Electric Load Forecasting in Sports Facilities using AI / ML
In this thesis project, the student(s) will develop AI/ML-based algorithms to forecast electrical loads in sports facilities. Furthermore, the thesis will investigate how these algorithms can be used to impute data under conditions with limited or missing data. The case study is Bosön Sports Development Center, which is a state-of-the-art facility for elite athletes and sports development including different types of sports facilities, as well as hotels, restaurants, school and other types of buildings. If there is interest, the thesis project can be conducted with an industrial partner.
Background
The increasing electrification of buildings and the transition towards more dynamic energy systems creates new opportunities for data-driven energy management. In this context, AI/ML becomes increasingly relevant tools for understanding complex energy systems, identifying patterns in energy use and forecasting future demand. However, data-driven energy management require large amounts of data, which are sometimes missing or has limited access.
Bosön Sports Development Center is powered by electricity with an internal GSHP facility and a small-scale district heating and cooling network, where high-resolution load data are collected by a research monitoring system. As part of the ongoing SPORT-RENOV research project, understanding and forecasting of electrical loads is important to manage the building energy systems. Therefore, this thesis will contribute both to research and development in AI/ML and to practical applications in a case study that can lead to more energy- and cost-effective energy systems.
Figure 1: Bosön Sports Development Center, Stockholm
Task description
-
Review and summarize the state-of-the-art for electric load forecasting algorithms and identify the main drivers of electricity demand in sports facilities.
-
Identify, collect and clean relevant training and testing data.
-
Identify suitable key performance indicators to assess the forecasting performance.
-
Develop, train and optimize a forecasting algorithm and analyse its performance and bias.
-
Analyze the algorithms’ applicability for data imputation under conditions where data are limited or missing.
-
Discuss the difficulty in predicting electric loads in sports facilities, the factors behind it and how they can be addressed.
-
Propose forecasting algorithms based on different objectives.
Learning outcomes
By the end of this thesis project, the student will be able to:
-
Review and compile scientific literature and collect data relevant to electric load forecasting.
-
Select and assess different algorithms and relevant key performance indicators applicable to electric load forecasting.
-
Provide recommendations for forecasting and data imputation under conditions where data are limited or missing.
-
Communicate complex technical findings effectively through structured reports and presentations tailored to both technical and non-technical audiences.
Duration
Starting in Spring 2027
Supervisors/Examiner
