LFU:online

Upon the successful completion of this course students will have gained a comprehensive understanding of time series analysis, linear dynamic systems, stochastic processes, and a selection of modelling techniques. Such knowledge can be used to predict and model behaviours in fields like e.g. finance, economics, epidemiology, climate science, or energy systems. Students will be equipped with the skills to select appropriate models, estimate parameters, and validate their models. The non-exclusive focus will be on grey-box approaches, i.e. including physical knowledge, for modelling and forecasting of energy systems.

- Introduction time series

• Time series general, linear dynamic systems, applications (forecasting, system characterisation, Anomaly detection, model predictive control)
• Overview models, State-space and grey-box models
• Deterministic vs. stochastic processes, stochastic vs. ODE’s
• Data management, Time series objects in R (recap Data management lecture)

- Stochastic processes

• Random variable, random walk, white noise
• Stationarity, ergodicity
• Correlation, Autocorrelation
• Backshift Operator
• Discrete time models: AR, MA, ARMA, Non-stationary processes (ARIMA)
• Continuous time models: state space, grey-box models, Kalman filter

- Selection, estimation and model validation

• Under-/ overfitting
• Akaike’s and Bayesian Information criteria (AIC, BIC)
• Selection procedure
• Estimation using Least-Square method
• Estimation using Maximum Likelihood method
• Model validation

- Application to real world problems, e.g. prediction of energy demand of a city, generating control models to maximise renewable energy usage or developing fault detection and/or predictive maintenance algorithms.

The course consists of a mix of lectures, i.e. presentation by lecturer, hands-on classroom exercises and small (individual and/or group) projects and their presentation by students.

The exercises in the course will be conducted in R, with specific packages such as CTSM-R. Instruction for use of R will be provided, however, it is recommended that participants familiarize themselves with the software (see prerequisites). Please make sure that R and RStudio are installed on your laptops. To download R, go to https://www.r-project.org/, for RStudio, go to https://posit.co/download/rstudio-desktop/.

Participants are graded by oral exam at the end of the course (50%) and by in-class presentation of student projects (50%). Further details will be announced in the first session.

The full list of references will be provided in the course. Selected preliminary references:

• Madsen, Henrik. Time Series Analysis. 2008. Chapman & Hall/CRC
• Ljung, Lennart. System Identification: Theory for the User. 2. ed., 1999. Prentice Hall PTR
• Sanchez, Juana. Time Series for Data Scientists: Data Management, Description, Modeling and Forecasting. 2023. Cambridge University Press
• Time Series Analysis with R, Nicola Righetti. 2022

The following tools will be used amongst others:

• R: https://www.r-project.org/
• CTSM-R: https://www.ctsm.info/ 

It is recommended but not required to …

• have proficiency in linear algebra,
• familiarity with probability and statistics,
• have prior exposure to R or a similar programming language.

Therefore this course is recommended after completing 198803 VU Introduction to Programming: Programming in R course.

Students advanced in completion of the Minor Digital Science get precedence, especialy these who passed module 1 (R). 

Students advanced in completion of the Minor Digital Science get precedence.