Endogenous Technological Learning Research Articles

Hindcasting to inform the development of bottom-up electricity system models: The cases of endogenous demand and technology learning

Bottom-up, technology-rich electricity system models are commonly used to generate scenarios for policy support at a national or global level. In literature, previous hindcasting studies (also called retrospective modeling or ex-post modeling) evaluated existing models rather than sought to inform model development from the beginning. In this study, we present a hindcasting exercise with D-EXPANSE model for national electricity systems in 31 European countries over the 1990–2019 period. We develop several model versions with or without elastic electricity demand and with or without endogenous technology learning, and use hindcasting to choose the most accurate configuration of the bottom-up model. The hindcasting results show that a model with endogenous elastic demand can capture well the real-world evolution of electricity demand, if elasticity factor is chosen well and if the countries did not undergo severe structural changes. Endogenous technology learning, however, increases the uptake of new emerging technologies in cost-optimal scenarios, but still cannot fully capture the real-world dynamics and at times even introduces further inaccuracies.

Coping with increasing tides: Evolving agglomeration dynamics and technological change under exacerbating hazards

By 2050 about 70% of the world’s population is expected to live in cities. Cities offer spatial economic advantages that boost agglomeration forces and innovation, fostering further concentration of economic activities. For historic reasons urban centers cluster along coasts, which are prone to climate-induced flooding and sea level rise. To explore trade-offs between agglomeration economies and hazards increasing with climate change, we develop an evolutionary agent-based model with heterogeneous boundedly-rational agents who learn and adapt to a changing environment. The model combines migration decision of both households and firms between safe Inland and hazard-prone Coastal regions with endogenous technological learning and economic growth. Flood damages affect Coastal firms hitting their labour productivity, capital stock and inventories. We find that the model is able to replicate a rich set of micro- and macro-empirical regularities concerning economic and spatial dynamics. Without climate-induced shocks, the model shows how lower transport costs favour the Coastal region fueling the self-reinforcing and path-dependent agglomeration processes. We then introduce five scenarios of floods characterized by different frequency and severity to study the complex interplay of hazards with agglomeration patterns affecting the performance of the overall economy. We find that when shocks are mild or infrequent, they negatively affect the economic performance of the economy. If strong flood hazards hit frequently the Coastal region before agglomeration forces trigger high levels of the waterfront urbanization, firms and households can timely adapt and migrate landwards, thus averting the adverse impacts of climate shocks on the whole economy. Conversely, in the presence of climate tipping points where the frequency and magnitude of flood hazards abruptly intensifies, we find that economic activities remain trapped in the hazard-prone region, generating lock-ins and leading to a harsh downturn of the overall economy.

Capacity planning with uncertain endogenous technology learning

Optimal capacity expansion requires complex decision-making, often influenced by technology learning, which represents the reduction in expansion cost due to factors such as cumulative installed capacity. However, having perfect foresight over the technology cost reduction is highly unlikely. In this work, we develop a multistage stochastic programming framework to model capacity planning problems with endogenous uncertainty in technology learning. To assess the benefit of the proposed framework over deterministic optimization, we apply a shrinking-horizon approach to compute the value of stochastic solution. Further, a decomposition scheme based on column generation is developed to solve large instances. Results from our computational experiments indicate substantial potential cost savings and the effectiveness of the proposed decomposition algorithm in solving instances with large numbers of scenarios. Lastly, a power capacity planning case study is presented, highlighting the stochastic optimization’s ability to anticipate significantly different expansion and production decisions in low- and high-learning scenarios.

Effects of learning curve models on onshore wind and solar PV cost developments in the USA

Technological innovation planning for developing and deploying clean energy technologies plays a key role in reducing greenhouse gas emissions and transition to a low-carbon future. Learning curve theory has been adopted as a common framework for exploring the relationship between endogenous technological learning and technology cost developments. The aim of this article is to analyse the effects of selecting different learning curve approaches (i.e. model formulations) to describe energy technology cost changes over time. Experience and knowledge stock are chosen as the sources of learning to be considered. A new definition of experience was developed to account for the interaction between global and local experience. The new definition of experience also accounts for learning sub-processes (i.e. learning-by-doing, learning-by-using, and experience spillovers) to estimate total experience gained through technology deployment. An integrative model is developed for estimating the effects of learning-by-deploying and learning-by-researching on cost developments for onshore wind and solar PV in the USA. Publicly available data from government departments and organisations were utilised. It was found that technology cost developments are better explained when: (1) experience is defined as a function of global and local experience; (2) knowledge stock is also considered in the model formulation; and (3) technological processes affect only a fraction of the total capital cost. The findings suggested that the application of learning rates for model-based energy planning is context-dependent and how technological factors are explicitly defined may have significantly different policy implications (i.e. different technology costs predictions based on alternative model formulations).

An expert survey to assess the current status and future challenges of energy system analysis

Decision support systems like computer-aided energy system analysis (ESA) are considered one of the main pillars for developing sustainable and reliable energy strategies. Although today's diverse tools can already support decision-makers in a variety of research questions, further developments are still necessary. Intending to identify opportunities and challenges in the field, we classify modelling topics into modelling capabilities (32), methodologies (15), implementation issues (15) and management issues (7) from an extensive literature review. Based on a quantitative expert survey of energy system modellers (N = 61) mainly working with simulation and optimisation models, the Status of Development and the Complexity of Realisation of those modelling topics are assessed. While the rated items are considered to be more complex than actually represented, no significant outliers are determinable, showing that there is no consensus about particular aspects of ESA that are lacking development. Nevertheless, a classification of the items in terms of a specially defined modelling strategy matrix identifies capabilities like land-use planning patterns, equity and distributional effects and endogenous technological learning as “low hanging fruits” for enhancement, as well as a large number of complex topics that are already well implemented. The remaining “tough nuts” regarding modelling capabilities include non-energy sector and social behaviour interaction effects. In general, the optimisation and simulation models differ in their respective strengths, justifying the existence of both. While methods were generally rated as quite well developed, combinatorial optimisation approaches, as well as machine learning, are identified as important research methods to be developed further for ESA.

Decarbonizing existing coal-fired power stations considering endogenous technology learning: A Turkish case study

Despite environmental threats, coal is expected to remain a remarkable energy source as the share of intermittent renewable sources in the energy mix grows; hence, managing pollutants such as carbon dioxide (CO2) is critical while pushing for clean alternatives. Carbon capture, utilization, and storage (CCUS) is a set of methods that removes CO2 from emissions and stores it safely for the long-term. As it stands now, CCUS is prohibitively expensive; however, learning-by-doing is a recognized phenomenon that assists novel technologies to become more affordable as employed continuously. Thus, for accurate long-term planning, this study investigates the impact of early adopters on CCUS market diffusion considering endogenous technology learning (ETL). A mixed-integer nonlinear programming model (MINLP) is developed which determines the optimal capacity and deployment time of post-combustion carbon capture units and the optimal source–sink matching in the CO2 supply chain, taking into account techno-economic constraints and regulations. The MINLP model is transformed into a series of mixed-integer linear programming models using a Stackelberg game approach and solved with a diagonalization method. Equilibrium solutions of the proposed model were obtained for independent and coordinated actions of players. The proposed model is applied to the Turkish energy market with multiple CO2 sources and sinks, where a cap-and-trade program is presumed to be in place. While the study demonstrates the interplay between CCUS and ETL in Turkey, the proposed model is universal, and it can be applied to other countries if the inputs are available.

Exploring China's carbon emissions peak for different carbon tax scenarios

Under the United Nations Framework Convention on Climate Change, China has made a commitment to peak CO2 emissions around 2030. China is still interested in fulfilling its commitment to address climate change. In this study, a diffusion model of energy technology based on endogenous technology learning under bounded rationality is developed to explore the possible impacts of different carbon tax conditions on the diffusion of energy technologies in China. It is concluded that the substitution of energy technology is a long-term and slow process. In the case of a high-carbon tax, transition technologies and new technologies will appear earlier and replace the original technology (6–8 years earlier than in the case of a lower carbon tax). The transition technology will become dominant only for a short period of time (around 2030) but will be quickly replaced by new technology. The traditional technology will cease to exist in the last thirty years of this century in all carbon tax scenarios. In the low-carbon tax or high-carbon tax scenarios, carbon emissions will peak in 2030 and then decline significantly. The peak in carbon emissions is lower under the high-carbon tax scenario than the low-carbon tax scenario, representing a 28% decrease.

The role of technology diffusion in a decarbonizing world to limit global warming to well below 2 °C: An assessment with application of Global TIMES model

Low-carbon power generation technologies such as wind, solar and carbon capture and storage are expected to play major roles in a decarbonized world. However, currently high cost may weaken the competitiveness of these technologies. One important cost reduction mechanism is the “learning by doing”, through which cumulative deployment results in technology costs decline. In this paper, a 14-region global energy system model (Global TIMES model) is applied to assess the impacts of technology diffusion on power generation portfolio and CO2 emission paths out to the year 2050. This analysis introduces three different technology learning approaches, namely standard endogenous learning, multiregional learning and multi-cluster learning. Four types of low-carbon power generation technologies (wind, solar, coal-fired and gas-fired CCS) undergo endogenous technology learning. The modelling results show that: (1) technology diffusion can effectively reduce the long-term abatement costs and the welfare losses caused by carbon emission mitigation; (2) from the perspective of global optimization, developed countries should take the lead in low-carbon technologies’ deployment; and (3) the establishment of an effective mechanism for technology diffusion across boundaries can enhance the capability and willingness of developing countries to cut down their CO2 emission.

A global and local endogenous experience curve model for projecting future uptake and cost of electricity generation technologies

A global and local learning model (GALLM) has been developed to project the cost and global uptake of different electricity generation technologies to the year 2050. This model features three regions, endogenous technological learning within and across those regions, various government policies to facilitate technological learning and a penalty constraint which is used to mimic the effect market forces play on the capital cost of electricity generation technologies. This constraint has been added as market forces have been a strong factor in technology pricing in recent years. Global, regional and component experience curves have been developed for some technologies. The model, with the inclusion of these features, projects a diverse range of technologies contributing to global electricity generation under a carbon price scenario. The penalty constraint leads to gradual and continual installations of technologies and because the constraint provides a disincentive to install too much of a technology, it reduces the impact of uncertainty in the learning rate. Alternative forms of the penalty constraint were tested for their suitability; it was found that, with a zero and lower-cost version of the constraint, photovoltaics are installed in a boom-and-bust cycle, which is not supported by past experience. When the constraint is set at a high level, there are fewer installations.

Western China energy development and west to east energy transfer: Application of the Western China Sustainable Energy Development Model

China is striving for coordinated regional economic development and to solve the energy shortage in eastern China through a western China development plan with one focus being energy development and west to east energy transfer. This paper describes Western China Sustainable Energy Development Model (WSED) to evaluate various energy development scenarios for western China. The model includes a Western China MARKAL model, a Computable General Equilibrium Model for Western China (WCGE), and an Energy Service Demand Projection Model (ESDP). The ESDP provides energy service demand projections for the Western China MARKAL model, while the WCGE provides macroeconomic inputs for the ESDP and analyzes the impact of different energy development scenarios on western China economy. A reference scenario and several different west to east energy transfer scenarios with and without consideration of the water constraints and the endogenous technology learning are presented. The modeling describes the energy consumption, carbon emissions, water consumption, energy investment cost, and the impact on western China GDP of the different scenarios through the year 2050. These results have implications on sustainable energy development policies and sustainable west to east energy transfer strategies.

Modeling technological learning and its application for clean coal technologies in Japan

Estimating technological progress of emerging technologies such as renewables and clean coal technologies becomes important for designing low carbon energy systems in future and drawing effective energy policies. Learning curve is an analytical approach for describing the decline rate of cost and production caused by technological progress as well as learning. In the study, a bottom-up energy-economic model including an endogenous technological learning function has been designed. The model deals with technological learning in energy conversion technologies and its spillover effect. It is applied as a feasibility study of clean coal technologies such as IGCC (Integrated Coal Gasification Combined Cycle) and IGFC (Integrated Coal Gasification Fuel Cell System) in Japan. As the results of analysis, it is found that technological progress by learning has a positive impact on the penetration of clean coal technologies in the electricity market, and the learning model has a potential for assessing upcoming technologies in future.

Global technology learning and national policy—An incentive scheme for governments to assume the high cost of early deployment exemplified by Norway

In this paper it is argued that technology learning may be both a barrier and an incentive for technology change in the national energy system. The possibility to realize an ambitious global emission reduction scenario is enhanced by coordinated action between countries in national policy implementation. An indicator for coordinated action is suggested. Targeted measures to increase deployment of nascent energy technologies and increasing energy efficiency in a small open economy like Norway are examined. The measures are evaluated against a set of baselines with different levels of spillover of technology learning from the global market. It is found that implementation of technology subsidies increase the national contribution to early deployment independent of the level of spillover. In a special case with no spillover for offshore floating wind power and endogenous technology learning substantial subsidy or a learning rate of 20% is required. Combining the high learning rate and a national subsidy increases the contribution to early deployment. Enhanced building code on the other hand may reduce Norway’s contribution to early deployment, and thus the realization of a global emission reduction scenario, unless sufficient electricity export capacity is assured.

Uncertainty in the learning rates of energy technologies: An experiment in a global multi-regional energy system model

The diffusion of promising energy technologies in the market depends on their future energy production–cost development. When analyzing these technologies in an integrated assessment model using endogenous technological learning, the uncertainty in the assumed learning rates (LRs) plays a crucial role in the production–cost development and model outcomes. This study examines the uncertainty in LRs of some energy technologies under endogenous global learning implementation and presents a floor-cost modeling procedure to systematically regulate the uncertainty in LRs of energy technologies. The article narrates the difficulties of data assimilation, as compatible with mixed integer programming segmentations, and comprehensively presents the causes of uncertainty in LRs. This work is executed using a multi-regional and long-horizon energy system model based on “TIMES” framework. All regions receive an economic advantage to learn in a common domain, and resource-ample regions obtain a marginal advantage for better exploitation of the learning technologies, due to a lower supply-side fuel-cost development. The lowest learning investment associated with the maximum LR mobilizes more deployment of the learning technologies. The uncertainty in LRs has an impact on the diffusion of energy technologies tested, and therefore this study scrutinizes the role of policy support for some of the technologies investigated.

CO 2, GDP and RET: An aggregate economic equilibrium analysis for Turkey

There is a worldwide interest in renewable electricity technologies (RETs) due to growing concerns about global warming and climate change. As an EU candidate country whose energy demand increases exponentially, Turkey inevitably shares this common interest on RET. This study, using an aggregate economic equilibrium model, explores the economic costs of different policy measures to mitigate CO 2 emissions in Turkey. The model combines energy demands, capital requirements and labor inputs at a constant elasticity of substitution under an economy-wide nested production function. Growing energy demand, triggered by economic growth, is met by increased supply and initiates new capacity additions. Investment into RET is encouraged via the incorporation of (a) endogenous technological learning through which the RET cost declines as a function of cumulative capacity, and (b) a willingness to pay (WTP) function which imposes the WTP of consumers as a lower bound on RET installation. The WTP equation is obtained as a function of consumer income categories, based on data gathered from a pilot survey in which the contingent valuation methodology was employed. The impacts of various emission reduction scenarios on GDP growth and RET diffusion are explored. As expected, RET penetration is accelerated under faster technological learning and higher WTP conditions. It is found that stabilizing CO 2 emissions to year 2005 levels causes economic losses amounting to 17% and 23% of GDP in the years 2020 and 2030, respectively.

Technological learning in energy–environment–economy modelling: A survey

This paper aims at providing an overview and a critical analysis of the technological learning concept and its incorporation in energy–environment–economy models. A special emphasis is put on surveying and discussing, through the so-called learning curve, both studies estimating learning rates in the energy field and studies incorporating endogenous technological learning in bottom-up and top-down models. The survey of learning rate estimations gives special attention to interpreting and explaining the sources of variability of estimated rates, which is shown to be mainly inherent in R&D expenditures, the problem of omitted variable bias, the endogeneity relationship and the role of spillovers. Large-scale models survey show that, despite some methodological and computational complexity related to the non-linearity and the non-convexity associated with the learning curve incorporation, results of the numerous modelling experiments give several new insights with regard to the analysis of the prospects of specific technological options and their cost decrease potential (bottom-up models), and with regard to the analysis of strategic considerations, especially inherent in the innovation and energy diffusion process, in particular the energy sector's endogenous responses to environment policy instruments (top-down models).

A MERGE model with endogenous technological change and the cost of carbon stabilization

Two stylized backstop systems with endogenous technological learning (ETL) are introduced in the “model for evaluating regional and global effects” (MERGE): one for the electric and the other for the non-electric markets. Then the model is applied to analyze the impacts of ETL on carbon-mitigation policy, contrasting the resulting impacts with the situation without ETL. We model research and development (R&D) spending and learning subsidies for the demonstration and deployment stage as control variables, and we investigate the ability of this extra spending to create path-dependent experience and knowledge to aid in the implementation of carbon-free technologies. Based on model estimations and sensitivity analyses, we conclude that increased commitments for the development of new technologies to advance along their learning curves has a potential for substantial reductions in the cost of mitigating climate change and thereby helping to reach safe concentrations of carbon in the atmosphere.

Implications of technological learning on the prospects for renewable energy technologies in Europe

The objective of this article is to examine the consequences of technological developments on the market diffusion of different renewable electricity technologies in the EU-25 until 2020, using a market simulation model (ADMIRE REBUS). It is assumed that from 2012 a harmonized trading system will be implemented, and a target of 24% renewable electricity (RES-E) in 2020 is set and met. By comparing optimistic and pessimistic endogenous technological learning scenarios, it is found that diffusion of onshore wind energy is relatively robust, regardless of technological development, but diffusion rates of offshore wind energy and biomass gasification greatly depend on their technological development. Competition between these two options and (existing) biomass combustion options largely determines the overall costs of electricity from renewables and the choice of technologies for the individual member countries. In the optimistic scenario, in 2020 the market price for RES-E is 1 €ct/kWh lower than in the pessimistic scenario (about 7 vs. 8 €ct/kWh). As a result, total RES-E production costs are 19% lower, and total governmental expenditures for RES-market stimulation are 30% lower in the optimistic scenario.

ETSAP-TIAM: the TIMES integrated assessment model Part I: Model structure

In this first part of a two-part article, the principal characteristics of the TIMES model and of its global incarnation as ETSAP-TIAM are presented and discussed. TIMES was conceived as a descendent of the MARKAL and EFOM paradigms, to which several new features were added to extend its functionalities and its applicability to the exploration of energy systems and the analysis of energy and environmental policies. The article stresses the technological nature of the model and its economic foundation and properties. The article stays at the conceptual and practical level, while a companion article is devoted to the more detailed formulation of TIMES equations. Special sections are devoted to the description of four optional features of TIMES: lumpy investments, endogenous technology learning, stochastic programming, and the climate module. The article ends with a brief description of recent applications of the ETSAP-TIAM model.

ETSAP-TIAM: the TIMES integrated assessment model. part II: mathematical formulation

This article is a companion to “ETSAP-TIAM: the TIMES integrated assessment model. part I: model structure”. It contains three sections, presenting respectively: the simplified formulation of the TIMES Linear Program (Sect. 1), the details of the computation of the supply demand equilibrium (Sect. 2), and the Endogenous Technology Learning Formulation (Sect. 3). The full details of these three formulations are available in the complete TIMES documentation at www.etsap/org/documentation.

Internalisation of external cost in the power generation sector: Analysis with Global Multi-regional MARKAL model

The Global MARKAL-Model (GMM), a multi-regional “bottom-up” partial equilibrium model of the global energy system with endogenous technological learning, is used to address impacts of internalisation of external costs from power production. This modelling approach imposes additional charges on electricity generation, which reflect the costs of environmental and health damages from local pollutants (SO 2, NO x ) and climate change, wastes, occupational health, risk of accidents, noise and other burdens. Technologies allowing abatement of pollutants emitted from power plants are rapidly introduced into the energy system, for example, desulphurisation, NO x removal, and CO 2 scrubbers. The modelling results indicate substantial changes in the electricity production system in favour of natural gas combined cycle, nuclear power and renewables induced by internalisation of external costs and also efficiency loss due to the use of scrubbers. Structural changes and fuel switching in the electricity sector result in significant reduction of emissions of both local pollution and CO 2 over the modelled time period. Strong decarbonisation impact of internalising local externalities suggests that ancillary benefits can be expected from policies directly addressing other issues then CO 2 mitigation. Finally, the detailed analysis of the total generation cost of different technologies points out that inclusion of external cost in the price of electricity increases competitiveness of non-fossil generation sources and fossil power plants with emission control.