Transfer Learning Based Short-Term Load Forecasting at Different Aggregation Levels

Short-term load forecasting has continued to grow in importance, prompting the development of new forecasting methods. However, access to historical data is often limited, especially for some areas in which smart meters have not been installed or cannot be installed. This paper proposes an approach...

Full description

Saved in:
Bibliographic Details
Published in2024 1st International Conference on Innovative Engineering Sciences and Technological Research (ICIESTR) pp. 1 - 5
Main Authors Jiang, Kunfan, Donaldson, Daniel L.
Format Conference Proceeding
LanguageEnglish
Published IEEE 14.05.2024
Subjects
Accuracy
Aggregates
Benchmark testing
Data models
Forecasting
Load forecasting
Load modeling
Long short term memory
residential load forecasting
short-term load forecasting
smart grid
Training data
Transfer learning
Online AccessGet full text

Cover

More Information
Summary:Short-term load forecasting has continued to grow in importance, prompting the development of new forecasting methods. However, access to historical data is often limited, especially for some areas in which smart meters have not been installed or cannot be installed. This paper proposes an approach to use transfer learning to improve the accuracy of short-term load forecasting for different aggregation levels subject to data scarcity. Specifically, the paper investigates the benefit from using regional data from a system operator to enhance local forecasting. Experiments forecasting demand at different aggregation levels ranging from one to 114 apartments evidence the ability of transfer learning to improve predictions of aggregate local demand. This demonstrates the value of well-maintained regional load data.
DOI:10.1109/ICIESTR60916.2024.10798263