DISTRIBUTED LOAD BALANCING OF DISTRICT HEATING

SYSTEMS: A SMALL-SCALE EXPERIMENT

Fredrik Wernstedt and Paul Davidsson

Department of Systems and Software Engineering, Blekinge Institute of Technology, Ronneby, Sweden

Keywords: Distributed control, Softwar

e agents, District heating

Abstract: We present results from an experiment where the effects of automatic flow control at a single substation is

compared to automatic cooperative concurrent flow control at multiple substations. The latter approach is

made possible by equipping individual substations with some computing power and integrating them into a

communications network. Software agents, whose purpose is to cooperate with other software agents

(substations) and to invoke reductions, are connected to each substation. The experiment show that it is

possible to automatically load balance a small district heating network using agent technology, e.g., to

perform automatic peak clipping and load shifting.

1 INTRODUCTION

District heating systems are by nature distributed

both spatially and with respect to control. Each

consumer (substation) can be viewed as a "black-

box" making local decisions without taking into

account the global situation. Thus, today a district

heating network is basically a collection of

autonomous entities trying to optimize operations

locally, which typically results in behaviour that is

not globally optimal.

ABSINTHE (Agent-based monitoring and

cont

rol of district heating systems) (WWW-

ABSINTHE) is a collaboration project between

Blekinge Institute of Technology and Cetetherm AB.

The goal of the project is to improve the monitoring

and control of district heating, e.g., by increasing the

knowledge about the current and future state of the

network at the producer side and by performing

automatic load balancing at the customer side.

Individual substations are equipped with some

computing power and integrated into a cooperative

system via a communications network. Each

substation is equipped with a software agent (Weiss,

1999, Wooldridge, 2002) that will enable it to

cooperate with other substations to perform

automatic load control.

A small-scale experiment has been performed in

a laborat

ory accredited by the Swedish Board for

Accreditation and Conformity Assessment (WWW-

SWEDAC) at Cetetherm AB. The objective of the

experiment is to validate a decentralized approach to

automatically reduce the effect that domestic hot

water consumption causes on the primary flow by

automatically and concurrently reducing the flow

caused by household heating at multiple substations

without any central control.

In the next section we discuss the problem

dom

ain and motivate a decentralized approach. This

is followed by a description of the architecture of the

multi-agent system used in the experiment. We then

present the results from the experiment. Finally, we

conclude and provide pointers to future work.

2 DISTRICT HEATING SYSTEMS

Current approaches to operate district heating

systems are centralized. The load of a district

heating system is mainly a consequence of the

customers' demand for household heating. As a

result, the operation of most district heating systems

is based on a simple mapping between the ambient

temperature and the supply temperature.

Furthermore, in order to ensure sufficient heat

supply, the tendency has been to produce more heat

than necessary to satisfy the demands of the

consumers (Arvastsson, 2001, Canu et al., 1994,

Petersson and Werner, 2003).

310

Wernstedt F. and Davidsson P. (2004).

DISTRIBUTED LOAD BALANCING OF DISTRICT HEATING SYSTEMS: A SMALL-SCALE EXPERIMENT.

In Proceedings of the First International Conference on Informatics in Control, Automation and Robotics, pages 310-313

DOI: 10.5220/0001132603100313

 SciTePress

A more advanced approach to decide the supply

temperature would be to use an optimization model.

In a general optimization model the network appears

as a set of constraints (where consumers have fixed

and given demand), and the objective function is

composed of cost for production. However, an

optimization model of a large district heating system

with many loops and more than one heat production

plant is extremely computationally demanding

(Aringhieri and Malucelli, 2003, Bøhm et al., 2002,

Tamminen and Wistbacka, 2001). In fact, it is

argued that when the complexity of a district heating

system reaches approximately 100 components and

restrictions, the present computer and software

technology is insufficient to find an optimum

operational strategy (Bøhm et al., 2002). Moreover,

the actual demand of the consumers is not known.

The quality of an optimization model is heavily

dependent of the quality of the predictions of future

demand (since direct measurements of consumption

typically will not suffice as the distribution time is

relatively large). Thus, even if we could implement

an efficient optimization algorithm, some type of

real-time handling of the discrepancies between

predicted and actual consumption is needed. Due to

the long distribution times, this probably needs to be

done in a decentralized manner.

A general argument against centralized

approaches for problems as complex as the

management of district heating systems is that when

the problems are too extensive to be analyzed as a

whole, solutions based on local approaches often

allow them to be solved more quickly (Rinaldo and

Ungar, 1998).

3 DISTRIBUTED CONTROL OF

DISTRICT HEATING SYSTEMS

It has been suggested that agent technology is a

promising approach to manage distributed complex

systems (Parunak, 1996), such as district heating

systems.

In this work we evaluate a decentralized multi-

agent system, where individual consumer agents are

locally taking interaction decisions. In a

decentralized multi-agent system agents interact

laterally to achieve coherence. The consumer agents

are given two fundamental goals to fulfil: to operate

on the behalf of the customers and to coordinate

with other agents to perform automatic load

balancing.

We assume that the flow of domestic hot water

varies in a non-controllable way, i.e., a process

disturbance that should be reduced. However,

control of domestic hot water consumption would

immediately decrease the comfort for the end user,

i.e., we should not limit the primary flow caused

directly by domestic hot water consumption.

However, since the heating of buildings is a slow

process, we should be able to make time-limited

reductions to the flow caused by radiators without

significantly affecting the comfort (Drysdale and

Stang, 2002, Österlind, 1982, Gieseler et al 2003).

The consumer agents will impose restrictions to

household heating when the consumption levels are

raised above specific limits. Each building has its

own unique heating characteristic, i.e., the limit for

reductions and the persistence of reductions will

need to be configured for each individual agent.

When enforcing a local reduction the consumer

agent also requests reduction assistance from other

nearby cooperative substations.

4 EXPERIMENTAL DESIGN

4.1 Laboratory Set-up

In the experiment we used two heat exchange

systems (600 kW and 400 kW respectively) and one

heat producing unit. The substations used were

developed by Cetetherm AB during the ABSINTHE

project and contain a built-in extendable I/O

platform with an expansion slot for a communication

card (WWW-Siemens). Access to sensor data is

provided by a Rainbow communication card (see

Figure 2). The I/O card contains a database that

continuously is updated with sensor data from the

I/O channels by a small real-time operating system.

Figure 2: The Rainbow communication and computation

card is here shown on top of the Saphir hardware interface

card.

Communication e.g. Ethernet

Protocol stack e.g. TCP/IP

Application e.g. Agent

A laboratory control computer handled

production and consumption as well as making all

measurements during the experiments. In Figure 3

we show a conceptual view of the laboratory set-up.

API

Heat exchanger controller

Rainbow

com. card

Saphir

I/O card

DISTRIBUTED LOAD BALANCING OF DISTRICT HEATING SYSTEMS - A SMALL-SCALE EXPERIMENT

311

Figure 3: Conceptual view of the set-up with the

laboratory control computer.

The consumer agents have the capability to change

the temperature set value for the household heating.

However, they may maximally reduce it by 15%. A

reduction by 15% may seem small but will actually

often result in a temporary shut-down of the heating

to the radiator system (since the water returning

from the radiator is sufficiently warm). As a result,

the consumption will start to decrease immediately,

but only at a relatively slow rate.

Another factor that slows down the control is the

way that the consumption is measured. We are only

able to discover changes in consumption at the same

rate as the rate of incoming pulses from the flow

gauges. The agents are instead using estimated

consumption values over an interval of one minute.

4.2 Consumption Scenario

Two substations, A and B, are set to have a constant

radiator demand of approximately 25kW and 15kW

respectively. The system is first allowed to reach a

steady state during five minutes. After five minutes

Substation B initiates a domestic hot water tapping

of 0,2 kg/s for a duration of five minutes. The

system is then given ten minutes to stabilize. After

the stabilization period Substation A initiates a

tapping of 0,2 kg/s, also with a duration of five

minutes. After the second tapping the system is

given time to stabilize before shutdown.

5 EXPERIMENTAL RESULTS

In the approach evaluated consumer agents

individually enforce reductions at their substation

when the consumption reaches a specific limit. The

limit for substation A is 30 kW and the limit for

substation B is 20 kW. When a consumer agent

enforces a local reduction the agent also requests

reduction assistance from the other substation. This

approach is evaluated against the normal situation

where substations are free to consume the amount

requested.

Figure 3 shows the total energy consumption for

the case with free consumption and the case with the

distributed control strategy during the experiment.

Figure 3: Total energy consumption for the two cases. The

desired maximal global consumption is 50 kW.

We see that the strategy to use a decentralized

multi-agent based approach clearly reduces the

consumption peaks. However, the multi-agent based

approach requires some time to assume the stable

level after reductions.

Figure 4 shows the amount of time of the total

experiment that the consumption has reached and

been above effects from 45 kW to 80 kW, e.g., the

agent system had a consumption of 60 kW and

above during 15% of the duration of the experiment.

Figure 4: Amount of time of the experiment that the

consumption has reached and been above effects from 45

kW to 80 kW.

Rainbow

Saphir

0 500 1000 1500 2000

Time [s]

Effect [kW]

No restrictions Decentralized MAS

Rainbow

Saphir

Production

10%

15%

20%

25%

30%

35%

40%

45 50 55 60 65 70 75 80

Effect [kW]

Time [%]

No restrictions Decentralized MAS

ICINCO 2004 - INTELLIGENT CONTROL SYSTEMS AND OPTIMIZATION

312

The multi-agent based approach only consumes 63

kW or more during approximately 10% of the

duration of the experiment. This is contrasted to

30% for no load balancing at all.

6 CONCLUSIONS

We have performed a small-scale experiment in a

controlled environment to evaluate the possibility of

distributed load balancing in district heating

systems. The results show that it is possible to

automatically load balance district heating systems

without any central control. Other possibilities that

integration of substations into a communications

network may have, besides environmental and

economical are for example the possibility to

prioritize certain customers, e.g., hospitals. To our

knowledge, agent technology has never been used

for monitoring and control of district heating

systems. There have been experiments performed

with centralized control of substations (Österlind,

1982), however we show that we can achieve

distributed concurrent automatic load balancing by

the use of agent technology. The experiments

described are only initial tests and there is much

room for improvements. For instance, because of the

flow gauges used, the agents had a limited and

delayed view of the environment, resulting in long

reaction times. By having continuous readings of

consumption we believe that it is possible to better

decide the persistence of household heating

reductions, which makes it possible to limit

unnecessary reductions to household heating. In

general, it is also possible to make more informed

decisions regarding reductions, e.g., if reduction

assistance should be requested from several

substations or just a few. Furthermore, it should also

be possible to develop strategies to even out the

negative effects of reductions over larger areas by

manipulating the willingness for agents to cooperate

and accept reductions. Future work includes:

– Investigating the scaling effects of the different

strategies using a simulation tool (Wernstedt et

al., 2003), as well as comparing this and other

strategies with centralized control strategies.

– Performing experiments in full-scale district

heating systems.

REFERENCES

ABSINTHE project web site. Retrieved June 21,2004,

from http://www.ipd.bth.se/absinthe.

Aringhieri, R., Malucelli, F., 2003. Optimal operations

management and network planning of a district

heating system with a combined heat and power plant.

Annals of Operations Research, 120, 173-199.

Arvastsson, L., 2001. Stochastic modeling and operational

optimization in district heating systems, Lund Institute

of Technology, Sweden, ISBN 91-628-4855-0.

Bøhm, B., Lucht, M., Park, Y., Sipilä, K., Ha, S., Won-tae,

K., Bong-kyun, K., Koljonen, T., Larsen, H., Wigbels,

M., Wistbacka, M., 2002. Simple models for

operational optimisation, Report 2002:S1, NOVEM.

ISBN 90-57-48021-2.

Canu, S., Duran, M., and Ding, X., 1994. District heating

forecast using artificial neural networks, International

Journal of Engineering,2(4).

Drysdale, A., Stang, J., 2002. Optimised DH systems

using remote heat meter communication and control.

In 8th International Symposium on District Heating

and Cooling. ISBN 82-594-2341-3.

Gieseler, U., Heidt, F., Bier, W., 2003. Combined thermal

measurement and simulation for the detailed analysis

of four occupied low-energy buildings. In 8th

International IBPSA Conference.

Larsson, G., Rossing, O., 2003. Flödesutjämnande

körstrategi (In Swedish). Forskning och Utveckling,

FOU 2003:86. ISSN 1402-5191.

Parunak, V., 1996. Applications of Distributed Artificial

Intelligence in Industry, in eds. O'Hare , G.M.P.,

Jennings, N.R., Foundations of distributed Artificial

Intelligence. Wiley Inter-Science.

Petersson, S., Werner, S., 2003. Långtidsegenskaper hos

lågflödesinjusterade radiatorsystem (In Swedish).

Forskning och Utveckling, FOU 2003:88. ISSN 1402-

5191.

Rinaldo, J., Ungar, L., 1998. Auction-driven coordination

for plantwide optimization. In FOCAPO’98,

Foundations of Computer-Aided Process Operation.

Siemens web site. Retrieved June 21, 2004, from

http://www.sbt.siemens.com/press/hvac_pr13.asp.

Swedish Board for Accreditation and Conformity

Assessment web site. Retrieved June 21, 2004, from

http://www.swedac.se.

Tamminen, E., Wistbacka, M., 2001. Capacity and cost

models for thermal power systems with random

outages of plants. VTT Energy, Espoo, Research

Report ENE6/44/01.

Weiss, G., 1999. Multi-Agent Systems, MIT Press,

Cambridge, MA. ISBN 0-262-23203-0.

Wernstedt, F., Davidsson, P., Johansson, C., 2003.

Simulation of district heating systems for evaluation of

real-time control strategies. In 1st European

Simulation and Modelling Conference.

Wooldridge, M., 2002. An introduction to multi-agent

systems, John Wiley & Sons. ISBN 0-471-49691.

Österlind, B., 1982. Effektbegränsning av fjärrvärme (In

Swedish). Rapport 63:1982, Byggforskningsrådet.

ISBN 91-540-3714.

DISTRIBUTED LOAD BALANCING OF DISTRICT HEATING SYSTEMS - A SMALL-SCALE EXPERIMENT

313