OPTIMAL CONTROL METHODS FOR MACROECONOMIC GROWTH AND ENERGY MODEL: AN EVALUATION OF COMPUTATIONAL ALGORITHMS.

A description is given of a new General Purpose System Simulation and Optimal Control (GPSSOC) computer program. The software called GPSSOC can solve nonlinear time-varying optimal control systems of arbitrary dimensions in explicit or implicit form. GPSSOC includes nonlinear programming algorithms as well as optimization algorithms based on optimal control theory. Numerical solution methods based upon differential dynamic programming seem to be more efficient and robust than the static unconstrained optimization methods from mathematical programming. The user supplies a user sub-program in Fortran to define the model and the cost functional and an input data file. GPSSOC then searches for the optimal open-loop and/or local feedback control with respect to the given cost functional. The computational results show the feasibility of using GPSSOC to numerically optimize the Long-Term Macroeconomic and Energy Model (LTMEM), which is also developed at M. I. T. Energy Laboratory. A specific important application is illustrated dealing with the evaluation of achieving minimum cost of Fast Breeder Reactors development subject to the limited available flow of investment over a long period of time.