Mitigation strategies and energy technology learning: An assessment with the POLES model

Highlights

• Apparent learning may be slower in mitigation scenarios with accelerated deployment.

• Solar benefits more from major technological breakthroughs than wind.

• Reductions in R&D budgets have significant impacts on long term technology costs.

• Ambitious stabilization targets can be met with limited cost increases.

• Deployment of renewables is key for stabilizing long-term electricity prices.

Abstract

This paper explores various dimensions of the learning process for low-carbon technologies under different mitigation scenarios. It uses the POLES model, which addresses learning as an endogenous phenomenon with learning curves, and a set of scenarios developed as part of the AMPERE project. It represents an analytical effort to understand the learning patterns of energy technologies in various contexts and tries to disentangle the different dimensions of the relation between these patterns and the deployment process. One result is, surprisingly, that apparent learning may be slower in mitigation scenarios with accelerated technology deployment when using two-factor learning curves. Second, the R&D analysis clearly shows that reductions in R&D budgets have significant impacts on long term technology costs. Third, solar technology which is more constrained by floor costs in the model benefits more from major technological breakthroughs than wind energy. Finally, ambitious stabilization targets can be met with limited cost increases in the electricity sector, thanks to the impact of learning effects on the improvement in technology costs and performances.

Introduction

The process of technological change is identified in the literature as a highly complex one, which involves multiple stages and categories of actors (researchers, developers, technicians, device producers, users, policy makers, etc.) [1], [2]. However in any modelling exercise associated with the use of Integrated Assessment Models for the analysis of greenhouse-gas mitigation strategies, simplifications must be made. In this perspective of applied modelling exercises, we have predominantly focused on two fundamental mechanisms by which technologies evolve. They are typically referred to as “learning-by-doing” [3] and “learning-by-searching” [4]. This reflects the fact that, barring approaches that consider technological change as purely exogenous, one must consider that both the proper deployment of a given technology and R&D effort dedicated to this technology have a dynamic impact on costs and performance: in the process of technology improvement over time, both doing and searching matter. These two aspects of the learning process have been integrated through the development, for modelling purposes, of two-factor learning curves [4]. They have been increasingly introduced into technology assessments [5], [6], and energy and Integrated Assessment Models as a means of representing endogenous technical change [7]. The adoption of two-factor learning curves (TFLCs) in energy system modelling has been an important development. It is however still debated as it introduces new uncertainties, raises data problems and fails to reflect important variables such as efficiency of research, network effects, spillovers, etc. [8], [9], [10], [11], [12]. Moreover, recent studies point out the impact of other factors than pure learning and research activities on cost reductions, commodity prices or scale effects, for example [13]. There is a clear need for continuous research on these topics [14]. Still, the TFLC may be relevant as it provides some insights into the influence of R&D budgets, a major variable for public policy.

As it focuses on an explicit representation of the energy system and its key technologies, the POLES model was one of the first models to implement two-factor learning curves (TFLCs) [4], [7]. The technology time-series database associated with the POLES model – the TECHPOL database – has benefitted from extensive research to develop both learning by doing and learning by searching for technologies in key sectors (power and hydrogen production, transport, energy-intensive industries, building), gathering information from the literature and energy experts throughout the EU. The model thus provides insights into the technological dimension of long-term energy-system transitions, subject to various conditions dependent on climate policies, energy and carbon prices, and technology performance trajectories or technology availability.

Many low-carbon technologies, particularly renewables, are relatively new and are thus characterized by considerable potential for learning. Several cross-cutting papers presenting an overview of this special issue address the topic of deploying low-carbon technologies and transforming the energy system in the framework of AMPERE scenarios [16], [17]. In this paper we analyse results from simulations carried out with the POLES model in this framework, in order to explore the relationships between the learning rates for low-carbon technologies and various mitigation pathways. AMPERE scenarios have two dimensions: a technological dimension with various technology portfolios and assumptions about technological change; and a climate-policy dimension (GHG targets). To reduce the complexity of the analysis we have chosen to focus on nine scenarios, out of 41, as detailed in Table 1.

In the paper we address important issues related to the relationship between technological learning of new energy technologies and climate policies: how do climate policies impact learning in wind and solar technologies? How do delays in policy implementation impact learning and, thus on the cost of technologies? What do accelerated or reduced learning rates imply for the long-term deployment of low-carbon technologies in various scenarios?

Section 2 provides some background on learning rates and detailed explanation of how TFLCs are implemented in the POLES model. In Section 3, we analyse the impact of climate policies on learning, based on a comparison between the Base-FullTech-OPT and 450-FullTech-OPT AMPERE scenarios, and the influence of delayed action scenarios (450-FullTech-HST, LST, and OPT) on technical change. Section 4 focuses only on the scenarios in which factors exogenous to the wind and solar industry impact the deployment of wind and solar energy, like LowEI where intensity is low or other technologies (NucOff and LimBio) are limited. Section 5 focuses on the sensitivity of results to changes in TFLC parameters, including sensitivities with different elasticities, technology floor-costs, and assumptions about cumulative R&D. The impacts of learning on mitigation costs and LCOE are examined in Section 6. Finally, Section 7 provides a summary of the findings and conclusions for further research.