ZURUECK HOCH VOR INHALT SUCHEN

» Back to overview
Proposing Institution

Institut für Informatik, Lehrstuhl für Rechnertechnik und Rechnerorganisation ,TUM Garching
Project Manager

Amir Raoofy
Boltzmannstr. 3
85748 Garching
Abstract
The recent trends in energy technology and corresponding shift towards the renewable energy has introduced new challenges for grid and power plant operators; to compensate for the fluctuations in power grids, the Gas power plants need to be operated at partial load or large variable load ranges. So, there is a great need for optimization with regards to efficient and low-pollutant operation of the Gas power plant. Such optimizations would potentially lead to efficient and economic operation of the power plant, reduced maintenance costs, longer machine life and higher availability of the power plant. In order to perform these optimizations, operational data from sensors in power plants are collected and analyzed using measurement and monitoring systems based on the desires of the producers and the operators.Huge amount of data are collected using the sensors in gas turbines and can potentially be used for optimization of the Gas power plants. For instance, raw data from dynamic vibration sensors with sampling frequency up to 25.6 kHz. Handling this huge amount of high-frequency data requires storage, computation and analysis techniques and in Gas Turbine Optimization (TurbO) project, we try to address this problem using Big Data; analysis of high-frequency data and improvement of storage techniques and lossy compression to store aggregated raw data are some examples of the research topics TurbO. Since we are dealing with a Big Data problem, namely analyzing huge amount of data, we need to use computer resources at LRZ to develop our Machine Learning programs in an HPC environment. At the moment we would like to test the Apache Spark for our data analysis and we are really interested to investigate the performance of Apache Spark in SuperMUC.

Impressum, Conny Wendler