Real-time monitoring of electricity grids' power flow, which reflects the physical reality of the power system, plays a crucial role in the power market, since the real-time market behavior often deviates from long-term market forecasts, due to unexpected supply-demand imbalances and the resulting price volatility. The real-time market results, in turn, have a major influence on the optimal dynamic economic dispatch of the power generation for stabilizing the power load in the grid. However, an appropriate market-grid coupling, in terms of a real-time interaction between the market and the grid, has not been designed to be available either from the grid network side or from the market structure side. In particular, in the context of Demand Response (DR), an incentive-driven load shedding or shifting for grid relief cannot be realized without an appropriate market-grid coupling. In this dissertation, a feedback control concept is proposed, designed and evaluated for modeling a market-grid coupling. The dissertation, also, addresses the research question of whether the market price as a feedback signal can effectively control the power dispatch in the grid, and vice versa. Recently, researchers have focused primarily on investigations of a complex interaction between the market and the grid, in terms of interoperability or controllability. This dissertation addresses rather a combination of both interoperability and controllability; namely an interoperable control between the market and the grid, by means of a closed-loop feedback control system. Essentially, this dissertation presents a novel approach with a closed-loop feedback control concept for the distributed market-grid coupling. One important part of the main contribution of this dissertation is the formal definition of the market-grid coupling. As the first requirement for the market-grid coupling, a real-time market model is designed and formulated as a power balancing option; Subsequently, a two-layer grid model is presented for an optimal dynamic dispatch (ODD) study. Based on both models, a definition of the market-grid coupling is formalized with a feedback control loop. Then, a further investigation and analysis of this formalized market-grid coupling is conducted in two different directions. A co-simulation framework that realizes a market-grid coupling is developed for studying the grid load's influence on the market price. In order to extend this unidirectional control towards an interoperable control between the market and the grid within the market-grid coupling framework, the system modeling of a MPC-based closed-loop feedback control system is presented, in which a market price optimization and a power dispatch optimization are performed concurrently. The problem formulation of the control system firstly focuses on a coupling model with a single grid unit and its correspondent local market. Subsequently, a distributed control architecture by means of a hierarchical MAS (Multi-Agent System) is presented for extending the centralized MPC problem of a local market-grid coupling to a distributed MPC problem of a distributed market-grid coupling. A distributed MPC strategy is adopted to decompose the overall grid into interconnected grid units, so that individual grid units achieve control objectives collaboratively. Different valuation use cases with IEEE bus system test cases are introduced. Simulation-based numerical results show that not only the centralized MPC formulation, but also the distributed MPC formulation, provide a clear stability of both the market price and the power load dispatch. Finally, an adaptive load forecasting framework is proposed to improve the STLF (Short-Term Load Forecasting) performance. The obtained accurate load forecasting result shows its benefit for solving the above MPC problems.
About the Author: Yong Ding received his Ph.D. degree of Computer Science in 2016 at Karlsruhe Institute of Technology (KIT) and his Diploma Degree of Electrical Engineering in 2008 at University of Karlsruhe. His employment experience included 15 months Trainee Program in the energy company RWE and more than one year project management of electrical engineering in the division of conveyor equipment and process data processing. His research focus is on context recognition and prediction in distributed systems as well as predictive control methods. Specially, he is interested in the combination of machine learning techniques and automatic control concepts. On the application side, the topics in the field of Smart Grid, Smart City and Big Data are his main focus.