Using Physics-Informed Machine Learning for reusing power system components

Purpose and goal:

The purpose of the project is to use innovative computational methods and take advantage of physics-informed neural networks (PINNs) to study the ageing of power components in a wind farm and find possible applications for their reuse. The goals of the project are to: - use PINNs to predict the component ageing

- study potential good applications for the method - identify methods for monitoring older components to ensure the system´s safety Expected results and effects: Expected results: - well-trained PINNs, which help to define and predict ageing of power components in a wind farm - proposed strategies for reuse and upcycling of power components in new applications

- discuss possible reliability impacts of the new application Expected effects: - novel end of life management strategies can reduce negative climate impacts of the renewable energy development

- Swedish and international businesses will be able to use our methodology to reduce capital investment on wind farms as well as get extra return on investment after wind farm decommissioning Approach and implementation: This project has started on the 1st of November 2021 and will continue until the 30th of June 2026.

The project will consist of 7 working packages. 1st working package will focus on identifying key assets for the study. WP 2 and 3 will be oriented towards the development of PINN methods and will compose the longest project stage. WP 4,5 and 6 will test the methods on several case studies and suggest possible new applications for the assets as well as give an economic evaluation of the results. WP 7 will focus on communication and dissemination of the project results.