The power grid is expected to undergo major transformations due to the increased penetration of renewable variable energy sources and electric vehicles. Uncertainty caused by the volatility of renewable energy production requires the adoption of new measures to maintain the balance between supply and demand, thus guaranteeing the reliability and stability of the grid. These challenges require the active engagement of consumer-side energy production and consumption to provide sufficient flexibility for the power grid through a mechanism known as demand response(DR). This dissertation focuses on DR, and more specifically on enabling residential consumers to participate in the balancing of the grid by providing frequency containment reserves. To provide frequency containment reserves, the dissertation determines the functional and non-functional requirements for using domestic energy storage resources. The functional requirements are identified in two use cases. The first use case defines the requirements for the planning phase of the DR, while the second specifies the requirements for the frequency reserves provision. In addition, non-functional requirements are specified for the developed DR system. The dissertation proposes the design of a DR system for providing frequency containment reserves based on the specified requirements. The design defines a hybrid ICT architecture capable of functioning during both the planning and execution of DR. For DR planning, the dissertation employs a partially distributed optimization algorithm to enable the day-ahead scheduling of consumer-owned energy storage resources. For frequency reserve provision, the dissertation contributes by integrating the proposed ICT architecture with a hybrid coordination algorithm. This auction-based task allocation algorithm enables consumer-owned energy storage resources to be allocated for the provision of frequency containment reserves. The developed DR system is validated through simulations in the two use cases. The dissertation investigates the required market intelligence for enabling a virtual power plant(VPP) to profitably exploit distributed energy resources. In the future, the VPP could exploit its resources on various markets, including frequency containment reserves. In order to bidintelligently on such markets, VPP should be capable of predicting market prices. Therefore, the dissertation focuses on providing a solution for predicting the prices of the frequency containment reserve for normal operation market. The dissertation provides a data-driven solution, analyzing the market prices and providing a methodology for predicting the day-ahead prices through the design and implementation of various machine learning regression models, including an artificial neural network model. The designed models are evaluated by comparing the prediction performance in an experimental setup.
|Tila||Julkaistu - 2019|
|OKM-julkaisutyyppi||G5 Tohtorinväitöskirja (artikkeli)|