Posters

Abstract

Wind measurements provide important information for scientific research and energy yield assessments, informing investment decisions during the wind farm planning stage. Strict standards are enforced to ensure high-quality continuous, long-term data to reduce uncertainty. To meet these requirements, measurement devices must exhibit high levels of technical robustness and possess reliable autonomous energy supply systems. Remote sensing devices (RSD) and meteorological masts are often deployed in remote and environmentally challenging locations. Therefore, they are typically powered by solar modules and methanol fuel cells which need to be regularly maintained. Failure in power supply or measurement components can lead to reduced data quality, data loss, extended measuring periods and increased cost.   This project aims to develop an intelligent monitoring system with integrated predictive maintenance capabilities. The focus lies on the detection, analysis, and — where feasible — prediction of faults and performance degradation in critical subsystems, such as energy supply and measurement hardware. Multi-year operational datasets from Pavana GmbH serve as the foundation, enabling the application of advanced AI techniques, including machine learning and deep learning. Depending on the model, feature extraction and feature engineering are sometimes applied to derive relevant indicators from raw sensor data. These features are then used in supervised and unsupervised learning models to identify correlations between sensor behavior and fault occurrences, enabling early failure detection and predictive maintenance.  The tool visualizes system states, issues early warnings, and supports predictive maintenance strategies. By improving fault detection and prediction, the system enhances maintainability, ensures higher data availability, and reduces operational risks and costs in the domain of wind energy measurement technology.