Publications

• A scalable, physics-based platform was developed for multi-building flexibilities.
• Benchmarking and analysing multi-scale building flexibility controls were enabled.
• A communication hub realized real-time links between models and control agents.
• MuFlex offered a modular, standardized design with the latest OpenAI Gym support.
• A well-tuned Soft Actor-Critic coordinated flexibility across four office buildings.

With the increasing penetration of renewable generation on the power grid, maintaining system balance requires coordinated demand flexibility from aggregations of buildings. Reinforcement learning (RL) has been widely explored for building controls because of its model-free nature. Open-source simulation testbeds are essential not only for training RL agents but also for fairly benchmarking control strategies. However, most building-sector testbeds target single buildings; multi-building platforms are relatively limited and typically rely on simplified models or data-driven approaches. To address these gaps, MuFlex, a scalable, open-source platform for multi-building flexibility coordination, was developed. MuFlex enables synchronous information exchange and co-simulation across multiple detailed building models programmed in EnergyPlus and Modelica, and adheres to the latest OpenAI Gym interface. The platform is released open-source on GitHub: https://github.com/BuildNexusX/MuFlex.

Wu, Z., Korolija, I. & Tang, R. MuFlex: A Scalable, Physics-based Platform for Multi-Building Flexibility Analysis and Coordination. Energy 140565 (2026). https://doi.org/10.1016/j.energy.2026.140565.

• Modeling of energy sharing in low-voltage distribution networks.
• Analysis of Energy systems’ configurations, pricing, and future projections.
• Assessment of Grid performance on loading, voltage, and reinforcement costs.
• Implementing real-world data from Munich, Germany, with upcoming regulations.
• Provision of insights into renewable energy communities for, e.g., regulators, utilities, and DSOs.

The shift towards renewable energy sources, which is especially significant in the residential sector, relies on distributed energy resources (DER) like PV systems, heat pumps, battery storage systems, and electric vehicles. Integrating DERs into low-voltage distribution networks presents challenges, including potential grid instabilities. Energy sharing is a consumer-centric market approach, allowing consumers and prosumers to establish renewable energy communities (RECs) and share energy generated via DERs. Existing literature concerning energy sharing often prioritizes highlighting the benefits it offers to participants, rather than examining its direct impacts on established system boundaries such as distribution grid infrastructure. To this end, we employ a sequential modeling approach to study the integration of energy sharing schemes facilitated by RECs and their impacts on grid performance metrics, such as component loading, voltage magnitudes, and grid reinforcement costs. We examine twelve scenarios reflecting different REC configurations, DER adoption levels, and pricing strategies for both current (2023) and future (2037) contexts in Germany. Our findings indicate that implementing energy sharing not only results in considerable cost savings at the community level (with potential savings of up to 80% compared to scenarios without energy sharing) but also brings about significant reductions in grid asset loading (with decreases in transformer loading of up to 68% and line loading of up to 62%, compared to baseline scenarios). Conversely, we show that energy sharing can significantly influence voltage magnitudes at various nodes within the grid, potentially leading to substantial increases in grid reinforcement costs in future scenarios (i.e., 2037). Our research provides valuable insights for REC participants, regulators, and DSOs to understand the impacts of energy sharing on European low-voltage distribution networks and explore mitigation options, such as grid reinforcement measures.

Lersch, J., Tang, R., Weibelzahl, M., Weissflog, J. & Wu, Z. Assessing the impacts of energy sharing on low voltage distribution networks: Insights into electrification and electricity pricing in Germany. Applied Energy 378, 124743 (2025). https://doi.org/10.1016/j.apenergy.2024.124743.

• Model-free and model-based reinforcement learning algorithms were implemented.
• Reinforcement learning of value-based, policy gradient and actor-critic are discussed.
• Training efficiency and control performance of algorithms were investigated.
• Data demand and robustness to hyperparameter tuning of algorithms were analyzed.

Reinforcement Learning (RL) has presented considerable potential as an advanced control technique in building controls to enable buildings operating more energy-efficient. As various types of RL algorithms have been studied on their performance of building controls, benchmarking these algorithms across the entire spectrum of features is essential to provide an overview and deepen the understanding of RL applications. Therefore, this study aims to compare and analyze the effectiveness of various RL algorithms, encompassing the entire RL categories featured by value-based, policy gradient, actor-critic and model-based RL considering model availability and policy representation. To provide a comprehensive analysis, in addition to the control performance quantified by the cumulative rewards based on the cost function of RL, data demand and robustness of hyperparameter tuning were investigated. The open-source Gym-Eplus framework was selected as the virtual environment to train and test different RL agents. The results showed that both model-free and model-based RL agents outperformed the baseline rule-based control in terms of energy consumption and thermal comfort, and RL agents were capable of evaluating both short-term and long-term rewards to achieve adaptive control optimization continuously along with the online control process. Model-based RL agent improved the data sampling efficiency but presented a relatively sacrificed control performance during the tested summer days.

Wu, Z., Zhang, W., Tang, R., Wang, H. & Korolija, I. Reinforcement learning in building controls: A comparative study of algorithms considering model availability and policy representation. Journal of Building Engineering 90, 109497 (2024).