Decentralized Model Predictive Controlfor smart grid applications
Student paper other, 20 credits / 30 HE creditsStudent thesis
This thesis focuses on a model predictive control scheme which allows organizing the power production of power plants in a decentralized fashion. That way a less computationally demanding control scheme can be applied to control the frequency at each power plant without a governing, high level controller for the whole power grid. The main contribution was to develop a communication scheme between power plants that can be applied in a competitive environment like the energy market. Most established schemes for decentralized power production require giving away a mathematical representation of a plant to all its neighbors. However, the developed control scheme only requires informing the neighboring power plants the expected future evolution of the power networks voltage angle. This is information which is more efficiently communicated by power plant operators. Additionally, this simplification does not only yield a notable reduction in communicated data but also reduces the computational complexity of the control problem for a single power plant. The aforementioned control scheme was applied to a network consisting of several different plants for each of which a model was developed. The modeled plants range from conventionally generated plants like hydro-, gas- or wind power plants to more modern converter coupled plants like photovoltaic installations. The plants were modeled such that energy buffers - in the form of aggregated Electric Vehicle Batteries - can be taken into account. For the power plants and the energy transfer between them, linear time-invariant models were augmented with linear matrix inequalities. These represent physical bounds which the model has to regard in order to have a realistic system evolution ie, maximal power line capacities or limiting power plant production capabilities. Proofs are given which indicate necessary properties of the developed algorithm to ensure nominal or robust stability. Simulations were carried out which verified the conditions obtained from the proofs. Also by simulation, the obtained control scheme was compared with a centralized approach, amongst others. Considering the developments towards a Smart Grid one can say that a power production which is organized in a decentralized way reduces the computational effort greatly with a tolerable loss of performance. This statement is backed up with results from the simulations. The findings also indicate that the addition of buffers is very beneficial regarding disturbance rejection in the power grid.
Place, publisher, year, edition, pages
2013. , 71 p.
IdentifiersURN: urn:nbn:se:kth:diva-133580OAI: oai:DiVA.org:kth-133580DiVA: diva2:662289