Self-learning sensor networks for weather-dependent overhead line monitoring
January 1st 2019 to December 31st 2021
The rapid growth of renewable energy and the growing amount of transborder electricity trade increasingly exceed the transmission capacity of the existing power grid. Therefore, policy makers in Germany favor the optimization of transmission capacity of the existing grid over its costly and lenghty physical expansion according to the NOVA-principle. The transmission capacity of overhead lines is limited by a maximum conductor cable temperature of 80°C. Higher temperatures lead to the sagging of the overhead line due to its thermal expansion. The distance of the overhead line to the ground, vegetation or buildings decreases. However, to avoid dangerous operating conditions (such as flashovers), a certain minimum distance must be maintained even under the most adverse operating conditions.
This increasingly leads to temporal shutdowns of a large number of renewable energy generation plants (especially wind and solar power plants) to avoid transmission bottlenecks. Since a quick extension of the grid network is unrealistic due to red tape and excessive costs, it is paramount to increase the transmission capacity of the existing grid. A very promising approach to achieve this goal is overhead line monitoring with precise weather data. The project therefore aims to develop smart, self-learning meteorological networks, which precisely model the cooling effect of weather on overhead power lines in order to increase their transmission capacity by monitoring them in real time.
UBIMET’s main tasks in the project are:
- Developing highly precise forecasts of meteorological parameters along power lines with 3d and 4d VERA and MOS (Model Output Statistics). The empirical foundation of these forecasts are physical measuring stations along the overhead line and “virtual weather stations”. These are analytical values at significant points which – under certain quality criteria – can substitute real observation / measuring stations.
- In interaction with a machine learning approach which is based on Model Output Statistics (MOS), exact forecasts for significant points of the power line should be produced and transmitted in real time. In this respect weather situations on a large and smaller scale, topographic and orographic conditions, land use as well as the quality of the original data used for modelling have to be explored.
- Karlsruhe Institute of Technology (KIT)
- unilab AG
- Wilmers Messtechnik GmbH
- GWU-Umwelttechnik GmbH (associated partner)
- Transnet BW GmbH (associated partner)
German Federal Ministry for Economic Affairs and Energy (BMWi) within the energy research program „environmentally friendly, reliable and affordable energy supply”.
For more information: https://www.itiv.kit.edu/6518.php