HeatingUsage of digitalization in District Heating
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 2 / 19
Digitalization in District HeatingSm
art M
eter
Dat
a
Critical Point
Data rich DH system:Large amount of available data from theproduction, end-users and critical points inthe network.Measurements of supply temp., return temp.,flow, pressure, etc.Smart meter at the end-user are usually justused to charge the customer.Critical points are usually used to controlpressure difference between supply and returnfor hydraulic modeling.
How can DH create value for the company and customers by utilizing all ofthe available data?
DATA DRIVEN MODELS
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 3 / 19
Digitalization in District HeatingSm
art M
eter
Dat
a
Critical Point
Data rich DH system:Large amount of available data from theproduction, end-users and critical points inthe network.Measurements of supply temp., return temp.,flow, pressure, etc.Smart meter at the end-user are usually justused to charge the customer.Critical points are usually used to controlpressure difference between supply and returnfor hydraulic modeling.
How can DH create value for the company and customers by utilizing all ofthe available data?
DATA DRIVEN MODELS
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 3 / 19
Methodologies for Data-Driven Models
Statistical ModelingArtifical IntelligencePhysical ModelingBig Data Analytics tools
By using these methodologies, we can create value from the district heatingdata by formulating new grey-box models based on the data and physicalintelligence of the network. Thus, Data-Driven Models enables toincrease the savings and the efficiency of the DH network.
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 4 / 19
Load Forecasting in District HeatingPhysical Modeling Statistical Modeling
Weather Forecast† Heat Load Forecast†
†ENFORhgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 5 / 19
Temperature Optimization in District HeatingUncertainty Real data
Temperature/FlowControllers
Critical Point
Listen to the grid, by identify few of thelowest temperature (critical) in the grid andoptimize based on the grid feedbackUsing controllers to control the flow andsupply temperature from the productionIncrease the flow, [mt ] before increasing thetemperature to produce the requried heatload, Qt = mtcp∆Tt
Regulating the flow first will increase thesavings of the production
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 6 / 19
Savings by using Temperature Optimization
The objective of utilizing the data and create data-driven models is toincrease the savings for the district heating company.A new report shows there is a potential of saving about 240-790Million DKK yearly by reducing the temperature of about 3-10 degreesusing data-driven temperature optimization*.More savings using intelligent heat control as the supply temperature islowered
*Potentialet ved dynamisk datadrevet temperaturregulering i fjernvarmesektoren, DANVAD Analytics & Dansk Fjernvarme, 2019-02
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 7 / 19
Usage Case of creating value from end-user data
How to use the end-user data to identify and create an artificial criticaltemperature.
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 8 / 19
Data - Kraftringen
The data was supplied by Kraftringen in realtion to the Smart CitiesAccelerator project.
Around 60 houses located in Lund.Three different set of houses behind a physical measurement.The smart data set include: Supply Temperature, Return Temperature,Flow, Total Energy.Hourly resolution over the period 01/10/2017 until 31/03/2018.The data were anonymized and the location of the houses is unknown.
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 9 / 19
The Smart Cities Accelerator (SCA)
EU funded project through the regional programme Interreg-ÖKSA research and innovation partnership working with optimizingenergy-systems away from fossile fuels towards a more sustainablefutureSCA involves 12 partners from universities, cities and companies in theGreater Copenhagen area and Skåne region who have ambitiousagendas on becoming CO2 free in a close future
Copenhagen aims to be CO2 free by 2025Malmö aims to be CO2 free by 2030
Budget: 6.468.035 Euro / 50Period: 1. September 2016 – 28. February 2020Projectleader: DTU Compute
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 10 / 19
Artifical Critical Temperature
40
60
80
100
Dec 04 Dec 11 Dec 18 Dec 25Time
Temp
eratu
re [C
]
Artificial CP CP
Estimating Artificial Critical Temperature
By using the end-user smart meter measurements to estimate criticaltemperature or Artifical Critical Temperature.Utilizing the data from the end-users gives value to district heatingcompany
Making the physical critical points redundantMore flexibility as the critical points can be moved in the grid with noeffort
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 11 / 19
Clustering the supply temperature
−0.75
−0.50
−0.25
0.00
0.125 0.150 0.175 0.200
Loadings v1
Load
ings
v2
Centroids 1 2 3 Cluster 1 2 3
−0.25
0.00
0.25
0.50
0.125 0.150 0.175 0.200
Loadings v1
Load
ings
v3
Centroids 1 2 3 Cluster 1 2
House.1.1
House.1.2
House.1.3
House.1.4House.1.5
House.1.6
House.1.7
House.1.8House.1.9
House.1.10
House.1.11
House.1.12
House.1.13
House.1.14
House.1.15
House.1.16 House.1.17
House.1.18
House.1.19
House.1.20
House.1.21
House.1.22
House.1.23
House.1.24
−0.25
0.00
0.25
0.50
−0.75 −0.50 −0.25 0.00
Loadings v2
Load
ings
v3
Centroids 1 2 3 Cluster 1 2 3
Using Principal Component Analysis to cluster the houses together byusing the eigenvector values as inputs for the Kmeanshgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 12 / 19
Clustering the supply temperature
0
25
50
75
100
Hou
se.1
.1
Hou
se.1
.2
Hou
se.1
.3
Hou
se.1
.4
Hou
se.1
.5
Hou
se.1
.6
Hou
se.1
.7
Hou
se.1
.8
Hou
se.1
.9
Hou
se.1
.10
Hou
se.1
.11
Hou
se.1
.12
Hou
se.1
.13
Hou
se.1
.14
Hou
se.1
.15
Hou
se.1
.16
Hou
se.1
.17
Hou
se.1
.18
Hou
se.1
.19
Hou
se.1
.20
Hou
se.1
.21
Hou
se.1
.22
Hou
se.1
.23
Hou
se.1
.24
House
Tem
pera
ture
[C]
Cluster 1 2 3
Critical Point One
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 13 / 19
Time of the week distribution
Two different temperature zones behind one of the critical pointsFaulty pipe?, two different house types?, leakage?
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 14 / 19
Clustering
Possible to cluster houses together based on the time-of-the-weekdistribution of their supply temperature using Jensen-Shannonprobability distance metric. Then using Laplacian eigenmaps to clusterthe houses together.Identify houses with problems, different temperature zones in thenetwork, etc.hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 15 / 19
Artifical Critical Temperature using Clusters
60
70
80
90
100
Dec 04 Dec 11 Dec 18 Dec 25Time
Temp
eratu
re [C
]
Artificial CP Artificial CP Clustered CP
Estimating Artificial Critical Temperature
Cluster houses together then taking the cluster with the highestmedian and create an artificial critical temperature based on them.
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 16 / 19
Multiple Temperature Zones inside a City
Combining the current temperature optimization, HEATTO† andsmart-meter data to control several different temperature zones insidea cityRemoving pre-selected critical points from the network using the meterdataControlling areas with new sustainable buildings with lower supplytemperature and large old inefficient buildings with higher temperatureUsing temperature mixing and pressure pumps in the network
†ENFORhgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 17 / 19
Usage of digitalization in District HeatingDigitalization gives the opportunity of getting the end-users measurement inreal-time which can be used for information, monitoring and controlpurposes.
Smar
t Met
er D
ata
Temperature/FlowControllers
Critical Point
Example of Value Creation from Digitalizationin DH:
Identify critical points in the network andcreating artifical critical point, i.e. makingthe physical critical point redundantTemperature optimization usingcontrollers with the feedback from theend-users in the grid.Hierarchical Temporal/Spatial LoadDemand Forecast.Utilizing local weather stations to improveheat load demand forecast.Clustering to create multilevel temperaturezones.hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 18 / 19
Thanks
Thank you
Contact Information:
1 Hjorleifur, [email protected] Henrik Madsen, [email protected] Henrik Aa. Nielsen, [email protected]
hgbe/hm ( DTU ) Identifying Critical Temp. 11/10/2019 19 / 19