In this paper we present the theoretical foundations for one of the me
thods used to achieve convergence in the National Energy Modeling Syst
em (NEMS). NEMS is a large model with several component models that ar
e built and operated by different branches in the organization and is
an example of a system without a hierarchical structure that cannot be
solved by traditional equation solving methods. Some of the component
models use linear programs to construct supply and demand curves. The
discontinuities that result lead to oscillations in the standard rela
xation algorithms. We explain where the convergence problems lie and h
ow the convergence theory with step functions links to the convergence
theory with continuous functions. To achieve convergence within the e
ntire system, a set of ad hoc techniques were developed to implement a
decomposition strategy that allows the individual models to be run se
parately. We present the theoretical justification for one of them her
e. The technique presented here has the potential to allow an organiza
tion to use operational models for planning without resorting to aggre
gation. It also facilitates decentralized computing over Internet.