A thermostatically controlled device in EPOS is able to generate a number of possible plans. These are possible energy consumption patterns that the operation of this device can support for the next period of time. A device selects the possible plan that contributes best to the stabilization of the total energy consumption. This can be achieved by a brute force computation of the combinations between all the possible plans for every device in the overlay network. Each combination is evaluated by summing all of the possible plans and computing the deviations of the total energy consumption. This approach guarantees the optimum selection of possible plans and therefore provides a maximally stabilized total energy consumption given certain possible plans.
However, the number of these combinations is very large and therefore the optimum combination cannot be computed efficiently using a brute force operation. Instead, EPOS first computes combinations between a small group of devices and then selects a combination based on the selections that other groups of devices have already performed. Technically, this adaptation is realized by summing the energy pattern consumption that a group of devices selected to each combination of possible plans that another group of devices computes. Therefore, the evaluation of deviations in each combination is more informative as the selections of possible plans by other devices are considered.
Aggregation is performed in a decentralized fashion based on peer-to-peer interactions between the devices. Two possible aggregation schemes are considered in EPOS:
- Structured aggregation using an overlay network organized in a tree topology. AETOS, the Adaptive Epidemic Tree Overlay Service, is an example of how tree topologies can be self-organized. EPOS is evaluated experimentally using this aggregation scheme.
- Unstructured aggregation using a dynamic overlay network that is able to (i) detect new information for aggregation, (ii) exclude duplicate information and (iii) update information that changes during runtime. Unstructured aggregation in EPOS is part of ongoing work.
The evaluation of EPOS shows that the stabilization of the total energy consumption improves compared to a system in which devices randomly select a possible plan or simply select their most stabilized possible plan. In the future Smart Power Grid, decentralized energy management mechanisms such as EPOS, can make the power network more robust to oscillations and peaks in energy consumption.
EPOS started as part of my PhD studies resulted in several publications and collaborations. It is still evolving work.
- Evangelos Pournaras, Mark Yao, Dirk Helbing, Self-regulating Supply-Demand Systems, Future Generation Computer Systems, 2017 © Elsevier (to appear)
- Peter Pilgerstorfer and Evangelos Pournaras, Self-adaptive Learning in Decentralized Combinatorial Optimization-A Design Paradigm for Sharing Economies, in the Proceedings of the 12th International Symposium on Software Engineering for Adaptive and Self-managing Systems-SEAMS-2017, Buenos Aires, May 2017
- Marinos Koutsomichalis and Evangelos Pournaras, The Sound of Decentralization-Sonifying Computational Intelligence in Sharing Economies, in the proceedings of the 23rd International Symposium on Electronic Art-ISEA-2017, Manizales, Colombia, June 2017
- Akshay Uttama Nambi, Evangelos Pournaras and RangaRao Venkatesha Prasad, Temporal Self-regulation of Energy Demand, IEEE Transactions on Industrial Informatics, Vol. 12, Nr. 3, pp. 1196-1205, 2016 © IEEE [Supplementary Information] (Nominated to IES ITeN )
- Evangelos Pournaras, Matteo Vasirani, Robert E. Kooij and Karl Aberer, Socio-technical Trade-offs in Self-regulating Smart Grids, in the proceedings of the International Conference on Computational Social Science-ICCSS 2015, Helsinki, Finland, June 2015 (compressed contribution of earlier papers)
- Evangelos Pournaras, Matteo Vasirani, Robert E. Kooij and Karl Aberer, Decentralized Planning of Energy Demand for the Management of Robustness and Discomfort, IEEE Transactions on Industrial Informatics, Vol. 10, Nr. 4, pp. 2280-2289, 2014 (to appear) © IEEE
- Evangelos Pournaras, Matteo Vasirani, Robert E. Kooij, Karl Aberer, Measuring and Controlling Unfairness in Decentralized Planning of Energy Demand, in the proceeding of the IEEE International Energy Conference-EnergyCon 2014, Dubrovnik, Croatia, May 2014. © IEEE [Presentation]
- Evangelos Pournaras, Multi-level Reconfigurable Self-organization in Overlay Services, PhD Thesis, Delft University of Technology, March 2013 (Chapter 6)
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, Local Agent-based Self-stabilisation in Global Resource Utilisation, International Journal of Autonomic Computing, Vol. 1, Nr. 4, pp. 350-373, 2010 © Interscience Publishers
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, A Distributed Agent-based Approach to the Stabilization of Global Resource Utilization, in the proceedings of the International Conference on Complex, Intelligent and Software Intensive Systems-CISIS 2009, pages 185-192, Fukuoka, Japan, March 2009. © IEEE [Presentation]
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, Towards Emergent Energy Synchronization using Agents, in the proceedings of the International Workshop on Agent Technologies for Energy Systems-ATES 2010, Toronto, Canada, May 2010 [Presentation]
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, Using intelligent agents for self-adaptation and self-optimization of energy consumption in power networks, in the proceedings of the International Workshop on Agents for Autonomic Computing-AAC 2008, Chicago, USA, June 2008 [Presentation]
- Evangelos Pournaras, Martijn Warnier, Frances M.T. Brazier and Elth Ogston, Towards Adaptive Energy Plan Aggregation over a Peer-to-Peer Tree Overlay, in the proceedings of the Performance for Peer-to-Peer Systems Workshop-P4P2P 2008, Warwick, UK, May 2008 [Presentation]
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, Energy Consumption Stabilization by Agent-Based Decentralized Tree Aggregation, poster presentation at SIREN 2008 Symposium, Vrije Universiteit Amsterdam, September 2008
- Evangelos Pournaras, Martijn Warnier and Frances M.T. Brazier, A Distributed Agent-based Approach to the Stabilization of Global Resource Utilization, in the proceedings of the 21th Belgian-Dutch Conference on Artificial Intelligence-BNAIC 2009, Eindhoven, The Netherlands, October 2009 (compressed contribution of an earlier paper) [Presentation]