Predicting electricity demand using machine learning - Case study of Oslo Airport Gardermoen - Master’s Thesis 2023

Aug 5, 2026

Upload by Hazel

Future electrical power systems face rising electricity needs and a more complex generation mix, which demands intelligent methods for efficient resource management. This thesis investigates Oslo Airport Gardermoen (OSL) as a case study to predict electricity demand using LSTM machine learning models, with emphasis on peak demand forecasting. Models are trained on 2022–2023 data using consumption measurements and exogenous inputs such as passenger numbers, outdoor temperature, and electricity prices.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages69

Reading Time2 h 54 min

AccessFree to download

1 Introduction

1.1 Background

1.2 Motivation

1.3 Problem Statement

2 Theory

2.1 Power Systems

2.2 Machine Learning

Master’s Thesis 2023 30 ECTS Faculty of Science and Technology (REALTEK) Predicting electricity demand using machine learning: Case study of Oslo Airport Gardermoen Prediksjon av elektrisitetsbruk ved bruk av maskinlæring: Casestudie av Oslo Lufthavn Gardermoen Sigurd Grøtan Environmental Physics and Renewable Energy ii Abstract With an increasing need for electricity in society and a more complex energy production mix, future electrical power systems require intelligent systems for efficient resource management. In order to realize the potential of flexible resources in the power grid, robust and accurate prediction methods are required. This thesis presents a case study of Oslo Airport Gardermoen (OSL) to explore the potential of Long Short-Term Memory (LSTM) machine learning models in predicting electricity demand, particularly focusing on peak demand forecasting. The models are trained on data from 2022 and 2023, utilizing electricity consumption measurements and exogenous factors including passenger numbers, outdoor temperature, and electricity prices. The models demonstrate high accuracy in demand prediction, particularly for peak hours. To improve peak prediction capabilities, the thesis implements two main strategies. First, models are trained using four different loss functions: Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), Negative Log Likelihood (NLL), and a new proposed Weighted Mean Squared Error (WMSE) . Second, a comprehensive grid search and cross-validation routine is performed to robustly determine the optimal model architectures. The best-performing models are characterized by simple model architectures with just 1 hidden layer and 64 or 128 units, suggesting that less complex models can efficiently capture the patterns of the data. These models achieve adequate MAPE scores, with the lowest being 4.53% . The new proposed WMSE loss function emphasizes peak hours and significantly enhances peak prediction reliability. Additionally, NLL enables probabilistic outputs, offering valuable uncertainty estimations for practical applications. This thesis provides a robust and versatile framework adaptable to various energy systems, enabling the development of optimized LSTM models for efficient electricity demand forecasting. The implications of this work extend beyond OSL, offering insights for managing flexible resources for efficient and sustainable power system operation. The thesis highlights the promising potential in using advanced machine learning methods for energy management systems, and demonstrates their ability in large-scale commercial buildings. iii Sammendrag Med økende elektrifisering i samfunnet og mer kompleks energiproduksjon trenger fremtidens kraftsystemer intelligente systemer for effektiv ressursbruk. For å kunne bruke fleksible løsninger i kraftnettet er det et behov for robuste og treffsikre prediksjonsmetoder. Denne masteroppgaven er en casestudie av Oslo Lufthavn Gardermoen (OSL) som undersøker potensialet i å benytte maskinlæringsmodeller basert på Long Short-Term Memory (LSTM) for å predikere strømforbruk, medet spesielt fokuspå forbrukstoppene. Modellene trenes på data fra 2022 og 2023, og bruker målinger av tidligere strømforbruk og forklaringsvariabler som passasjertall, utetemperatur, og strømpriser. Modellene predikerer strømforbruket med god treffsikkerhet, særlig med tanke på forbrukstoppene. For å forbedre prediksjonene av toppene benyttes to hovedmetoder. Den første er at modellene trenes med fire forskjellige tapsfunksjoner: Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), Negative Log Likelihood (NLL), og en ny foreslått tapsfunksjon Weighted Mean Squared Error (WMSE) . Den andre metoden er at det gjennomføres en omfattende grid search medkryssvalidering for å finne optimaliserte modellarkitekturer. Modellene med best ytelse har enkle modellarkitekturer bestående av bare 1 skjult lag og enten 64 eller 128 noder, n

FAQs

Which machine learning approach is used to forecast electricity demand?

The thesis uses Long Short-Term Memory (LSTM) models for electricity demand prediction, with a focus on peak demand forecasting.

What data inputs are used for training the forecasting models?

Training uses electricity consumption measurements and exogenous factors including passenger numbers, outdoor temperature, and electricity prices, based on data from 2022 and 2023.

How does the thesis improve peak demand prediction performance?

It applies two strategies: training with four loss functions including a proposed Weighted Mean Squared Error (WMSE), and using grid search with cross-validation to identify optimal model architectures.

Aug 5, 2026

Upload by Hazel