Access to clean, affordable, and reliable energy is crucial for climate action and achieving the Sustainable Development Goals (SDGs). However, progress remains uneven, particularly in developing countries. Official development assistance (ODA) is a key driver of the transition to clean energy in these countries, but funding needs are unclear. Here, we use machine learning to estimate country-specific ODA needs to reach clean energy goals from SDG~7 (i.e., affordable and clean energy for all). Using counterfactual predictions of SDG outcomes under different ODA volumes, we quantify the ODA needs to reach the clean energy goals of SDG~7, namely, (i)~access to electricity, (ii)~access to clean fuels, (iii)~renewable energy consumption, and (iv)~energy efficiency. We apply our machine learning framework to a global dataset covering more than 120 recipient countries and USD~65.8 billion in ODA flows.

In collaboration with Prof. Stefan Feuerriegel and Kerstin Forster (University of Munich, Munich Center for Machine Learning)

Covering 26 coal-producing countries, this study integrates a systematic literature review of 106 studies and cluster analysis to scrutinize policies related to coal mining phase-out and just transition. Using a typology based on vulnerability criteria and types of policy instruments, we identify policy interventions and practices that mitigate or exacerbate vulnerabilities in coal-producing countries. Our review highlights the need for more emphasis on economic diversification policies for upcoming transitions, for which effective policy mixes are required.

In collaboration with Paola Yanguas Parra, Prof. Pao-Yu Oei, Andrea Furnaro, Lisa Wijayani (University of Flensburg, ZHAW Zurich)

This journal article analyzes the opportunities, costs, and risks of Carbon Capture and Storage (CCS) in Germany, comparing various deployment scenarios and quantifying project risk through a Monte Carlo simulation. Based on current research and the German Federal Government's Carbon Management Strategy, annual and cumulative costs for CO₂ capture, transport, and storage are calculated through 2045, with a cost range of 39.2 to 81.5 billion euros in the medium scenario. Project risks were estimated using a Monte Carlo simulation that mimics the CCS-capacity buildup under different failure rates.

In collaboration with Prof. Pao-Yu Oei, Jonas Gothe, and Philipp Herpich (University of Flensburg, TU Berlin)

Provides a comprehensive overview of China's complex energy landscape, exploring the paradox of its simultaneous expansion in renewable energy and continued reliance on coal. We looked at the structural evolution of China's energy system, highlighting past and recent challenges of governance, market dynamics, and technology integration. Additionally, the report explores China's influence on global energy and critical minerals markets and the international implications of its energy policy (Belt and Road Initiative).

In collaboration with Prof. Pao-Yu Oei (University of Flensburg)

Active learning approaches are essential in domains where labeled data is scarce or difficult to obtain. In this project, I critically review the deep active learning framework using Bayesian CNNs with Monte Carlo dropout for image data by Gal et al. (2017). Acquisition functions like (Batch)BALD, Max Entropy, and Variation Ratios are evaluated in the Bayesian and determinist active learning setting. Based on Beluch et al. (2018), they are likewise compared against ensemble and coreset methods using an adapted version of Munjal and Chandra's toolkit. Seminar presentation available here. Link to paper.

Seminar project at the University of Munich (LMU)

Interactive presentation at FossilExit Research Group on principles and usage of modern AI technology with a focus on Large Language Models (LLMs) and their applications in qualitative scientific research. Target audience: PhD students in social sciences and energy transitions research. Presentation slides available here.

Talk at FossilExit Research Group

Maintenance and unplanned outages are key sources of blackouts ('load shedding') in South Africa's power system. Using gradient boosted trees and convolutional neural networks on historic time series data, we attempted to predict outage occurence and magnitude, as well as other relevant patterns in the data.

In collaboration with Bryce McCall and Bruno Merven (University of Cape Town)

Examines the influence of raw metal product beneficiation on export price elasticity, focusing on chromium and manganese. Given the supply oligopolies present, chromium and manganese ore exporters are exploring the potential of refining raw chromium into more advanced products like ferrochromium within their borders. This could potentially stabilize prices, offering a more resilient export economy. Using the World Bank’s WITS COMTRADE database and a two-stage ordinary least-squares regression model, we assess the Price Elasticity of Demand (PED) for these metals and their processed products along the value chain. Our analysis of the interaction between product category and price suggests that processed chromium and manganese products might have a lower PED than their raw counterparts.

In collaboration with Bryce McCall (University of Cape Town)

Remoteness has been a popular yet ambiguous concept in the fields of geography, international relations, and economics, often associated with areas of seemingly limited economic and social importance. However, the concept has so far eschewed proper quanitification due to the concept's inherent relativity. This method aims to tackle this issue by proposing a new quantitative measure of remoteness based on weighted geometric medians. Two versions of the metric were developed: a monocentric and a polycentric (hierarchical clustering). The resulting metrics are flexible, scaleable, and easy to implement in other quantitative models, such as regression models. We aim to provide a precise, practical metric that brings clarity to the concept of remoteness, ultimately enhancing its applicability and relevance across multiple disciplines. To exemplify and operationalize the remoteness measurement method, we applied it to measure the relative remoteness of coal-producing regions of the 10 largest coal-producing countries.

In collaboration with Paola Yanguas Parra (TU Berlin, IISD, ZHAW Zurich)

Evaluates the potential impacts of the global energy crisis in 2022, following the invasion of Ukraine by Russia, on international coal markets and the risks and opportunities for coal exporting countries. Based on the COALMOD World v2.0 modelled short-term reaction and medium to long-term plans of the main coal consumers and producers, we claim that the temporary increase in demand and prices in mid 2020s will likely be followed by a rapid decrease in international coal demand that could lead to a ``boom and bust'' scenario. In contrast, if managed wisely, the additional short-term earnings that governments and companies are receiving as a result of the crisis (high coal prices) can become an important financing source for kicking off just transition processes in coal-dependent regions and communities.

In collaboration with Paola Yanguas Parra (TU Berlin, IISD, ZHAW Zurich) and several other colleagues at TU Berlin/EU Flensburg.

Analyzes the techno-economic, institutional and political factors that contribute to China's coal-related policies following a novel approach that blends different theories and frameworks to establish an interdisciplinary dialogue between various strands of research in economics, political economy and energy studies that were hitherto unconnected. This is accomplished by applying a political economy framework through the lens of techno-institutional carbon lock-in theory in three case studies encompassing China itself, as well as China's climate and energy policy abroad in Pakistan and Mozambique. The article concludes that China's energy paradox is driven by its coal-based growth imperative, suggests leveraging its decentralized governance to promote renewable energy, calls for addressing carbon lock-in by empowering renewable advocates, and advocates for halting coal financing abroad while promoting cleaner alternatives.

In collaboration with Paola Yanguas Parra (TU Berlin, IISD, ZHAW Zurich), Felipe Corral (TU Berlin) and Pao-Yu Oei (TU Berlin, EU Flensburg).

Through a series of publications in Spanish and a journal article, we explored political, economic, and sociological aspects of coal mining, export, and domestic consumption in Colombia. Furthermore, the country's continued reliance on coal is contrasted with its vast renewable energy potential. Impacts of coal mining on the surrounding communities in Cesar and La Guajira departments are highlighted as well as power dynamics leading to these outcomes examined. Publications in Spanish can be found here and here.

In collaboration with Felipe Corral (TU Berlin, Hertie School) and Max Telias (TU Berlin, FU Berlin); various colleagues at Universidad del Magdalena, Santa Marta, and Universidad Nacional de Colombia, Bogotá, Colombia.