Utilizing open electricity consumption data in assessing household consumption flexibility
Duan, Yunge (2025)
Kandidaatintyö
2025
School of Energy Systems, Sähkötekniikka
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2025042932570
https://urn.fi/URN:NBN:fi-fe2025042932570
Tiivistelmä
This thesis examines the consumption flexibility of three household electricity customer groups in Finland during the year 2024, using the most recent national data from Fingrid. The study focuses on how spot market price variations affect household electricity usage, aiming to quantify and forecast demand-side flexibility.
The primary objective of this study is to:
- Develop a statistical model that describes the mathematical relationship between hourly household electricity consumption and spot market prices.
- Uncover temporal patterns in consumption flexibility.
- Quantify the flexibility potential of households for comparison between groups.
- Forecast future household electricity consumption using machine learning techniques.
The research methodology involves data processing, regression modeling, flexibility estimation, and time-series forecasting using LSTM networks. The results reveal that BE02 (electrically heated households) exhibit the highest flexibility, both in relative and per-household terms. Flexibility varies significantly throughout the day and year, with the strongest responsiveness occurring during late-night hours. Surprisingly, in winter, no negative demand response was observed in the long-term analysis; however, shorter-term surge events did show some evidence of demand response. These findings underscore the value of time- and group-specific analysis in assessing household demand response.
The primary objective of this study is to:
- Develop a statistical model that describes the mathematical relationship between hourly household electricity consumption and spot market prices.
- Uncover temporal patterns in consumption flexibility.
- Quantify the flexibility potential of households for comparison between groups.
- Forecast future household electricity consumption using machine learning techniques.
The research methodology involves data processing, regression modeling, flexibility estimation, and time-series forecasting using LSTM networks. The results reveal that BE02 (electrically heated households) exhibit the highest flexibility, both in relative and per-household terms. Flexibility varies significantly throughout the day and year, with the strongest responsiveness occurring during late-night hours. Surprisingly, in winter, no negative demand response was observed in the long-term analysis; however, shorter-term surge events did show some evidence of demand response. These findings underscore the value of time- and group-specific analysis in assessing household demand response.