Optimization methods in multilevel systems: A methodological survey (Q1072972): Difference between revisions

A critical review of currently prevalent optimization methods in multilevel systems is attempted. Special emphasis is laid on such methods which offer maximum scope for real-world applications. The paper begins by introducing the fundamental notion of ''coordination'' of a multilevel system. Two ''coordination principles'', viz. (i) Interaction prediction (INPRE) and (ii) Interaction balance (INBAL), are singled out for extended discussion. The main text of the paper focusses on the optimization of dynamic multilevel systems, characterized by the LQ format (linear system equations and quadratic performance criterion). For such systems, the following optimization methods are considered. (a) (i) Feasible decomposition method and (ii) penalty function coordination method. Both these methods are INPRE-based; (b) Non-feasible decomposition method, which is an INBAL-based method. (c) (i) Combined coordination method and (ii) Takahara decomposition method. These two methods are hybrid methods, based on a combination of INPRE and INBAL. Computational problems associated with these methods are discussed in a separate section. The paper concludes with a brief overview of certain areas where multilevel techniques have found substantial applications such as power systems, water resources, urban traffic systems, river pollution control etc.