LAUSR

Abstract Today’s applications, like smart grids or complex HVAC systems, get more and more complex. This leads to enormous communication when using centralized controllers. Controller networks can reduce this effort by splitting the control task to smaller nodes, which do not need the full system information. The complex behavior of such systems can be modeled by multilinear time-invariant (MTI) systems in situations where linear models may fail. In this paper a method is proposed that determines a decentralized state feedback controller structure, that has an acceptable performance loss compared to the centralized case by multiple linearizations of the MTI plant. During operation of the controller the feedback gain is adapted to the current operating point. Application to a complex heating system example shows promising results.