Dear Reader,

For the 50th issue of our heatbeat Research Newsletter, we take a look into two research papers in the field of heat load prediction and optimal excess heat integration in district heating networks.

The first research paper is "A Novel Heat Load Prediction Algorithm Based on Weighted Fuzzy C-Mean Clustering and Feature Attention Enhanced Informer" from Song et al. evaluating the performance of a new heat load prediction algorithm in comparison to other state of the art algorithms. The second research paper "Strategic Integration of Urban Excess Heat Sources in District Heating: A Spatio-Temporal Optimisation Methodology" from Kumar et al. focuses on a geospatial optimization model to find the optimal integration of excess heat locations in Stockholm’s district heating system.

The first research paper proposes a novel heat load prediction model for district heating systems, combining Weighted Fuzzy C-Mean (WFCM) clustering with a Feature Attention Enhanced Informer (FAE-Informer). The aim is to improve heat load forecasting, which is critical for optimizing the efficiency and operational cost of smart district heating systems (SDHS).

The WFCM clustering assigns weights to features to better capture the spatial distribution of heat load data, particularly in ambiguous regions. This unsupervised approach enhances clustering accuracy by effectively revealing the significance of different features in the heat load dataset. On the other hand, the FAE-Informer incorporates dual attention mechanisms: feature attention for dynamically identifying critical features and sparse attention for efficiently modeling long-term dependencies. Together, these techniques significantly improve prediction accuracy and computational efficiency.

The model was evaluated using data from four heat exchange stations in northern China, with the proposed algorithm achieving a mean absolute percentage error (MAPE) of 2.812 %, 2.980 %, 2.518 %, and 1.665 % across these stations, outperforming state-of-the-art algorithms such as LSTNet, Informer, and DeepAR. Ablation experiments confirmed the individual contributions of WFCM and feature attention mechanisms to the model's overall performance.

The study demonstrates the potential of hybrid clustering and attention mechanisms to address the complexities of heat load prediction, particularly in handling nonlinear and highly volatile datasets. Future research should explore incorporating indoor temperature data, further optimizing model architecture, and extending applications to other time-series prediction scenarios.

The second research paper introduces a spatio-temporal optimization methodology for integrating urban excess heat (UEH) sources into district heating systems (DHS), addressing the need for sustainable heating solutions amidst the EU's transition away from fossil fuels. UEH, derived from sources such as data centers, sewage plants, and metro tunnels, represents a largely untapped potential to meet urban heating demands. However, challenges like low temperatures, high connection costs, and variable energy market dynamics have hindered its adoption.

The study develops a framework combining three open-source tools: geospatial optimization for mapping heat sources and network extensions, long-term planning to assess investment viability, and short-term dispatch analysis for operational feasibility. Applying this framework to Stockholm's DHS, where 80 % of buildings are connected, reveals that proximity to existing pipelines (within 5 km) and connection volume significantly affect UEH integration feasibility. Data centers emerge as the most viable sources due to higher temperatures and lower connection costs. Scenarios highlight that lower grid temperatures improve UEH uptake, but higher electricity prices constrain the economic viability of low-temperature sources like supermarkets and ice rinks.

The methodology underscores the importance of aligning spatial, economic, and temporal considerations for UEH integration. It demonstrates that, with optimized investment and operation strategies, DHS can effectively incorporate UEH sources, contributing to lower emissions and enhanced energy resilience. Future research could refine the framework to include more detailed operational constraints, dynamic pricing models, and stakeholder collaboration strategies, broadening its applicability to diverse urban contexts.

Further Information

As always, we recommend reading the articles in full. Both research topics are interesting and innovative fields in the field of district heating network. Therefore, we are currently part of two active research projects investigating these topics. The first one is "BeStWärmKI" where, together with our project partners, we combine forecast prediction models with our simulation models. The second one is called "AI-X-Heat" where we develop new geospatial optimization models for our heatbeat digital twin. The next issue of our newsletter will be published on January 8, 2025. If you do not want to wait until the next newsletter, you can attend our webinar about heating network projects with and without municipal heating planning under "https://www.enerpipe.de/seminarinfo/online-seminar-waermenetze-mit-und-ohne-kommunale-waermeplanung."