Decomposition of a Large-Scale Energy Model

The modeling of energy systems generally involves the solution of very large scale linear programming problems which include descriptions of the energy transformation chains. The scale of the problem and the variety of processes considered are such that the model should, ideally, be composed of submodels, each developed by experts in the appropriate field. However, this is not usually possible for a number of reasons. One of the most important of these is the absence of efficient methods for linking or making consistent the various submodels, which may be based on different time-scales and different degrees of aggregation, and which may involve different policy variables and economic agents. Another reason for the infrequent use of this modular approach may lie in the many reported failures of attempts to implement decomposition approaches in large-scale optimization systems. These considerations, combined with the practical necessity of squeezing a large-scale model into a small computer, encouraged members of the IIASA Energy Systems Group and the Systems and Decisions Sciences Program to work together on the decomposition of the IIASA energy supply model MESSAGE II. The decomposition algorithms developed as part of research on nondifferential optimization played an important role in the study. The results suggest a method of constructing an integrated system of energy models that could provide a detailed representation of the energy supply system itself and its interaction with the major energy-intensive sectors. A thorough investigation of this interaction, in terms of the energy flows represented by the linking variables, could be valuable in determining an interally consistent national energy policy