Being technology a fundamental driving factor of the evolution of energy systems, it is essential to study the basic mechanisms of technological change and its role in achieving more efficient, productive and clean energy systems. Understanding its dynamics constitutes a guide for policy formulation and decision-making and the conception of effective intervening instruments. Technology development does not occur as an autonomous independent process, but evolves from a number of endogenous interactions within the social system. Technologies evolve and improve only if experience with them is possible. Cumulative learning, both in R&D activities and the marketplace constitutes one of the basic mechanisms in the emergence and replacement of technological regimes.

Thus, efforts must be devoted to improve our analytical tools and decision-support frameworks concerning the treatment given to the technological variable. Despite its undeniable importance, several technological factors have been traditionally addressed in an oversimplified way in energy optimisation models, without recognising the cumulative and gradual nature of technological change and the important role that learning processes play in achieving cost/performance improvements in a given technology or clusters of them.

This dissertation addresses the endogenisation of some aspects of technological change in energy systems optimisation models. Here, learning curves, an empirically observed manifestation of the cumulative technological learning processes, are endogenised in two energy optimisation models: MARKAL, a widely used bottom-up model developed by the ETSAP programme of the IEA and ERIS, a model prototype developed together with other partners during the EC-TEEM project, for assessing different concepts and approaches (TEEM 1997, 1999). The incorporation of the curves provides the models with a mechanism to represent path-dependent and self-reinforcing phenomena intervening in shaping the technological trajectories of the system.

The methodological approach is described, illustrative analyses presented and insights derived from the analyses outlined. The incorporation of learning curves results in a non-convex non-linear mathematical program. Here, using Mixed Integer Programming techniques, a linear approximation to such problem is applied. When endogenous learning is present, model outcomes are significantly different than those obtained when applying static or exogenous cost trends, common in traditional approaches. New, innovative technologies, hardly considered by the standard models, are introduced to the solution. Up-front investments in initially expensive, but promising, technologies allow the necessary accumulation of experience to render them cost-effective.

The learning rates of the technologies are, however, uncertain. In order to capture this aspect, a two-stage stochastic programming approach is applied. With uncertain learning rates, a more prudent intermediate path of installations for learning technologies is followed, and a more diversified technological choice takes place. However, even under uncertainty, technological learning in emerging technologies continues to be an important hedging mechanism to prepare for future actions. Uncertainty in many other factors also plays a relevant role in the stimulation or delay of learning. For instance, when uncertainty in emission reduction commitments is considered, the results point also in the direction of undertaking early action as a preparation for future contingencies. Early investments stimulating technological learning prove beneficial in terms of both lower costs and emissions in the long run. Increasing returns, associated to the effects of learning, and technological uncertainty emerge as interacting core mechanisms of the technological change process.

The spatial aspects of the technological learning process are also highlighted. Learning is a network phenomenon and the spatial configuration of the learning network is of considerable relevance in the scope and effectiveness of the process. Thus, the spatial scale of learning plays an important role in the global competitiveness of emerging technologies and, therefore, its variation influences significantly model outcomes. The mutual interactions between different scales of learning and several modalities of emissions trading are examined and the importance of their combined effects on the technology choice underlined. The results reveal the significant potential of international co-operation in fostering the diffusion of more efficient and clean energy technologies and the necessity of deepening the understanding of spill-over effects in the learning process.

In addition, recognising that besides market experience R&D efforts also constitute an important factor for technological progress, a simplified analysis is presented regarding the representation of this factor as part of the technological learning mechanism. A so-called two-factor learning curve is applied, where both capacity deployment and R&D expenditures contribute to the accumulation of knowledge. Although the exercise is preliminary and the formulation still depends on a meaningful statistical estimation to be supported, the analysis shows the necessity of incorporating such a factor as one of the decision variables of the models as to gain insights about the optimal configuration of R&D portfolios and continuing work on the poorly understood role and effectiveness of R&D in technological progress.

The results obtained using this modelling approach provide some important policy insights. Early investments in R&D, demonstration projects and deployment in niche markets (the so-called ERD3 strategy) of sustainable technologies, are required in order to ensure that they move along their learning curves and achieve long run competitiveness. New technologies will become competitive only if experience with them is possible. Their successful introduction requires then the promotion of innovation and learning at multiple technological, social and institutional levels. It is necessary to advance further in the endogenisation of technological change into energy planning models. The treatment given to technology dynamics affects our understanding of a number of issues concerning the future structure of global energy systems and their environmental impacts (e.g. contribution to climate change). An adequate framework is necessary to gain insights about the underlying forces that drive this evolution.