ONTOLOGY BASED DESCRIPTION OF DER’S LEARNED

ENVIRONMENTAL PERFORMANCE INDICATORS

Managing the Environmental Performance of Distributed Energy Resources

org Bremer

Department of Computing Science, University of Oldenburg, Uhlhornsweg, Oldenburg, Germany

Keywords:

Distributed Energy Resources, Environmental Performance Indicator, Optimization, Constraint Modelling.

Abstract:

The current technology’s upheaval in the electricity sector is leading to a need for new distributed control

schemes for distributed electricity generation. In order to enable optimization of electricity generation and

consumption for global objectives in distributed search spaces, meta-models of constrained spaces of operable

schedules are indispensable for efﬁcient communication. In order to qualify for being a green technology,

indicators for individual environmental performance have to be incorporated into these meta-models. We here

present insight into ongoing work concerning the integration of environmental performance indicators into the

distributed control of energy resources. In order to ensure interoperability on the indicators, we will discuss

the need for an ontology deﬁning the description of such indicators.

1 INTRODUCTION

In order to allow for a transition of the current cen-

tral market and network structure of today’s electric-

ity grid to a decentralized smart grid, an efﬁcient man-

agement of numerous distributed energy resources

(DER) will become more and more indispensable.

We here consider rather small, distributed elec-

tricity producers that are supposed to pool together

with likewise distributed electricity consumers and

prosumers (like batteries) in order to jointly gain more

degrees of freedom in choosing load proﬁles. In this

way, they become a controllable entity with sufﬁcient

market power. In order to manage such a pool of

DER, the following distributed optimization problem

has to be frequently solved: A partition of a demanded

aggregate schedule has to be determined in order to

fairly distribute the load among all participating DER.

Optimality usually refers to local (individual cost) as

well as to global (e.g. environmental impact) ob-

jectives in addition to the main goal: Resemble the

wanted overall load schedule as close as possible.

When determining an optimal partition of the

schedule for load distribution, exactly one alterna-

tive schedule is taken from each DER’s search space

of individual operable schedules in order to assem-

ble the desired aggregate schedule. For optimization,

a scheduling algorithm (whether centralized or not)

must know for each DER which schedules are opera-

ble and which are not. Therefore, the set of alterna-

tive, operable schedules (obeying multiple constraints

like allowed power or voltage bands or buffer charg-

ing levels) has to be encoded by an appropriate, stan-

dardizable model for inclusion into optimization. An

example for such a model has been presented by (Bre-

mer et al., 2010). If alternative solutions are to be

chosen environmentally conscious, appropriate indi-

cators must be included into the model of the search

space and therefore into the description of each alter-

native schedule. Such an indicator allows a conclu-

sion about a single type of environmental impact (e.g.

emissions) of a schedule.

Each DER has to serve the purpose it has been

built for. But, usually this purpose may be achieved

in different alternative ways. For example, it is the

(intended) purpose of a CHP (a small in-house plant

for combined heat and power production) to deliver

enough heat for the varying heat demand in a house-

hold at every moment in time. Nevertheless, if heat

usage can be decoupled from heat production by us-

ing a thermal buffer store, different production pro-

ﬁles may be used for generating the heat. This leads,

in turn, to different respective electric load proﬁles

that may be offered as alternatives to a scheduling al-

gorithm.

With each of these alternatives, different environ-

mental impacts are associated. If it becomes pos-

sible, to make information about different environ-

107

Bremer J..

ONTOLOGY BASED DESCRIPTION OF DER’S LEARNED ENVIRONMENTAL PERFORMANCE INDICATORS - Managing the Environmental

Performance of Distributed Energy Resources.

DOI: 10.5220/0003978001070112

In Proceedings of the 1st International Conference on Smart Grids and Green IT Systems (SMARTGREENS-2012), pages 107-112

ISBN: 978-989-8565-09-9

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

mental impacts (globally) available, then optimization

for environmental conscious operation of all DER be-

came true even in distributed optimization environ-

ments with a control scheme that comprises (large)

groups of DER.

We here present insights into an ongoing research

that aims at the integration of ontology based descrip-

tions of environmental performance indicators (EPIs)

and support vector based meta-models of distributed,

constrained search spaces for DER as have for ex-

ample been developed by (Bremer et al., 2011a) or

(Blank et al., 2011).

Developing a set of standardized EPI descrip-

tions and a software solution for collecting and man-

aging environmental performance information from

distributed sources and reporting based on EPIs is

the main goal of the OEPI-project (http://www.oepi-

project.eu). We will extend these ideas to the ﬁeld of

optimizing DER in a smart grid environment.

In this position paper, we present our road map

and preliminary concepts that aim at an adaptation

of ontology based services for monitoring and cal-

culating EPIs within enterprises production planning

schemes and supply chains to the new use case of

DER optimization. We discuss our approach for in-

tegrating EPIs with search space meta-models and

present a preliminary formal representation. We con-

clude with a discussion of the ontologies role for in-

teroperability.

2 DESCRIBING

ENVIRONMENTAL

PERFORMANCE

Environmental performance indicators like carbon,

water or energy footprint usually measure the envi-

ronmental impact of the activities of organizations. In

this way, they are a measure that reﬂects the perfor-

mance in achieving the actual objectives with respect

to environmental issues (cf. (H

reb

cek et al., 2007)).

Up to now, no general standardized description

model for exchanging information on environmen-

tal performance in general exists. It is later argued

that also for the smart grid domain there is a need

to communicate environmental issues that have to be

commonly understood among communicating and in-

teracting distributed devices. As a starting point for

further development, we chose our ontology from

the OEPI project. The OEPI project aims at devel-

oping a set of standardized EPI descriptions and a

software solution for collecting and managing envi-

ronmental performance information from distributed

3,5 4 4,5 5 5,5 6

electric power / kW

94,5

95,5

96,5

overall efficiency/ %

Figure 1: Relationship between generation rate and over-

all efﬁciency for a SOLO-Stirling-CHP. Modiﬁed from

(Thomas, 2007).

sources and reporting based on EPIs (Meyerholt et al.,

2010). The vision of the OEPI project aims at busi-

ness users – across industries and supply chains – and

at a continuous reduction of environmental impacts

of daily operations. To achieve this, the visibility and

comparability of EPIs of alternative decisions in cor-

porate and supply chain operations is enlarged. In the

smart grid domain, individual EPIs must be made vis-

ible for interacting devices what entails a much higher

degree of automation and way shorter time periods

than in the business sector.

Some examples for EPIs concerning DER in the

context of a virtual power plant (VPP) are:

• Static losses, i.e. the loss of energy due to imper-

fect insulation e.g. in a thermal buffer store. The

higher the temperature is in the store, the higher

are the losses.

• Too many cold starts entail increased fuel con-

sumption and abrasion.

• Low generation rates often lead to a declined efﬁ-

ciency as Figure 1 shows using the example of the

SOLO-Stirling-CHP (Thomas, 2007).

If EPIs from different DER are to be compared

during a design or optimization process for environ-

mentally conscious decision making, a uniﬁed de-

scription language is required that unambiguously de-

ﬁnes how an EPI was computed, what data was con-

sidered and – most important – how it may be fur-

ther processed and related to other EPIs. OEPI de-

velops such a language using a reference-ontology

to describe EPI semantics and to achieve interoper-

ability among different EPI related services and pro-

cesses along supply chains in the business domain.

Currently, such measures are - if at all - applied on

a long term (mostly annual) or on a one-time basis,

what limits the scope of application for such EPIs.

Annual sustainability reporting is one example for to-

day’s use of EPIs - with the aim of merely gaining le-

gal compliance. At present, the need for a shift from

such a strong operational focus to a more strategically

oriented realignment of e.g. corporate environmen-

tal management information systems (CEMIS) can

be observed (Teuteberg and G

omez, 2010). In the

smart grid domain, much shorter terms are required

SMARTGREENS2012-1stInternationalConferenceonSmartGridsandGreenITSystems

108

Figure 2: Excerpt from OEPI’s preliminary EPI ontology with respect to DER impacts.

for spreading environmental information. One goal of

OEPI is the development of a standardized exchange

format for EPIs. Figure 2 shows an excerpt from an

early version of the reference ontology emphasizing

already inherent aspects that are important to DER:

• Energy input, i.e. primary energy usage and

avoidable losses.

• Direct green house gas emissions (scope 1); these

might be counted as scope 2 emissions for pro-

cesses using the generated energy.

• Waste refers to needs due to possibly increased

maintenance requirements.

3 DESIGN FOR ENVIRONMENT

Before discussing sustainable procurement for dis-

tributed energy management, we will have a short

look at the related processes in the business domain.

In order to get environmentally sound products out

of the design phase, it is important to provide prod-

uct engineers with concepts and tools that make them

aware of the environmental impacts of their design

decisions on the whole product life cycle. This will

be achieved by integrating EPIs into the design opti-

mization process. As a result, evaluation of design al-

ternatives with lower overall environmental footprint

will become much faster.

The term IT-for-Green has recently been estab-

lished in order to distinguish clearly between sustain-

ability issues of hardware/software and sustainability

achieved by means of IT (Rapp et al., 2011). The

realm is an increased environmental friendliness of

companies and their processes. However, conven-

tional CEMIS are not sufﬁcient to achieve this ob-

jective, as they serve mainly for ensuring legal com-

pliance with relevant environmental regulations in or-

der to avoid ﬁnancial sanctions from state authori-

ties. With such a strong operational focus, the require-

ments entailed by the concept of sustainable develop-

ment can only be fulﬁlled to a very limited degree.

But, companies may achieve proﬁts by applying

sustainable development measures and by implement-

ing new CEMIS: they reduce costs through mate-

rial and energy efﬁciency and increase their turnovers

through sustainable products and services, corporate

image improvement and advantages in competition.

Similar incentives should be considered for the smart

grid domain to enforce an environmental conscious

electricity grid.

Sustainable procurement is another example for a

use case calling for EPIs being available virtually on

demand. Such procurement integrates a concept for

supplier-dependent EPIs into business processes for

procurement. Here, one goal is to enable a business

user to take action (and responsibility) according to

different environmental impacts of different alterna-

tive sources for raw materials for production. This

task is currently hardly achievable because individual

variations are not captured in the procurement infor-

mation systems (Dada et al., 2010).

4 THE CASE OF VIRTUAL

POWER PLANTS

These business domain use cases are now to be trans-

ferred to the smart grid domain using the example of

virtual power plants.

We consider a VPP as an orchestrated group of

DER that communicate with and are controlled by

a central controlling unit that is responsible for see-

ing that the group as a whole fulﬁls certain objectives

like bundling for greater market power or grid stabi-

lizing by reduced stochastic feed-in. A VPP helps in-

tegrating DER into the (smart) energy grid of the fu-

ture (Lukovic et al., 2010). We do not deal with any

concrete objective here, but we assume that a central

scheduling unit has to search the space of alternative

load schedules for each DER in order to ﬁnd an ap-

propriate one that ﬁts bests for the problem at hand,

although we do not exclude further extensions to fu-

ture distributed control schemes.

If the central scheduler is supposed to select al-

ternative schedules according to their individual envi-

ronmental performance as well as according to their

appropriateness for the main objective, then such in-

formation must be incorporated as additional fea-

ONTOLOGYBASEDDESCRIPTIONOFDER'SLEARNEDENVIRONMENTALPERFORMANCEINDICATORS-

ManagingtheEnvironmentalPerformanceofDistributedEnergyResources

109

Figure 3: Integration scheme for scheduling algorithms.

ture sub-space into the schedule search space and its

model. In this paper, we follow the support vector ap-

proach in (Bremer et al., 2010) for encoding the load

schedules of a DER. This approach already has been

proven as ﬂexible enough to incorporate additional

information about the individual environmental per-

formance of each alternative into the learned model

(Bremer et al., 2011b). While learning the geomet-

rical structure of the space of feasible schedules, the

approach is able to concurrently learn the functional

relation of assigned performance indicators. In this

way, each individual alternative load schedule can at

least be annotated with numerical values of respective

environmental performances.

4.1 Adapting the Use Cases

Both use cases from the business domain can be

adapted for the load planning process within a VPP.

In this way, we consider the procurement of a differ-

ent product. In a VPP, each DER may offer several al-

ternative load schedules. These are the products that

are sought after by the controller. As every material

(physical) product, it is also possible in such scenario,

to assign measures of individual environmental per-

formance to each alternative load schedule.

• Different DER and the controller make up a sup-

ply chain.

• Collaborating DER jointly produce a product (an

aggregated schedule).

• Environmental impacts from one DER has an im-

pact on the performance of all others and therefore

on the whole VPP.

• Reducing the planning horizon or precautious in-

tegration of appropriate information may lead to

strategic avoidance of negative impacts.

• In both cases, environmental performance has to

be communicated.

On an abstract level, a VPP is a business like any

other, but with an highly automated and highly fre-

quent product design and procurement process.

4.2 Encoding the Indicators

The format of the information (incl. EPIs) for a sched-

ule that might be offered by a DER is currently dis-

cussed to be of the following format:

x = (x

, . . . , x

, e

, . . . , e

)

= (x

, . . . , x

, f

, . . . , x

), . . . , f

, . . . , x

)).

Herein x denotes a schedule (with x

denoting load at

time interval i) with associated ancillary information

about the individual environmental performance rep-

resented by one or more EPI e

, . . . , e

. The EPI in

turn, usually is a function of the schedule itself, e.g.

the heat losses resulting from that speciﬁc operation:

, . . . , e

) = ( f

(x), . . . , f

(x)).

In this form, the schedules (together with the at-

tached indicator values) may be taken as input for the

mentioned support vector models for learning and en-

coding the set of operable schedules for communi-

cation to the scheduler without a need for adaption.

After decoding on scheduler side, the schedules will

still have the same information on load per time in-

terval and the values of the indicators (Bremer et al.,

2011b). Now the scheduling unit needs to know the

individual meanings of the indicator values.

Figure 3 shows the envisaged integration scheme

for scheduling algorithms. When the model of the

search space is built-up from a DER upon request

from the scheduler, the model will be accompanied

by a set of meta-information objects – one for each

requested EPI. Each meta-object contains information

structured as shown in Figure 2. Associated to each

EPI-deﬁnition, is a numeric data item that, in the case

of VPP, will be an instance of a referenced data item.

This means, that the actual value of the EPI is not

directly included (like in the other cases), but is a ref-

erence to some external value. We here consider a

relative addressing of elements of a schedule (starting

from the ﬁrst element referring to an EPI).

Thus, only one meta-information object for each

EPI is needed for all schedules in the model, whereas

the actual EPI values are reconstructed by the en-

coded functional relationship from the search space

model. After reconstruction, schedules have the for-

mat shown in Figure 3. For each EPI value in the

schedule, a meta-information object exists (commu-

nicated from the DER together with the model), that

serves for interpreting the values.

If comparability is ensured for all participating

DER, respective referenced EPI values may then be

integrated into the objective function of the scheduler

and are therefore used for environmentally conscious

judging of different solutions during the scheduling

process.

SMARTGREENS2012-1stInternationalConferenceonSmartGridsandGreenITSystems

110

Figure 4: Referencing external data values with the EPI on-

tology.

4.3 The Role of the EPI Ontology

In engineering, ontologies always comprise a special-

ized vocabulary as well as an explicit set of assump-

tions regarding intended meanings of data. Entities

within an ontology either describe concepts (e.g en-

ergy losses) or relations between them (e.g. CO

is a

greenhouse gas). Such shared agreement within a sys-

tem environment with distributed entities allows for

a common understanding of exchanged information.

In the case of a virtual power plant, or (more gen-

eral) in the case of interacting distributed devices and

appliances in a smart grid environment, the need for

several standards for ensuring interoperability arises.

On a technical integration level, standards for integra-

tion – e.g. (IEC 61850-7-420:2009, 2009) – exist (al-

though rarely implemented in practice) and the com-

mon information model (CIM) is discussed as an on-

tology based model for operating data exchange (Us-

lar, 2005).

For describing the individual environmental per-

formance of alternative schedules and their produc-

tion schemes that might be considered in a VPP, cur-

rently no appropriate description standard exists. For

this reason, we will take the OEPI ontology as a start-

ing point.

A cost effective and high-quality ontology heav-

ily relies on the reuse of existing ones. The reuse of

already existing ontologies (or at least parts of them)

not only prevents reinventing the wheel in some cases

but also reduces the effort required for ontology build-

ing.

Ontology engineering has meanwhile grown to a

mature discipline. Different methodologies and tools

for support are readily available. Nevertheless, ac-

cording to (Bontas, 2005) most ontologies are still

a result of some ad-hoc application- and domain-

dependent engineering process. Building ontologies

from scratch is time-consuming and error-prone. In

order to weaken this challenge for the development

of new ontologies, knowledge sources available on

the web should be harnessed. If such existing onto-

logical knowledge is used as an input source for the

creation of new ontologies; this process is called on-

tology reuse.

Reusing the OEPI ontology will mean using it as

Figure 5: Integration of EPI-ontology into distributed en-

ergy management schemes.

a base ontology and derive a DER environmental per-

formance related one from it by extending the con-

cepts for communicating environmental performance

of energy schedules as discussed. The most impor-

tant extension (as a prerequisite to our approach) will

be the deﬁnition of EPI classes whose actual value is

just a pointer to some externally stored data (Figure

4) rather than included data values of any sort. As

we have seen in the previous section, individual en-

vironmental performance can – even in more than on

dimension – easily be incorporated in a schedule de-

scription as mere number values. But, what are the

meanings of these values? Different appliances may

consider different performance indicators and there-

fore communicate different sets of indicators. Some

indicators might be related to money instead of en-

vironmental cost. As an example, static losses do not

make sense for systems without a thermal buffer store.

On the other hand, these are energy losses and might

be mapped (or at least compared) to other energy loss

related indicators like e.g. losses through conversion.

This gives rise to some questions:

• What is the impact of different indicators?

• Which indicators can be mapped for comparison

and – more important – how can this mapping be

done?

• What does optimization mean in the light of a

given indicator?

Clearly, an optimization algorithm that wants to pro-

cess EPIs attached to individual schedules in order

to choose and evaluate schedules in a many objective

scenario, needs a description of meaning and relation-

ships of the EPIs.

The role of our ontology as interoperability ensur-

ing entity within a VPP use case is depicted in Figure

5. Each individual DER has access to the ontology as

common knowledge and therefore can create a proper

meta-description of each indicator that is encoded to-

gether with the schedules. One description is created

for each indicator position in the schedule (Figure 3).

A central scheduler in charge of ﬁnding optimal

schedules for each DER may then harness the indica-

ONTOLOGYBASEDDESCRIPTIONOFDER'SLEARNEDENVIRONMENTALPERFORMANCEINDICATORS-

ManagingtheEnvironmentalPerformanceofDistributedEnergyResources

111

tor descriptions and use the ontology to ﬁnd out, how

to treat them during evaluating a schedule, how to

convert indicators or how to map them to each other.

5 CONCLUSIONS AND FURTHER

WORK

It is to be expected that future architectures for smart

grids will call for the ability of DER to frequently at-

tach themselves to different groups depending on the

situation at hand. Unlike VPPs, such groups will be

drawn together rather by market forces. This transi-

tion to volatile groups of independent DER will be

gradual. Due to universal applicability to central,

decentral and distributed scheduling approaches, our

method should be able to serve the needs during the

whole transition process as the ontology approach is

independent of speciﬁc smart grid architectures.

Research has just started out. Up to now we

have gained a clear understanding of how the inte-

gration of distributed knowledge about individual en-

vironmental performance into a centralized (and in

the long run decentralized) energy management con-

trol scheme might be done. A meta-model of search

spaces that is based on geometric subspace descrip-

tions, provides an ideal connecting point for the in-

tegration of environmental information about alterna-

tives by simply extending the mathematical model of

a schedule to additional dimensions for EPIs. In this

way, necessary information is directly incorporated

into the meta-model of the load schedules and there-

with into the schedules themselves.

At the same time, correct interpretation of the EPI

values may be ensured with the help of a extended

EPI ontology. Our next steps will be the extension of

the OEPI ontology as discussed and the deﬁnition of

a standard set of EPIs for the sketched scenarios be-

fore we will start implementing a planned simulation

environment to test our approach.

ACKNOWLEDGEMENTS

This work was funded by the European research

project OEPI (Solutions for Managing Organizations

Environmental Performance Indicators, 748735).

REFERENCES

Blank, M., Gerwinn, S., Krause, O., and Lehnhoff, S.

(2011). Support vector machines for an efﬁcient rep-

resentation of voltage band constraints. In Innovative

Smart Grid Technologies. IEEE PES.

Bontas, E. P.; Mochol, M. (2005). Towards a reuse-oriented

methodology for ontology engineering. In 7th Inter-

national Conference on Terminology and Knowledge

Engineering, Copenhagen, Denmark. Shaker Verlag,

Aachen.

Bremer, J., Rapp, B., and Sonnenschein, M. (2010). Sup-

port vector based encoding of distributed energy re-

sources feasible load spaces. In IEEE PES Confer-

ence on Innovative Smart Grid Technologies Europe,

Chalmers Lindholmen, Gothenburg, Sweden.

Bremer, J., Rapp, B., and Sonnenschein, M. (2011a). En-

coding distributed search spaces for virtual power

plants. In IEEE Symposium Series in Computational

Intelligence 2011 (SSCI 2011), Paris, France.

Bremer, J., Rapp, B., and Sonnenschein, M. (2011b). In-

cluding Environmental Performance Indicators into

Kernel based Search Space Representations. Infor-

mation Technologies in Environmental Engineering

(ITEE 2011).

Dada, A., Staake, T., and Fleisch, E. (2010). Reducing envi-

ronmental impact in procurement by integrating mate-

rial parameters in information systems: The example

of apple sourcing. AMCIS 2010 Proceedings.

reb

cek, J., Misa

rov

a, P., and Hyr

slov

a, J. (2007). Envi-

ronmental Key Performance Indicators and Corporate

Reporting. In Environmental Accounting and Sustain-

able Development Indicators (EA-SDI 2007), pages

978–980. Universita Jana Evangelisty Purkyn

IEC 61850-7-420:2009 (2009). Communication networks

and systems for power utility automation - Part 7-420:

Basic communication structure - Distributed energy

resources logical nodes. IEC, Geneva, Switzerland.

Lukovic, S., Kaitovic, I., Mura, M., and Bondi, U. (2010).

Virtual power plant as a bridge between distributed en-

ergy resources and smart grid. Hawaii International

Conference on System Sciences, 0:1–8.

Meyerholt, D., Marx G

omez, J., Dada, A., Bremer, J., and

Rapp, B. (2010). Bringing sustainability to the daily

business: The oepi project. Proceedings of the Work-

shop ”Environmental Information Systems and Ser-

vices - Infrastructures and Platforms.

Rapp, B., Solsbach, A., Mahmoud, T., Memari, A., and Bre-

mer, J. (2011). It-for-green: Next generation cemis for

environmental, energy and resource management. In

Pillmann, W., Schade, S., and Smits, P., editors, En-

viroInfo 2011 - Innovations in Sharing Environmental

Observation and Information, Proceedings of the 25th

EnviroInfo Conference ’Environmental Informatics’,

pages 573 – 581. Shaker Verlag. ISBN 978-3-8440-

0451-9.

Teuteberg, F. and G

omez, J. M. (2010). Green comput-

ing & sustainability - status quo und herausforderun-

gen f

ur buis der n

achsten generation. Praxis der

Wirtschaftsinformatik, 47(274):6–17.

Thomas, B. (2007). Mini-Blockheizkraftwerke: Grundla-

gen, Ger

atetechnik, Betriebsdaten. Vogel Buchverlag.

Uslar, Mathias; Schmedes, T. L. A. L. T. W. L. A. H.-J.

(2005). Interaction of ems related systems by using

the cim standard. In Proceedings of the ITEE 2005

Magdeburg, pages 596–610. Shaker Verlag, Aachen.

SMARTGREENS2012-1stInternationalConferenceonSmartGridsandGreenITSystems

112