Back to search

NAERINGSPH-Nærings-phd

Applied Transfer Learning in the Energy Domain

Alternative title: Anvendt læringsoverføring i energidomenet

Awarded: NOK 1.8 mill.

Project Manager:

Project Number:

329073

Application Type:

Project Period:

2021 - 2025

Funding received from:

Organisation:

Machine learning is becoming increasingly important for operating the electrical power system, and machine learning algorithms are highly dependent on historical data. Modern machine learning algorithms based on deep neural networks have superior results compared to traditional machine learning methods. They do require a significant amount of training data to achieve these results. However, the amount of available historical data for specific electrical systems is however often limited, particularly for newly built systems. This project aims to investigate the use of transfer learning for machine learning problems within the energy sector. Transfer learning is an approach to machine learning that aims to utilize data and learned knowledge from different, but not identical problems for solving a target problem. In this project, different transfer learning methods will be developed, implemented, and tested on energy systems managed by TrønderEnergi. Currently, a literature study on transfer learning is being done. The first application area is wind power forecasting.

Machine learning is becoming increasingly important for operating the electrical power system, and machine learning algorithms are highly dependent on historical data. Modern machine learning algorithms based on deep neural networks have superior results compared to traditional machine learning methods. They do require a significant amount of training data to achieve these results. However, the amount of available historical data for specific electrical systems is however often limited, particularly for newly built systems. This project aims to investigate the use of transfer learning for machine learning problems within the energy sector. Transfer learning is an approach to machine learning that aims to utilize data and learned knowledge from different, but not identical problems for solving a target problem. In this project, different transfer learning methods will be developed, implemented, and tested on energy systems managed by Aneo AS. A successful project will lead to the development of new methods deployed in production systems used by TrønderEnergi to improve the efficiency of the energy system. While the results will increase the competitiveness of TrønderEnergi, they will also be published in scientific and trade journals to have a broader impact.

Funding scheme:

NAERINGSPH-Nærings-phd