Javascript must be enabled for the correct page display

Distributed Model Predictive Control in the DC Power Network

Apel, Maximiliaan (2019) Distributed Model Predictive Control in the DC Power Network. Research Project, Industrial Engineering and Management.

[img]
Preview
Text
mIEM_2019_APELMAS.pdf

Download (2MB) | Preview
[img] Text
toestemming.pdf
Restricted to Registered users only

Download (122kB)

Abstract

DC microgrids are increasingly utilized, because they have favourable characteristics over AC networks. The orientation of this research is to establish a distributed model predictive control based consensus algorithm for current sharing in DC microgrids. In centralized MPC control approaches for the DC microgrids the dynamics are coupled, while currents flow between the nodes. In large-scale networks centralized control is considered infeasible, nonscalable, too costly or too fragile due to the fact that every control action is executed by one controller. Therefore the establishment of a distributed MPC scheme for in the DC microgrid is essential. In this research the dynamics are decoupled by means of dual decomposition and subgradient iterations to reach a distributed formulation. Wherein each local controller, present at each node, solves its own subproblem, solely based on local information. Together they will arrive to the solution of the original problem, but without solving a centralized MPC problem. The proposed controllers also ensure that both the buck converter output voltages as well as the load voltages remain within acceptable bounds. The dual decomposition is performed on the physical system, as the DGUs share currents with each other. In previous research, dual decomposition was usually performed on a simple matrix, ensuring relatively steady convergence properties for the subgradient algorithm. In this research it is performed on a more complex coupling matrix representing the physical properties of the DC microgrid. As a result, it was discovered that the convergence behaviour of the subgradient algorithms was greatly affected by the configuration of the system in the optimization. Considering the effects of these parameters, it has been observed that MPC via dual decomposition and gradient iterations is a suitable design approach for reaching consensus in the DC microgrid in a distributed manner.

Item Type: Thesis (Research Project)
Supervisor name: Scherpen, J.M.A.
Degree programme: Industrial Engineering and Management
Thesis type: Research Project
Language: English
Date Deposited: 08 Apr 2019
Last Modified: 09 Apr 2019 09:39
URI: https://fse.studenttheses.ub.rug.nl/id/eprint/19340

Actions (login required)

View Item View Item