TY - CHAP A1 - Alasmar, Odai A1 - Neumayer, Martin A1 - Bucker, Dominikus T1 - Data Augmentation Technique for Dealing with Multi-Resolution Issues in Segmentation of Photovoltaic Systems in Aerial Imagery T2 - 2024 International Conference on Electrical, Computer and Energy Technologies (ICECET N2 - Given the urgent global challenges posed by climate change the transition to renewable energy and the reduction of carbon emissions is of paramount importance. Automatically detecting photovoltaic (PV) systems in aerial imagery is crucial for understanding, planning and optimizing our energy infras- tructure. However, the task is complicated by the significant variability in the ground sampling distance (GSD) of available aerial imagery. This directly affects the spatial resolution and consequently the quality and applicability of training data for deep learning models. Typically, available datasets have high- resolution imagery, making them preferable for annotation due to their detailed visual information. In contrast, real-world applications often have to deal with lower-resolution images. This discrepancy poses a challenge for training models that can make accurate predictions under varying real-world conditions. Our research presents an approach that deliberately degrades the resolution of high-quality training images to match the lower- resolution images encountered in real-world applications. By training on degraded resolution data, we ensure that the models are not overly tuned to high-resolution features that are often not present in the target application scenarios. The comparative analysis of three state-of-the-art models, with and without ap- plying our resolution degradation method, shows a considerable improvement in prediction accuracy and model generalisability when applying our resolution degradation method. This study highlights the importance of matching the resolution of training data to that of real-world applications to develop robust and universally applicable PV detection models. Y1 - 2024 U6 - https://doi.org/10.1109/ICECET61485.2024.10698167 SP - 1 EP - 6 PB - IEEE ER - TY - JOUR A1 - Stecher, Dominik A1 - Neumayer, Martin A1 - Ramachandran, Adithya A1 - Hort, Anastasia A1 - Maier, Andreas A1 - Bücker, Dominikus A1 - Schmidt, Jochen T1 - Creating a labelled district heating data set: From anomaly detection towards fault detection JF - Energy N2 - For an efficient operation of district heating systems, being able to detect anomalies and faults at an early stage is highly desirable. Here, data-driven machine learning methods can be a cornerstone, particularly for fault detection in district heating substations, where the availability of heat meter data keeps increasing. However, the creation of data sets suitable for training such machine learning models poses challenges to researchers and practitioners alike. To address this problem, we propose a systematic and domain-specific process for data set creation for fault detection in the form of practical guidelines. This process concretizes the data science and data mining cross-industry standard CRISP-DM for the district heating domain and focuses on the process steps of goal definition, data acquisition and understanding, and data curation. We aim to enable researchers and practitioners to create data sets for fault detection in the district heating domain and therefore also enable the creation or improvement of machine learning models in this domain. In addition, we propose a minimum viable feature set for fault detection in district heating networks with the goal of enabling better cooperation between researchers and easier transfer of the resulting machine learning models, to better proliferate new progress in the field. KW - Machine Learning Y1 - 2024 U6 - https://doi.org/10.1016/j.energy.2024.134016 VL - 313 SP - 134016 PB - Elsevier ER -