Conference Agenda

Energy poverty, a multidimensional socio-economic challenge, significantly affects the welfare of many people across Europe. This paper aims to alleviate energy poverty by exploring sustainable energy practices and policy interventions, using household survey data from Portugal and Denmark. A Multidimensional energy poverty index (MEPI) is developed to assess energy poverty through different dimensions such as heating and cooling comfort, financial strain, access to energy-efficient appliances, and overall health and well-being. In a next step, for selecting features, machine learning techniques are employed including recursive feature elimination and random forest analysis. These methods help to reduce the number of irrelevant and mutually correlated predictors. Subsequently, a logistic regression model is used to predict energy-poor households based on selected socio-economic, and policy-related factors. The logistic regression results indicate that sustainable energy-saving behaviors and supportive government policies can mitigate energy poverty. Furthermore, for analyzing the impact of determined features the Shapley additive explanations (SHAP) method is being utilized. Finally, the main findings are evaluated further via scenario simulation analysis. The result shows that fully adopting waste-compositing and energy-efficient microwave ovens can decrease the proportion of energy-poor households by 93% and 79%, respectively. This paper contributes to the literature by providing a thorough analytical framework that enables to identify key features for alleviating energy poverty via effective policymaking and promoting energy-conscious consumption. Its framework can be also applied for studying broader implications in other regions of Europe.

Energy system optimization models are key to investigate energy transition paths towards a decarbonized future. Since this approach comes with intrinsic uncertainties, it is not sufficient to compute a single optimal solution to provide a profound basis for decision makers. The paradigm of modeling to generate alternatives enables to explore the near-optimal solution space to a certain extent. However, large-scale energy models require a non-negligible amount of computation time to be solved. We propose to use warm start methods to accelerate the process of finding close-to-optimal alternatives. In an extensive case study for the energy transition of the Berlin-Brandenburg area, we make use of the sector-coupled linear programming oemof-B3 model to analyze several scenarios for the year 2050 with a resolution of one hour and up to 100% reduction of greenhouse gas emissions; and we demonstrate that we can actually achieve a significant computational speedup.

The electrification of buildings and transportation is one of the key pillars to combat climate change by reducing carbon emissions. The rapid growth of distributed energy assets such as rooftop solar, heat pumps, and electric vehicles necessitates a more decentralized grid architecture. In this context, local electricity markets can facilitate the integration of distributed energy assets and enhance their utilization by aligning distributed generation with consumption. However, existing market designs lack either realistic models of intelligent agents or physical grid constraints, or both. To this end, we propose an opengate forward market design that integrates realistic grid constraints by solving a security constraint optimal power flow problem using linear AC power flow approximation. The market is implemented as an open-source simulation framework based on the OpenAI Gym environment. This Python-based environment allows for the seamless integration of residential agents with state-of-the-art decision-making algorithms. To demonstrate the effectiveness of the proposed framework, we compare two optimization techniques, mixed integer linear programming, and deep reinforcement learning, for solving the residential agent energy management control problem. The first experiment examines the performance of the reinforcement learning optimizer under perfect forecasts, using the linear programming solution as an upper bound. The second experiment examines the robustness of the optimizers to uncertainty by comparing their performance under different degrees of forecast uncertainty. Finally, we present the potential of the proposed framework to support future research on multiagent dynamics, strategic bidding behavior, and the impact of intelligent agents on grid infrastructure.