A FRAMEWORK FOR THE ANALYSIS OF TRANSITION PATH SIMULATION MODELS

DAVID PRICE YANCEY, Purdue University

Abstract

This thesis is concerned with the development of a framework for constructing and analyzing transition path models. A transition path is the economic, social, technological, and environmental behavior that is exhibited by a system over a given period of time. Typically, various sectors of real life systems are linked such that the behavior of, say, the economic sector influences the behavior of the environmental sector. Transition models attempt to capture this sector interaction where appropriate and report data from the relevant model sectors in ways that describe the transition path for desired analysis. The guiding principles employed in designing the framework for transition path modeling included the following. (1) The framework must be designed such that all stages throughout model construction, testing, and analysis are supported. (2) Models are to be constructed and analyzed using high level language statements. (3) The user must retain the capability to make intricate modeling specifications. (4) A variety of modeling methodologies must be included in the framework to support the variety of approaches commonly used in capturing technological, social, environmental, or other phenomenon relating to transition path models. (5) The transition analysis framework must be open ended to allow additional user selected algorithms to be merged into the model structure. (6) The framework must provide a package that promotes the integration of algorithms such that submodels (algorithms) have a generalized capability to dominate and direct the operation of other submodels (algorithms). (7) The significance of this research is the framework, not individual algorithms; thus, effort is dedicated to the integration of existing algorithms rather than developing new algorithms. Based upon the above principals, a conceptual design was presented. Subsequent functional design specifications were made and software was selected from the fields of simulation, data base management, optimization, and econometrics to manifest the design. A prototype of the Transition Path Analysis Language was synthesized from selected computer programs to simulate transition path models. This simulation language has the capability to employ linear and nonlinear programming, polynomial curve fitting, continuous and discrete simulation, stochastic and deterministic modeling, equilibration of submodels, dynamic editing and restructuring of data within models, and statistics collection and reporting. An attached data base management facility provides a long-term data storage facility for both collected and simulation generated information. Capabilities are provided such that ad hoc analyses may be performed and data may be restructured for model fitting experiments. Four classes of models were described and programmed in TPAL: a continuous model of automotive fuel consumption; a nonlinear optimization of least squares to best fit parameters to a model; a network of R&D events that lead to the mass production of Stirling engines; and an energy process model, formulated as a linear program, with an economic model, formulated as an input-output model. Embellishments to these models were programmed to illustrate various modeling features. It was concluded based upon this research that it is feasible to build transition models using the transition path modeling framework. In addition, TPAL satisfied the design objectives of this research. It was concluded that TPAL is a useful tool for transition modelers, users, and policy analysts. The modeling framework and language developed in this research for transition path analysis contributes significantly by its design, construction, and philosophical approach.

Degree

Ph.D.

Subject Area

Industrial engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS