Market Integration of Virtual Power Plants

Global efforts to reduce emissions of carbon dioxide drives the introduction of renewable power production technologies into the existing power system. The real-time balance between production and consumption must, however, still be maintained at all times.
Unfortunately, this is becoming increasingly challenging due to the intrinsic variability of production technologies such as photovoltaics and wind turbines. In a Smart Grid system the balancing task will therefore be handled by mobilizing flexibility on the consumption side.
This Thesis assumes that the Smart Grid should be commercially based rather than funded by subsidies. Consequently the Smart Grid provides a business opportunity for so-called Virtual Power Plants. A Virtual Power Plant is an independent commercial operator, which provides Smart Grid capabilities to flexible consumers. This means that the Virtual Power Plant is the technical and commercial entity in charge of mobilization and control of flexible consumers. This Thesis addresses some of the challenges relating to the Smart Grid and Virtual Power Plant visions with special attention to flexibility, value creation and portfolio coordination.
The term flexibility is central to the Smart Grid discussion, but it is difficult to give a precise definition of flexibility. This Thesis therefore suggests the use of a simple taxonomy for modelling consumer flexibility. The taxonomy consists of three archetypal flexibility models, but it does not exhaust the flexibility term. It does however significantly sharpen the discussion of the flexibility concept and provides a categorization of flexible systems.
This Thesis also investigates what value can be created from the different types of flexibility by assuming that the Virtual Power Plant will generate profit by trading flexibility in electricity markets. Based on the taxonomy and a model of the Nordic electricity markets it is explored how differences between flexibility types affects profit margin. It is found that revenue potential depends strongly on the quality of flexibility.
Finally the subject of portfolio coordination is addressed, since a major challenge in developing the Smart Grid is that thousands or even millions of flexible consumers must be coordinated to operate in a sensible, interconnected manner. Due to the sheer size of the coordination problem, however, the computation time associated with coordination can be problematic. This issue is first investigated through analytical contributions on the circumstances under which portfolio coordination becomes computationally challenging.
Next, several options for finding optimal solutions of the coordination problem
are investigated, namely use of the software package CPLEX, Dynamic Programming and Dantzig-Wolfe Decomposition. Since non of these efforts scale to large problem instances the option of heuristic optimization is explored. Several methods are investigated and promising results are found both regarding computation time and solution quality.