Learn-by-doing and carbon dioxide abatement

There are inherent difficulties in solving learn-by-doing (LBD) models. Basic to such models is the idea that the accumulation of experience leads to a lowering of costs.
This paper is intended to explore some of the algorithmic issues in LBD modeling for carbon dioxide abatement. When using a standard algorithm for nonlinear programming, there is no guarantee that a local LBD optimum will also be a global optimum. Fortunately, despite the absence of guarantees, there is a good chance that one of the standard algorithms will produce a global optimum for models of this type - particularly if there is an artful selection of the starting point or of the terminal conditions. Moreover, there is a new procedure named BARON. In the case of small models, a global optimum can be recognized and guaranteed through BARON.
Eventually, it should be possible for BARON or a similar approach to be extended to large-scale LBD models for climate change. Meanwhile, in order to check for local optima, the most practical course may be to employ several different starting points and terminal conditions.
Keywords: Learn-by-doing model; carbon dioxide abatement; BARON

VIEW CONTENT