A Learning-by-Doing Energy Model on Dynamic Programming
Abstract
The concept of learning by doing (LBD) rests on the assumption that the more we do something, the more efficient we become at it. the inclusion of this phenomenon in our models results in a non-convex formulation and the possibility of multiple local optimal solutions. In this paper, we present a dynamic programming formulation of a model with learning-by-doing. The main advantage of this formulation is the guarantee of a global optimal solution, as conventional nonlinear solvers generally return local optimal solutions with no guarantee of global optimality. We also present two nonlinear extensions to the model that are not easily solved with some other heuristics. We conclude by running the model based on three carbon tax cases and a discussion of the results.