Servicio de Información Comunitario sobre Investigación y Desarrollo - CORDIS

Two Factor Learning Curve Specification and Estimation

The specification and estimation of the two factor learning curve relations for use in the large models as well as in PROMETHEUS is the backbone for meaningful policy analysis in the context of the SAPIENTIA project as they constitute the main vehicle and first step through which R&D actions translate into impacts. Some specific properties have been sought for the specification of the two factor learning curve formulation: the TFLCs should incorporate both learning-by-doing and learning-by-research (which is crucial in order to be able to perform the R&D policy exercises), endogenise as much of the technical progress as possible, constrain to technical possibilities as they emerge from perspective analysis, include “Clustering” as fully as available information allows, take carefully into account initial conditions regarding cumulative R&D and equipment stock and capture as much of the above with as few parameters as possible (to render their model incorporation practically feasible). A general algebraic specification has been derived and considerable effort has been devoted to standardise the formulation as much as possible. Some exceptions however were deemed necessary due to specifities of the technologies. As a result, the technologies are classified into five categories, the Cluster Matrix Technologies, for which a cluster matrix is supplied and the general algebraic formulation can be applied, the Stand Alone Technologies, which are orthogonal technologies and do not need to consider clustering, the Perfect Clustering Technologies, the technical and economic characteristics of which are directly related to the corresponding characteristics of other technologies in the same cluster (for example the wind offshore and wind onshore technologies), the On-board storage technologies, (a sub-category of the perfect clustering technologies) that are shared by different types of vehicles and finally the Fuel Cell technologies. Particular attention was given to the estimation of the TFLCs so as to ensure that apart from statistical fit they also displayed sufficient robustness for use in the wide variety of models and especially that they performed credibly in view of the R&D policy analysis. For the estimation historical time-series for R&D, equipment stock, capital costs and projections of technical and economic characteristics of technologies were used (as derived from the TECHPOL database), projections of installed equipment from the provisional POLES Baseline, projections of public and private R&D by technology elaborated by ICCS/NTUA, and finally cluster information from MARKAL. Clustering of technologies has been incorporated through learning by doing. The learning parameters were estimated by applying Maximum likelihood estimation over the historical period; yet this was by no means the only estimation criterion. All properties sought in the TFLCs specification figured among the objectives of the estimation. In addition, simultaneous equations estimation has also been applied (along clusters) in order to improve estimates and obtain appropriate co-variances of learning parameters. In all, technology dynamics have been estimated for a total of 51 technological options covering power generation, CO2 capture and sequestration, Hydrogen-related technologies, conventional and non-conventional vehicles and Fuel Cells. Learning parameters were estimated for capital costs, fixed Operation and Maintenance costs, variable Operation and Maintenance costs, efficiencies and CO2 capture rates.

Reported by

Institute of Communication and Computer Systems of National Technical University of Athens
9, Iroon Polytechniou Street
157 73 Zografou
Síganos en: RSS Facebook Twitter YouTube Gestionado por la Oficina de Publicaciones de la UE Arriba