Forecasting for Energy Systems

Details

Level	Master
Language	English
Semester	Summer
ECTS (SWS)	3 (2)
Lecturer	Prof. Dr. Anke Weidlich
Lecture in HisInOne	Link

Content

This lecture covers the most prominent forecasting algorithms currently applied in energy system analysis. Mostly time series are considered. We discuss forecast quality measures and time series components, address statistical approaches such as linear regression or (S)ARIMA, and learn several machine learning approches based on artificial neural networks

Course outline

• Preparatory topics
  • Linear correlation
  • Linear regression
  • Forecast quality measures
• Naϊve approaches & time series decomposition
  • Time series components
  • Simple forecasting methods
• Statistical modeling
  • Multiple linear regression
  • Exponential smoothing
  • ARIMA, SARIMA
• Machine learning approaches
  • Multi-layer perceptron
  • Recurrent neural networks, LSTM
  • K-nearest neighbor regression
  • Random forest regression
  • Support vector machines
• Feature engineering

Highlights

This lecture appeared in the Top 5 of all lectures in Sustainable Systems Engineering twice (see ranking website)

• #2 in winter semester 2022/23 (under the former name Optimization and Forecasting for Energy Systems)
• #2 in winter semester 2021/22 (under the former name Operations Research for Energy Systems)

We do practical calculation examples, primarily using MS Excel, so programming skills are not required. However, for understanding the machine learnign-based methods, a basic understanding of Python is helpful.