177 / 215
Thesis - Machine Learning for short-term wind-speed forecasting
The steady progress in wind-turbine automation introduces high demands on the mathematical modelling of the turbine and its environment.
Especially new, sophisticated control methods like nonlinear model predictive control (NMPC) could benefit from a precise short-
term prediction of the relevant wind velocities. Compared to classical, statistical prediction methods (e.g. AR, ARMA, ARIMA), data-
based methods like machine- / deep-learning show promising room for improvement in wind prediction models.
Within the scope of your thesis you should develop and evaluate the suitability of machine- / deep-learning algorithms for the short-
term wind prediction ( 30s) of wind velocities, based on real measurement data. Your tasks include :
Comprehensive literature research regarding existing approaches
Development of a suitable neural network (NN) modelling approach
Implementation of the chosen NN in a suitable machine-learning framework (e.g. Tensorflow)
Performance evaluation and optimization with meaningful metrics
Quantitative assessment of the results with respect to state-of-the-art methods
Studys in the field of mathematics, computer science or electrical engineering
Experience in the field of machine learning
Good programming skills in MATLAB and Python
Experience with Tensorflow desirable
Fluent in German and / or English
Autonomous working and goal-oriented style
You will be working in a highly motivated team on demanding and trend-setting assignments with a great deal of scope for your ideas.
Our qualification programme promotes your professional and personal development. All of that offered in a modern, attractive working environment with a constantly growing, globally operating successful company.