ENE471 Machine Learning for Energy Markets Data

• Topics

  In several situations the mechanisms behind demand and price formation are very complex and market participants may be purely interested in predicting market outcomes. Machine learning is a large set of quantitative methods that attempts to predict and classify those outcomes. In this course we examine some of those standard prediction methods and we use them on data from energy markets. The use of these methods has its advantages and disadvantages with respect to a full modelling of the mechanisms at play in a given market, but the growing availability of data and computational tools have made machine learning methods easily accessible.

  The main topics to be covered are:

  1. Cross-validation vs classical estimation methods
  2. Linear models of prediction
  3. Classification methods
  4. Demand and price prediction in energy markets

• Learning outcome

  Upon successful completion, the students will:

  Knowledge

  • Have an in-depth understanding of linear models of prediction such as ordinary least squares, ridge regression, LASSO, and elastic net methods and how to implement them using data from energy markets.
  • Know how to classify demand load profiles using the k-means clustering method.
  • Have an understanding of the concept of cross-validation, which is used in all machine learning methods, not just those seen in this course.

  Skills

  • Have the ability to decide in what situations linear models of prediction or classification are well suited to solve a problem when dealing with data from energy markets.
  • Be able to apply linear prediction models to datasets from energy markets.
  • Be able to apply classification models and recognize when these methods are recommended.

  General competences

  • Be able to determine the data requirements to apply the abovementioned methods.
  • Capable of building a project to predict outcomes in energy markets.

• Teaching

  One week-long intensive course with lectures and hands-on sessions to apply the methods seen in the lectures.

• Credit reduction due to overlap

  None.

• Compulsory Activity

  Attendance is mandatory.

• Assessment

  • Mini quizzes during class (20%)
  • An individual mini paper to be submitted two weeks after course ends (80%)

  Both elements have to be taken in the same semester.

• Grading Scale

  Pass-Fail

• Computer tools

  R and Python. No previous knowledge of these languages is required. The course is self-contained. 

• Literature

  Bajari, P. et al., 2015. Machine Learning Methods for Demand Estimation. American Economic Review.

  Burkov, B., 2019. The Hundred-Page Machine Learning Book.

  James G. et al., 2018. An Introduction to Statistical Learning. Springer.

  Mullainathan, S. and Spiess, J. 2017. Machine Learning: An Applied Econometric Approach. Journal of Economic Perspectives.