You are on page 1of 14

Unified Energy Agents as a Base for the Systematic Development of Future Energy Grids

Christian Derksen1, Tobias Linnenberg2, Rainer Unland1, and Alexander Fay2


DAWIS, Universitt Duisburg-Essen, Schtzenbahn 70, 45127 Essen {Christian.Derksen,Rainer.Unland}@icb.uni-due.de 2 IfA, Helmut-Schmidt-Universitt Hamburg, Holstenhofweg 85, 22043 Hamburg {Tobias.Linnenberg,Alexander.Fay}@hsu-hh.de
1

Abstract. The need for the application of software agents and agenttechnologies in highly diversified future energy grids is widely accepted today. Nevertheless, the very general concept of the agent paradigm still leads to misunderstandings and to the fact that agents are meant and utilized for very different tasks. Accordingly, the approaches that were presented in the Smart Gird area have major weaknesses in terms of comparability and a subsequently largescale use. We claim that the introduction of a unified definition of an Energy Agent will help to create a coherent picture that can accelerate further discussions and the conversion of the energy supply. Considering a development cycle that consists of modeling and implementation, simulation, test-bed application and the deployment to real systems, we present here our definition of an Energy Agent that takes into account the law of conservation of energy. Further, we present a classification of Energy Agents according to their sophistication and integration level and outline the need for individual but standardized energetic option models.

Introduction

In the course of the European and global energy-revolution some problems weaved aside as side issues at the beginning now turn out to be the major challenges. One of the most imminent threats is the introduction of a vast amount of isolated and incompatible control applications obstructing a sustainable integration of energy systems into a comprehensive energy control solution. Due to the highly diversified nature of tomorrows energy grid, new control solutions have to be found that tackle the rising demand for coordination. Coping with the challenge of fluctuating electricity production by a growing number of ever shrinking energy converters, dependent on meteorological factors as wind or sun, or heatoperated combined heat and power plants, already strains conservative control solutions. Meeting the high requirements with regard to availability of electricity supply, minimization of frequency- and voltage-deviations and other regulatory and socioeconomic factors does finally overstrain these control concepts as we may find them in todays electricity grids. While the physical grid infrastructures and the processes therein become more and more complex, the control solutions are lagging far behind.
M. Klusch, M. Thimm, and M. Paprzycki (Eds.): MATES 2013, LNAI 8076, pp. 236249, 2013. Springer-Verlag Berlin Heidelberg 2013

Unified Energy Agents as a Base for the Systematic Development

237

The needed change in paradigms has not yet left well controlled laboratory-like setups, where visionary thinking is allowed and contained in restrictive boundaries. In these setups different approaches have been implemented to approach the general problem of an ever growing complexity in number and solutions. Recent literature presents several Smart Grid concepts. Most of them focus either on stabilizing measures as needed for grid recovery during abnormal load conditions [1], the pooling of different types of energy converters to virtual power plants [2], or the control of connected energy consumers by means of demand side management [3]. All those efforts are united by one drawback - the central control. Todays central control solutions are generally limited in their number of nodes [4]. Due to the recent shift towards an increasing amount of small power plants it seems to be more promising to examine grid control concepts, featuring decentralized decision [5], [6]. One of the key features of Multi Agent Systems (MAS) is their proven reliability in the field of distributed problems. They are typically made up of a high number of basic components which are easily replaceable as each one is a separate entity. This high number of small actors gives a better tolerance towards failures in single modules and will therefore facilitate the compensation of grid failures or undesired grid conditions. In case of abnormal grid conditions it is possible to separate the affected regions from the grid and put them into a so-called islanding mode, while stabilizing the detached area allows reconnecting. This implies that the area in question is selfsustainable in regard to its energy needs [7], [8]. Besides all these benefits, there are some downsides, connected with the use of decentralized decision making in the context of energy management: All multi-agent based power control systems in literature known to the authors (e.g. [9], [10], [11]) do apply proprietary data exchange formats and architectures. This leads to compatibility and comparability issues. Even though a multitude of available standards like IEC61850 or the Common Interface Modell (CIM) and proprietary solutions for the different aspects of data exchange and control and command needs in the energy domain are available, none of them is totally suited for the needs of distributed processing and control. Furthermore do most of these approaches only focus on one form of energy, omitting the potential of solving problems common to other energy domains. As we can find a volatile energy demand and the tracking and quantification of energy flows not only in the domain of electrical energy, it would be desirable to implement one heuristic capable of solving problems common to the management processes of several forms of energy. Even though the agent based DEcentralized MArket based POwer control System (DEMAPOS) [12], developed at the Helmut Schmidt University in Hamburg, is capable of tackling some of the above mentioned challenges in regards to electricity networks, several pitfalls and difficulties, common to most alike systems, were discovered in the process of merging it with University Duisburg-Essens Agent.GUI [13] multi-domain simulation environment. The major challenge was found when designing the interfaces between the control and the simulation. To overcome these problems in the design process and the final implementation we suggest in the following the design and utilization of a generalized Energy Agent based on a unified agent model throughout all phases of the development cycle. This agent features standardized interfaces and inner workings, which will be able solve

238

C. Derksen et al.

the compatibility- and comparability issues not only for the domain of electrical energy, but also for the associated fields of gas-, water- and heat distribution. A differentiation between the agent itself and the surrounding system in the context of a desirable development process will be pointed out. It will be indicated on how to use the same piece of software throughout the entire development cycle - from concept to realization. Further individual problem areas connected with energy distribution and management will be considered as well. The structure of the paper is as follows: in section 3 a short introduction to the theoretical background is given, explaining the agent concept, its communication needs and the thereof derived requirements towards a testing environment. Afterwards the problem statement is specified on the basis of some examples. In section 4 a prototypic design process is introduced and in section 4 the thereof resulting Energy Agent is discussed. Section 5 presents an option model to be used in combination with the Energy Agent. A short synopsis concludes the paper.

Theoretical Backgrounds and Problem Statements

For the purpose of this paper an agent can be seen as an autonomous computational entity that is situated in an arbitrary environment and that is capable of autonomous interactions with this environment. Other optional properties and abilities that are also often associated with intelligent agents are their social ability, a reactive or proactive behavior and the capability to learn (e.g. [14]). If the overall system consists of a set of loosely coupled agents that act and cooperate with each other in the same environment, a multi-agent system (MAS) is formed. MAS are usually implemented to deal with problems whose complexity overstrains the capabilities of a single agent. For the practical problem solution the agents have to rely on communication capabilities and possibly also on the ability to collaborate, negotiate, delegate responsibility, and trust. These subjects are discussed in detail in literature [cf. e.g. 15]. Multi-agent based simulations (MABS) as used in this paper represent a specific scenario that is modeled and conceptualized as MAS. The involved active and (more or less) independent real world entities are mapped as agents that are embedded in an abstracted, shared and virtual environment. Thus, the foundation of MABS is the modeling and abstraction of the real world scenario, which is basically the definition of a comprehensive environment model and the agents that are situated in this environment. The advantages of MABS are often described by the benefits that result from the conceptualization of a system as an agency. These are: 1. Provision of natural description of a real system, 2. Capture of emergent phenomena in a system and 3. Flexibility [16]. The mapping of individuals from the real-world to an agent based model for simulations can naturally be done by mapping actively acting individuals to agents. Emergent behavior results from the interaction of these individuals. Thus, examples can be found in real social groups as well as in social groups that are simulated by a MABS. Unlike a system that results from a strict mathematical modeling (e. g. with differential equations) with clear overall rules, a MABS may be able to capture group behaviors that are enriched with individual aspects for considerations and reasoning. Those can be differently motivated and reach different decisions for a single situation.

Unified Energy Agents as a Base for the Systematic Development

239

MABS can be considered as a tool for simulations that are able to capture similar and nearly realistic aspects as conventional simulation frameworks, but with a more enriched and meaningful level of detail on the individual agent side [17]. The mentioned flexibility is achieved by the dynamic adaptation of a simulated situation. For such a MABS it is irrelevant how complex a simulation is, as long as the agents are able to couple themselves and interact with an underlying model of the environment. The latter basically reflects the overall complexity of a simulation. For example, a traffic simulation may have to deal with up to 100s, 1,000s or even 100,000s simulated cars. Moreover, agents may be provided with a higher degree of rationality or the ability to learn, which will change their reactions to occurring situations and will again increase complexity. Introductions, descriptions and several exemplary scenarios can be found, e. g., in [18], [19]. In a similar manner MABS is used in large-scale simulation scenarios when testing and validating future smart grid control concepts. These simulations feature a different degree of complexity and may vary in regard to their focus and runtime performance. In most cases it is taken into account when a performative simulation responding to a multitude of inputs is required. In parallel to classical simulations MABS may thereby be limited to simulating certain domain or use-case specific factors and calculating key performance indicators (like voltage or frequency deviations), essential to assess the regarded control. Even though the degree of complexity, which may be reproduced in MABS is sizeable, real-life implementations feature problems, which may not be considered in a simulation environment. Therefore hybrid test beds exist, which allow for the investigation of such unforeseen event. These feature a combination of real-world devices connected to a simulation environment. The operating data and in return the control commands of these real appliances are exchanged with the simulation, enabling enhanced test scenarios for the control software featuring real delays, and unforeseen events, difficult to model in a software simulation. In parallel to the described simulation environments, do the current implementations of decentralized grid control solutions have to adapt their behavior and performance in accordance to the existing electricity network. Most of the solutions envisioned do therefore form virtual power-plants in which they aggregate decentralized energy converters. Those clusters usually contain a small number of fluctuating electricity production facilities extended by some power generations equipment capable of balancing the unsteady nature of their fellow workmates. The advantage of this type of systems is their controllability in a wide power range, which makes them comparable to conventional power plants. This enables them to provide a predefined amount of energy as agreed upon in earlier negotiated supply agreements between them and external consumers of electrical energy. However, problems in regard to voltage stability may occur due to the dislocated nature of the energy production. An exemplary real-life implementation of such a system is the regenerative model-region Harz [20]. Other projects in the domain of energy management and control are trying to match the demand depending on the availability of energy from renewable energy sources. This technique is called demand side management. Most of these implementations are based on smart meters. In connection with time variant electricity rates, incentives are given to use energy-intensive applications in times of low energy

240

C. Derksen et al.

prices, which usually correlate with elevated energy availability [21]. Projects like EDeMa [22], eTelligence [23], Smart Watts [24] or the Model city of Mannheim [25] do follow this track. These projects do aim at studying the implementation of smart devices in a large infrastructure, focusing on different aspects of electricity generation, consumption and storage while paying special attention to policies and rules. All these projects have one drawback in common, which they share with the majority of alike projects in the international research community - they focus on the electricity domain only. This limitation does not pay tribute to the fact, that the different energy systems are interconnected. An electricity system has to be regarded in the context of the surrounding supply and distribution infrastructure. Not only the gas-, heat- and water network are important counterweights, featuring volatile processes, affecting each other across domain borders, but also meteorological information have to be taken into account. Examples can be found in the linkage of the electricity grid with the heating demand by e.g. night storage heating or the conversion of spare electrical energy to gas by using Power2Gas.

An Agent Development Process for Smart Energy Systems

In the following we outline a desirable, systematic development process that covers the development of agents from an initial idea to the deployment in real hardware on site. The main idea herein is that an agent can be moved through different development stages, while the immediate environment of an agent marginally changes. Similar approaches were already presented through methods like Hardware-in-the-loop simulations or Rapid Control Prototyping [26]. In addition, different process models of the Software Engineering can be applied here, as for example the V-Model as it is used by the Germany government [27]. Based on the ideas of these methods, a standardized development cycle for smart energy systems has to include the following project milestones and working packages: 1. Specification and modelling: Based on the knowledge of available subsystems, interfaces and information, the desired autonomous software-driven functionalities of an energy entity, as well as the internal and external interactions have to be described by using suitable modelling techniques like UML1, SysML2 or O-MaSE3. 2. Implementation: Following the modelling phase, a system, capable of using local information and communicating with other systems in a uniform manner, has to be implemented. This does not require the use of a specific programming language, like C++ or Java, however the communication processes have, at least, to be based on commonly-used data models and standardized interaction methods. 3. Simulation: The simulation environment should be able to embed the developed software artefact by providing similar information sources as a real system. Especially when focussing on a decentralized software-use, a simulation framework has to support the required interfaces and data provisioning for the software, in order to
1 2

http://www.uml.org http://www.sysml.org 3 http://macr.cis.ksu.edu/O-MaSE

Unified Energy Agents as a Base for the Systematic Development

241

provide or simulate the required independence of a component. For the concurrent simulation of several interacting systems the aspect of autonomy is of great importance, which emphasizes the use of agent based simulations as a precondition and the necessary simulation paradigm. Here, it must be ensured that the overall simulation works at least with a similar temporal behaviour as the real system. 4. Test bed-application: in hybrid or entire real-life test-beds the developed software artefact has to be assessed with the hardware used in a later stage. In contrast to simulations, the software is exposed to real-world conditions in terms of available computing capacity and time requirements. Here the developed smart device should be subjected to systematic testing, and so obtain a type of certification for conformity and suitability. If large scale networks have to be tested a hybrid approach, as described in section 2, may be applied. In this type of testbed selected positions are served by real life components, forming an integral part of the environment. 5. Deployment to real systems: The final application in real systems puts high demand on hardware and software requirements in terms of longevity and reliability. For this the hardware used for running the software agent should have a similar life-time as the surrounding electronic system while the software should be maintainable, e. g. through software updates.

Fig. 1. Desirable development cycle for the systematic development of Energy Agents

As described in the first point, specification and modelling already requires the knowledge about the later on controlled system, its interfaces and its information base. For the entire development process and the concrete development stage we see the actual system in which an Energy Agent is located in as intermediate environment, with accessible (and maybe standardized) interfaces. Since larger energy systems, like power plants, CHP or wind turbines already own local controlling and monitoring processes that are reacting on sensor information, we do not expect that such hard real-time requirements as defined in the domain of automation technology have to be fulfilled by software agents. This is also justified by the fact that, due to

242

C. Derksen et al.

their possible indeterministic behaviour, currently available agent systems appear to be unfavourable in critical real-time systems, demanding fixed execution cycles. Furthermore we see the main application of the agent paradigm in the coordination and harmonization of planning processes in the context of a higher level system, e. g. for the creation of timetables and a coordinated scheduling. In the foreseeable future critical real-time systems - as for example in power plants and other - will still rely on conservative systems located in the domain of the classical automation technology. Focussing on the real application of agents for smart energy systems shows that a systematic development cycle requires to make assumptions about system boundaries and interfaces that are, for the present time, quite far away from reality. But in the opinion of the authors team such standards are essential for a systematic development of future energy systems. A consistent definition of possible local sub-system boundaries and its interfaces for an autonomous software system called Energy Agent would thus help to better understand the roles and functions of such an entity. Furthermore it would be possible - on the long term - to provide comparable scientific results and subsequently allow a sustainable conversion of our energy infrastructure.

Definition and Scope of an Energy Agent

After describing its local environment above, we are now focussing on the definition and the thinkable internal processes and structure of an Energy Agent in a global context. Following [17] and the VDI directive 2653-1 we define thus: An Energy Agent is a specialized autonomous software system that represents and economically manages the capacitive abilities of the energy consumption, production, conversion and storing processes for a single technical system and that is embedded and thus part of one or more domain specific energy-networks, capable to communicate therein and with external stakeholders. This definition is based on the considerations of the previous section and with the goal to define a generalized prototype for an Energy Agent that is applicable in various situations and in different energy domains. With the concept of energy domains, we consider not only electrical networks (e. g. Smart Grids) to be more sophisticated in the sense of more dynamic in their resource allocation by using an additional piece of software. Rather, the Energy Agent is supposed to integrate all those energy systems that are subjected to the fundamental first law of thermodynamics, stating that energy can neither be created or destroyed, and which are thus able to store energy or convert it from one form to another. Mathematically written, this means that the sum of consumed and emitted energy of a single system is always equal 0 over time. Thus energy can neither be created nor destroyed. Formula 1 resumes this circumstance.

Ei = 0

(1)

This theorem of the conservation of energy applies for every system that converts energy from form i to another, like mechanical energy to electrical energy (electricity generator) or chemical energy to thermal energy (heating system). Taking into

Unified Energy Agents as a Base for the Systematic Development

243

account this relationship, we assume all following Energy Agents (and their underlying technical systems) to be individual, but connected by (possibly) hybrid transport networks, each with its specific form of energy. In the sense of the first law of thermodynamics and the capacitive abilities of an energy system, the main task for an Energy Agent, connected to a network, is its ability to describe and handle the individual positively or negatively directed (produced/emitted or consumed/obtained) amount of energy over time, maybe in cooperation with its neighbouring entities. This base-requirement leads to the need of a well defined, suitable generic data or option model that allows an Energy Agent (or a later external optimization procedure) to utilize this information. Derived from this requirement is the question of the necessary and sufficient understanding of time and the system changes over time. On the basis of current implementations of data acquisition systems and for safety reasons it is clear that time has to be discretized and monitored somehow (possibly in equidistant steps) and that it will be dependent on the managed energy usage or production processes. What an ideal step size could be depends basically on the dynamics of the handled amount and type of energy and thus on the physics of the underlying technical system. Before we discuss a more substantial integration of the energy option model as needed by the Energy Agent and introduced above, we want to divide our definition of an Energy Agent in different levels of sophistication first. These levels describe the local capabilities of onsite software. By doing so, we want to reflect the fact that current systems are usually not equipped with functionalities that could enable complex interactions with their environment, as they are commonly envisioned in the current scientific discussion. For reasons of illustration we define in the following table a coherent set of integration levels for Energy Agents by means of an electricity meter; functionalities and limitations may be transferred to other energy domains as well.
Table 1. Integration level of Energy Agents Integration Level IL0 Overall Control Central Description

IL1 IL2 IL3

IL4

IL5

Initial situation: old state of the art from the 80s (e. g. Bakelite ferrite electric meters and newer meters without information exchange) Central Current meter systems: enables information transfer of energy usage, but requires a central data analysis Central Advanced meter systems with predictions: enables information transfer of energy usage with locally aggregated data Central & Advanced local controller: Can act on the underlying local sysLocal tem and react autonomously to external signals (e. g. price signals for local optimization) Central, Advanced local area controller: restricted, but independent local Distributed systems that can dynamically build coalitions in order to keep & Local track of optimization goals (e.g. intelligent local power transformers, responsible for one network segment) Distributed Fully distributed control of energy production, distribution and & Local supply

244

C. Derksen et al.

The above classification shows that the consideration of an Energy Agent is advisable if additional control processes are required on a local level. Accordingly, Energy Agents are only necessary from integration level IL1 onwards. Additionally, the table shows that with an increasing integration level a shift to a more decentralized control of energy supply systems is addressed. Integration level IL5 describes thereby a fully autonomous and decentralized energy supply system, based on the most sophisticated Energy Agents. The latter is primarily an extreme, of which we assume that it not will be fully implemented in the near future. Rather we assume that a multitude of future energy supply systems and their parallel development will lead to a concurrent presence of differently sophisticated Energy Agents. This results in another hybrid dimension and thus a further degree of complexity for future energy supply systems. Taking into account the high diversity of the systems to be considered in the context of the Energy Agent definition, different levels of sophistication can already be observed today. Independent of recent scientific solutions for energy-wise autonomous and self-aware systems, we can observe that todays technical reality is mostly located in the integration level IL0 and IL1: The broad mass of households is still not equipped with smart-meters that are able to communicate their energy consumption. Larger automated plants are designed to follow one or more given set points with their local control systems. The information transfer to central control points is realized by using existing connectivity, while an evaluation of this data is often done offline and only by partly automated processes that were not necessarily designed to optimize energy usage. Even though these degrees of sophistication may not be seen as use cases for agents at mere sight, the usage becomes clear when regarding integrated systems containing these components. Therefore these basic Integration Levels are needed for simulating current integrated layouts and systems. In order to have a common starting point for discussions and comparisons in the framework of a systematic development approach for future energy systems, the classification of an Energy Agents integration level as discussed briefly in this article will be of great support for reaching a common understanding, especially when refined in future research. In the above context, another important question lies in the design objective of an Energy Agent or even of a group of Energy Agents, as indicated for the integration levels IL3 to IL5. Here, purely technical or economical goals can be identified and differentiated in the course of developing a holistic system. While technical specifications mostly only describe the parameter set in which a technical system can operate, technical parameters are to be seen as constraints that have to be fulfilled in every case during operation. After fulfilment of such constraints, an optimization can take place. Assuming a kind of cost or income optimization of an Energy Agent, a relation between energy conversion and storage processes and a corresponding accounting system have to be integrated. A high level quantification of energy flows in units of watt or watt seconds requires an explicit calculation, based on the domain specific form of energy. While for electrical networks voltage und current are the important base units for such calculations, the gas quality and the physical states for pressure, temperature and other

Unified Energy Agents as a Base for the Systematic Development

245

factors are important for gas transportation networks. Thus such detailed domain specific information is also required by an Energy Agent, especially if an energy flow is considered to exceed the boundaries of the underlying single system. In summary, from the above considerations, this results in the following systemic picture for an Energy Agents and for the needed, aggregated information that have to be gathered, processed and deployed by the internal control processes (see Fig. 2): Running on available or additional hardware, the Energy Agent is connected to the technical system and can partly or fully control and monitor it. By using the domain specific knowledge about the local technical system over time, the agent is able to synthesis an energetic option and cost model (see next section) that can either be used for local limit compliance or for cost optimizations. A further sophistication of an Energy Agent, as it was described with IL3, would allow communication as for example for the processing of external price signals or for coordination processes and negotiations with neighbouring entities, in order to comply with non-local goals.

Fig. 2. Structure, interfaces and processes of an Energy Agent of IL3 and higher

Option Models: What Energy Agents Can Provide or Not

As stated in the previous section, the description of the time-dependent and quantified energy-related abilities of an Energy Agent or the respective underlying technical system requires summarizing domain-specific information and an abstraction to a high-level, domain-independent system. The necessity for this is derived from the individual optimization goal of Energy Agents at integration level IL3 and above, where a clear relationship between energy price and usage is addressed. Especially, the superior levels IL4 and IL5 require an abstract means of information exchange, in order to communicate their energetic degree of freedom and maybe to negotiate for their concrete energy consumption or production. This particularly applies for the

246

C. Derksen et al.

cross-domain exchange of information and energy, as for example in all relevant intermediate technologies between electricity and gas. Consequently, a unified data model is required that enables the description of individual degrees of freedom and their energy-related abilities of single entities over time. This means first of all that individual systems (or their managing Energy Agents) need to know their technical capabilities, which is trivial for simple consumer such as bulbs, but more difficult for complex systems like cogeneration plants and other. Based on the information, this internal model has to be extended by a temporal behaviour of a single entity. Again, this is simple for a bulb since it can be easily switched on or off, if the control over the bulb is given and if there are no constraints that require the bulb to be permanently switched on. For more complex technical systems, this becomes more difficult since such systems are subject to their current operation state and can not simply be moved to any other operation point without taking into account a multitude of (possibly also external) facts. Therefore, transitive states have to be considered in addition, as for example start-ups, shutdowns, maintenance periods and other. For the creation of energetic option models, we propose a fundamental twodimensional subdivision that describes either the typical characteristic of an energy system, while the second dimension classifies a possible temporal behaviour. Here, a possible classification for the energy systems consumption, production and its storage capabilities (C. P. S.) could be as follows:
Table 2. Classification of the characteristics of energy systems Energy C. P. S. Constant working systems Task dependent or batch systems Repetitive systems Environment dependent systems Dynamic and flexible ondemand-systems Examples from different domains bulbs, irons, certain types of power plants and other white goods (like washing machines and dishwasher), industrial facilities (start-ups and shutdowns) and other fridges, central heating wind turbines, photovoltaic plants and other pumped storage plants, gas power plants, gas turbines, compressors, gas storages

A possible temporal behaviour of a single energetic system could range between an uncontrollable usage or production of energy on one side and a fully time-controllable system on the other. Thus, the question is addressed how predictable an energyrelated system is and who (or what) controls the system. In addition to manually and thus - from a control perspective - ad hoc operating systems, semi-automated and fully automated systems have to be considered here. Further aspects are the ability to dynamically reduce, increase, pause, shift, switch on or switch off the energy flow at the boundaries of a single system. The medium-term objective of the above classifications is to build up example cases that can help to create a unified model that is capable of describing all of these cases and that, based on domain-specific information, enables interaction processes

Unified Energy Agents as a Base for the Systematic Development

247

based on a commonly known grammar or language that can be used by every Energy Agent. A necessary extension to that grammar is a corresponding and first internal cost model for actions of Energy Agents and their underlying technical systems. Based on the above described flexibility aspects of Energy Agents, the question has to be discussed if these internal cost models have to be disclosed, converted and communicated in a standardized manner, in order to enable market oriented control mechanisms and thus the formation of Smart Markets, similar to existing systems featuring automated trading processes. The concrete implementation of a comprehensive, energetic option model is task of our further work and will be presented in a future publication.

Conclusion

In the previous sections we have presented a novel approach to and perspective onto the topic of energy transformation and -storage control design and implementation. The suggested Energy Agent as autonomous software system is described as a design concept with several degrees of freedom, covering the majority of use cases in different energy domains. It is supposed to represent and economically manage the abilities of energy related technical systems and shall be capable of communicating with other stakeholders. For evaluating the functions and capabilities required by the individual Energy Agent several integration levels, referring to the level of evolution, are defined. Six integration levels are differentiated reaching from IL0, denoting entirely local systems incapable of communicating and higher logic, to IL5, representing completely decentralized systems, distributing data acquisition, knowledge and decision onto a variety of entities. Based on the presented integration levels an analysis of information relevance and flow in a generalized energy management environment is performed. The outcomes are summarized in a schematically model of the Energy Agent illustrating the most important steps regarding information acquisition, processing and deployment while complying with given constraints in regards to the encountered technical and economical context. Derived from this model an option model is outlined, supporting the classification of energy usage patterns, allowing for an efficient provision and processing of vital information. Furthermore a systematic development approach covering the core functions of the Energy Agent was introduced. This approach is based on strategies developed and successfully deployed in the fields of automation technology. It comprises the following steps: specification and modelling, implementation, simulation, test bedapplication and finally the deployment to real systems. This work is the offspring of a scientific cooperation, in which a decentralized energy- and load- management system was linked to an agent based multi domain simulation. In the course of the study several pitfalls and shortcomings in regard to communication and data management processes in the energy domain were identified. The subsequently envisioned Energy

248

C. Derksen et al.

Agent is supposed to serve as a basis for further development and discussion in the domain of decentralized energy management. Questions in regard to standardization shall be raised with a special focus on communication and system design. While systematically specifying the Energy Agent in further iteration cycles, generalizable requirements shall be derived, supporting the development of future energy control systems. Until now the described Energy Agent lacks a proper definition of stakeholders and their corresponding user roles and privacy requirements. These factors shall be taken in account when implementing the first exemplarily release in the near future.

References
[1] Ilic, M.D., Allen, H., Chapman, W., King, C.A., Lang, J.H., Litvinov, E.: Preventing Future blackouts by Means of Enhanced electric Power Systems Control From Complexity to Order. Proceedings of the IEEE (11) (2005) [2] Setiawan, E.A.: Concept and Controllability of Virtual Power Plant. Kassel University Press (2007) [3] Nestle, D., Ringelstein, J.: Integration of DER into Distribution Grid Operation and Decentralized Energy Management. In: Smart Grids Europe 2009, Barcelona (March 19, 2009) [4] Daneels, A., Salter, W.: What is SCADA? In: International Conference on Accelerator and Large Experimental Physics Control Systems, Trieste (1999) [5] Huhns, M.N., Stephens, L.M.: Multiagent Systems and Societies of Agents. In: Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence. The MIT press (1999) [6] Jennings, N.R., Wooldridge, M.: Applications of intelligent agents. In: Agent Technology. Springer (1998) [7] Keilar, N.: Concept to Integrate Distributed Generation into the German Power Grid Intended for Multi-Agent System Realisation. Thesis, http://www.hsu-hh.de/ evhs/index_WCPxnDvauo2WCG70.html (version: April 2013) [8] Li, H., Sun, H., Wen, J., Cheng, S., He, H.: A Fully Decentralized Multi-Agent System for Intelligent Restoration of Power Distribution Network Incorporating Distributed Generations. IEEE Comput. Intell. Mag. 7, 6676 (2012) [9] Kok, K., Warmer, C., Kamphuis, R., Mellstrand, P., Gustavsson, R.: Distributed Control in the Electricity Infrastructure. In: Proceedings of the International Conference on Future Power Systems (2005) [10] Lehnhoff, S.: Dezentrales vernetztes Energiemanagement - Ein Ansatz auf Basis eines verteilten Realzeit-Multiagentensystems. Vieweg + Teubner (2010) [11] Platt, G.: The Decentralised Control of Electricity Networks- Intelligent and Self-Healing Systems. In: Grid-Interop-Forum 2007, pp. 16 (2007) [12] Linnenberg, T., Wior, I., Schreiber, S., Fay, A.: A Market-Based Multi-Agent System for Decentralized Power and Grid Control. In: Proceedings of IEEE-ETFA (2011), doi:10.1109/ETFA.2011.6059126 [13] Derksen, C., Branki, C., Unland, R.: A framework for agent-based simula-tions of hybrid energy infrastructures. In: Ganzha, M., Macias-zek, L.A., Paprzycki, M. (eds.) FedCSIS (2012)

Unified Energy Agents as a Base for the Systematic Development

249

[14] Sudeikat, J., Braubach, L., Pokahr, A., Lamersdorf, W., Renz, W.: Validation of BDI agents. In: Bordini, R.H., Dastani, M., Dix, J., El Fallah Seghrouchni, A. (eds.) PROMAS 2006. LNCS (LNAI), vol. 4411, pp. 185200. Springer, Heidelberg (2007) [15] Wooldridge, M.: An Introduction to MultiAgent Systems, 2nd edn. Wiley & Sons (July 2009) [16] Moss, S., Davidsson, P. (eds.): MABS 2000. LNCS (LNAI), vol. 1979. Springer, Heidelberg (2001), http://books.google.de/-books?id=l5eRqAZh0JgC [17] Davidsson, P.: Multi agent based simulation: Beyond social simulation. In: Moss, S., Davidsson, P. (eds.) MABS 2000. LNCS (LNAI), vol. 1979, pp. 97107. Springer, Heidelberg (2001) [18] Bonabeau, E.: Agent-based modeling: methods and techniques for simulating human systems. Proceedings of the National Academy of Sciences of the United States of America 99(suppl. 3), 72807287 (2002) [19] Drogoul, A., Vanbergue, D., Meurisse, T.: Multi-agent based simulation: Where are the agents? In: Sichman, J.S., Bousquet, F., Davidsson, P. (eds.) MABS 2002. LNCS (LNAI), vol. 2581, pp. 115. Springer, Heidelberg (2003) [20] RegenerativeKraftwerke Harz: REG MOD HARZ. Rege- nerative Modellregion Harz Homepage, https://www.regmodharz.de/das-projekt/ (version: April 2013) [21] Faruqui, A., Sergici, S.: Household Response to Dynamic Pricing of ElectricityA Survey of the Experimental Evidence. Journal of Regulatory Economics (2013), http://www.hks.harvard.edu/hepg/Papers/2009/The%20Power%20of %20Experimentation%20_01-11-09_.pdf (version: April 2013) [22] Laskowski, M.: E-DeMa development and demonstration of locally networked energy systems to the E-Energy marketplace of the future. E-DeMa Homepage, http://www.e-dema.com/en/index.html (version: April 2013) [23] AG, EWE: eTelligence. eTelligence Homepage, http://www.etelligence.de/ etelligence.html (version: April 2013) [24] Utilicount GmbH & Co. KG: Smart Watts - Startseite. Smart Watts Homepage, http://www.smartwatts.de/home.html (version: April 2013) [25] MVV Energie AG: moma The power supply system gets intelligent. Modellstadt Mannheim Homepage., http://www.modellstadt-mannheim.de/moma/ web/en/home/index.html (version: April 2013) [26] Boehm, B.: Guidelines for verifying and validating software requirements and design specifications. In: EURO. IFIP, vol. 79, pp. 711719 (1979) [27] Abel, D., Bollig, A.: Rapid Control Prototyping - Methods and Application. 1. Aufl. Springer, Berlin (2006) ISBN 3-540-29524-0

You might also like