Report

Teaser, summary, work performed and final results

Periodic Reporting for period 2 - CONENE (Control of Large-scale Stochastic Hybrid Systems for Stability of Power Grid with Renewable Energy)

Teaser

Increasing penetration of renewable energy sources and liberalization of electricity markets are significantly changing power system operations. Analysis and control of this system is highly challenging because of (a) large dimensions arising from a complex network of...

Summary

Increasing penetration of renewable energy sources and liberalization of electricity markets are significantly changing power system operations. Analysis and control of this system is highly challenging because of (a) large dimensions arising from a complex network of transmission and distribution lines and generators, (b) uncertainties, such as fluctuating power output from weather-based renewable generation, and (c) multi-agent complex interactions due to the participation of a large number of producers and consumers with individual objectives and coupling constraints. My research agenda is focused on developing fundamental theory and practical algorithms for control of large-scale stochastic multi-agent systems in order to guarantee the stability and efficiency of the power grid.

Work performed

The work performed has been perfectly aligned with the goals of my ERC project. In particular, I have significantly advanced fundamental understanding of multi-agent decision making under uncertainty, developed novel scalable and provably safe algorithms for these systems and have verified the methods on realistic power system simulation testbeds in collaborations with power system researchers. The work can be discussed in four main categories.

Distributed control of large-scale uncertain systems: First, we quantified necessary and sufficient conditions for tractable formulation of the problem of optimal distributed control under safety constraints [Furieri, Kamgarpour, 2017]. Second, for the class of systems that do not satisfy these stringent conditions (such as power systems) we developed a tractable control synthesis algorithm with provable performance guarantee [Furieri, Kamgarpour, 2018]. Third, we developed a scalable approach using advanced graph theoretic decompositions, for verification and control synthesis for large-scale systems [Zheng, Kamgarpour, Sootla, Papachristodoulou, 2018].

Multi-agent decision making in electricity markets: First, we have advanced tools from economics and in particular game theory and auction theory, to analyse electricity markets and design market mechanisms with improved efficiency [Karaca, Kamgarpour, 2017]. Second, we have utilised advances in mechanism design mainly developed for auction of multiple items such as online adverts, to markets involving continuous goods, stochastic costs and constraints [Karaca, Kamgarpour, 2018]. Third, we have verified our proposed mechanisms with several realistic electricity market data, including the well-established IEEE LMP markets as well as Swissgrid control reserve markets. Leveraging advances in computer science, we further developed novel conditions on the electricity markets to minimise potential of strategic manipulations, including shill-bidding and collusion [Karaca, Kamgarpour, 2018].

Learning in multi-agent systems: To understand how multi-agent systems can learn optimal decision making in an uncertain dynamic environment, we developed a novel distributed decision-making algorithm with provable convergence guarantees to an equilibrium state [Tatarenko, Kamgarpour, 2017, 2018]. The proposed algorithm requires the agents to only observe their payoffs (or costs) for their played actions, rather than knowing the functional form of the costs or constraints. We showed the applicability of the method in electricity markets, where prosumers aim to optimize their production/consumption profiles, without knowing the functional form of the price of electricity a priori. Using observed prices, they can adjust their profiles to ensure convergence to optimality.

High-Voltage-Direct-Current (HVDC) power line placements: We proposed an approach to determine placement of offshore HVDC links into a multi-terminal electricity network for maximizing stability of the DC voltage [Elahidoost, Furieri, Kamgarpour, Tedeschi, 2017, 2018]. We have verified our proposed approach in a standard IEEE HVDC network testbed. Furthermore, given that renewable energy intermittency and uncertainty aspects are decisive criteria in HVDC expansion problems with high penetration of solar or wind, we accounted for the power injection from connected wind farms using historical data. To this end, we tested our proposed approach with wind data from the North Sea wind farms as well as the current configuration of these wind farms.

Final results

The CONENE research has been extremely successful and rewarding, in terms of developing general theory for control of multi-agent systems as well as practical applications of the theory to power system analysis and control. We are thrilled to continue to unravel theoretical challenges and to develop novel algorithms to this end. In particular, my main goals for the remaining duration of the project are to address two ambitious goals. First, I aim to develop a local synthesis approach to realise the optimal distributed controllers we have identified for large-scale systems. In particular, almost all past research has focused on optimality conditions and characterisation of distributed controllers. There has been very little work on how to compute these controllers using local information and while respecting data-privacy of each subsystem. I plan to address this challenge with my doctoral student Luca Furieri, as well as my Oxford collaborations. Second, I aim to develop novel pricing and incentive mechanisms for electricity markets with nonlinear and stochastic constraints. In particular, almost all proposed mechanisms have analytical properties only for the linearised power flow models and under deterministic conditions. Furthermore, the proposed aggregation mechanisms do not discuss how to fairly distribute the market deficit/surplus to players in such complex constraint setting. With my doctoral student Orcun Karaca and my collaborators in Technical University of Denmark, we plan to embark on this problem. I look forward to sharing the successes in the final reporting period.