Date: TBA
Time: TBA
Place: TBA

Energy consumption is a critical cost driver and a substantial sustainability challenge for large businesses and organizations.

This 4hrs workshop provides an interactive session on applied time series forecasting in energy analytics and gives an overview of the most common forecasting methods used in retail to predict energy consumption.

The session starts laying out core challenges and classical time series forecasting methods, then advances to the supreme discipline: time series foundation models.

The workshop concludes with an open discussion allowing participants to bring their own time series challenges to the table.

By participating in this tutorial, beginners will gain a practical introduction to time series forecasting in energy analytics, while advanced attendees will acquire insights into real-world pitfalls and knowledge of state-of-the-art foundation models in this evolving field.

Outline

1. Introduction and Baselines
   • Overview of challenges in forecasting energy consumption for large retailers.
   • Establishing robust baseline forecasts and evaluation schemes.
   • Walkthrough of basic time series forecasting methodologies.
   • Problems in real world energy consumption data.
2. Advanced Methods. Introduction to time series foundation models.
3. Hands-On Coding session. Application of the discussed forecasting techniques against:
   • Hindcasting / Backtesting (Anomaly detection)
   • Counterfactual Analysis (What-if scenarios)
   • Predictive Forecasting (Resource planning)
4. Discussion. Practitioners challenges
   • Pitfalls and learnings in bringing forecasts into business applications.
   • Open discussion about participants’ challenges.

Target Audience

This workshop is designed for researchers, data scientists, and industry practitioners interested in applied time series analysis, energy data, and data-driven decision-making.

Prerequisites

Participants should have working knowledge of Python (e.g., Pandas, Scikit-learn) and basic familiarity with time series and machine learning concepts.

• Suitability: Ideal for both academic researchers looking to explore complex, real-world industrial datasets and professionals seeking to deepen their applied analytical toolkit.

Materials Needed

• Laptop with access to the internet and terminal access
• Git account (GitHub) for accessing workshop materials, datasets, and starter code
• Python 3.9+
• Basic proficiency in Python programming and tabular data manipulation

Presenters

Dr. Dominik Heinisch ([Lead Data Scientist Real Estate], Lidl Data & AI). Dominik holds a PhD in Economics and brings over 15 years of experience in Big Data Analytics, including seven years focused on industry forecasting. He also serves as a guest lecturer in forecasting at the University of Kassel. His professional interests are centered on bridging the gap between technical analytics and industry operations by integrating data science into practical business applications.

Benedikt Lugauer ([Data Scientist], Lidl Data & AI). Benedikt is a trained psychologist with over five years of experience in the field of data science. His core expertise lies in statistical methodology, specifically causal inference, machine learning, and variable selection. He also worked as a guest researcher at the Max Planck Institute (MPI) in Leipzig. His current work and research interests are strongly focused on the application and development of foundation models.