In the meantime, agents and agent technologies are widely accepted as a conceptual approach for controlling energy systems in a Smart Grid context, and many research projects and pilots have already shown their applicability. From a global scope however, building proprietary software artifacts should not be the goal in critical system environments like the energy supply. Rather, it should be intended to build verified, scalable and sustainable on-site software that is able to face the different requirements coming from the affected disciplines, like engineering, automation, energy markets and other.
On-site software agents should be systematically tested before they are applied in real systems, which implies to validate them in simulations or automation testbeds. Furthermore, they should be able to adapt, since organizational or market rules may change over time – as it is the case with the introduction of an energy blockchain. This requires suitable tools and structures that continually support the life cycle of such agents.
A general and comprehensive tool for designing, developing, testing and deploying such agents was still missing. As a result, currently developed agent systems are not able to interact with other systems or provide adaptivity, which, in turn, hinders to discuss required data and behavioral standards for the control of distributed energy system, like distribution grids or Smart Homes. Thus, it basically increases uncertainty and costs, while it slows down the overall progress in research and industry.
With the goal of bridging the gap between research and industry, we designed and developed a holistic approach that focusses on monitoring and controlling the flexibilization potentials of single energy conversion systems. We defined the core abilities and the structure of the required software agents and introduced the resulting software artefacts as so-called “Energy Agents”; an earmarked autonomous software component that may be used to control distributed energy systems in a decentralized or centralized manner.
Furthermore, we enabled our Energy Agents to dynamically generate aggregations of technical systems, which in principle allows us to hand-over the control of any type of system aggregation to Energy Agents. Thus, focusing on single energy systems first and aggregate them as “system of systems” by considering different energy carriers like electricity, heat and gas enables us to tackle a wide application ranges in the energy sector, as for example Demand Side Management, Demand Response, Virtual Power Plants, Smart Homes, the control of electrical distribution grids and other.
Methodology and Evaluation
Within the funded project Agent.HyGrid (see www.agent-hygrid.net), we developed and refined the systematic development process for Energy Agents in the last three years. Within this process, the developer can first develop and improve Energy Agents in interactive, time-dependent simulated scenarios, considering the complex system relationships (e.g. network states). After a functional validation in the simulation environment, these Energy Agents can be deployed to dedicated hardware, which enables testbed applications and further validations (e.g. Hardware-In-The-Loop, Software-In-The-Loop, test of hardware sizing with respect to the required calculation abilities). Finally, Energy Agents may be applied on-site, where they then can monitor or control associated systems or aggregations of systems.
In the context of the above project and the former years, a comprehensive framework in combination with an end-user toolkit and a runtime environment was created. Here, a general-purpose framework for agent-based application – Agent.Workbench – formed the basis for the specialization towards the energy domain. Agent.Workbench is based on the Eclipse Rich Client Platform architecture, making use of its’ OSGI-based modular structure to provide adaptability and extensibility. Theron based, an energy system modelling framework – the Energy Option Model (EOM) – was added as a feature. Herewith, Energy Agents can be enabled to in-detail “understand” the flexibility potentials of energy systems or aggregations and use this knowledge for planning or real-time control. On top of this, the structure and core abilities of Energy Agents were defined and bundled in the Agent.HyGrid feature, which in turn enables to extend and customize Energy Agents by concurrently using the above described development process, covering simulations, testbed-applications and the on-site usage.
The combination of OSGI/Eclipse-based modularization, the update mechanisms of p2 and the automated deployment process for Energy Agents covers the whole software lifecycle of Energy Agents. Thus, the Energy Agent and the toolchain represent a sustainable approach that can be utilized and extended over years, since the knowledge about technical systems and processes remains, even if the regulatory framework or markets evolve.
Tutorial & Hackathon Content
During our tutorial, we will introduce the core concepts and functionalities of Energy Agents. Further, we will introduce an example scenario and demonstrate the abilities of the simulation environment. In addition to the introduction of the concepts, we would like to give you (above all) the opportunity to develop your own solutions with our toolchain for Energy Agents in the future.
That’s why you as a participant of the hackathon and tutorial will be asked to prepare your Notebook and join to exchange and discuss. Further details will be provided over the SEN-MAS site (see www.sen-mas.org) and LinkedIn over the next weeks.
If you have further questions or suggestions, please join the SEN-MAS groups at LinkedIn (see https://www.linkedin.com/groups/13577165) or post your question in the Issue-corner of the Agent.Workbench project at github (see https://github.com/EnFlexIT/AgentWorkbench/issues, please mark your “issue” with the label “question” on the right hand site) .
Conclusion and Future Work
The framework and toolchain for Energy Agents represent a powerful approach to support the required joint learning process for research and industry in the energy sector. Such joint learning and development, we believe, is required to achieve commonly accepted standards for “smart” energy systems and avoid proprietary solutions that lead to artificial area monopolies in liberated markets.
Based on our approach, we plan to further improve the domain specific abilities of our frameworks (Agent.Workbench and the EOM) in funded and industry-related projects. Furthermore, we would like to share and improve the overall topic, since the possible application ranges exceed our own personnel capacities – especially because it is required bringing together experts from different domains.