HVDC cables are important for the ongoing electrification of the society to more sustainable. Solar parks and wind farms are used as renewable energy sources; however, these are often located far from where the energy is needed. To transfer electrical energy long distance, high voltage direct current (HVDC) technology is used, where as high voltage as possible is preferred in order to limit the losses. Increased state-of-the-art understanding of the underlying physical processes is important to push the efficiency, compactness, and reliability to a next level. This put high demands on the insulation material used. Even though HVDC cables can be long, 100-1000 kilometers, the high demands on the material goes down to properties at atomic scale.
The concentration and diffusion of different chemical species play an important role in the conductivity and breakdown of a high voltage cable. The present accepted model is simulation of diffusion process by using the Fick's laws. The Fick's laws relate the chemical flow due to diffusion process to the spatial concentration. Here, the most well-known parameter is the diffusion coefficient that depends on the type of chemical species, temperature, and concentration. For the present problem, a Physics-Informed Neural Network model is the combination of loss function from Fick's differential equation and traditional neural network. The trained neural network model would be able to predict the diffusion process and concentration in a complex dependence of diffusion coefficient to temperature, concentration and other unknown possible parameters.
**Tasks**
In this thesis we will use neural network combined with the governing equation of the diffusion process from the Fick's law. The first step will be to make a literature review of where PINN used for simple cases. The next step, is installing required libraries and practical learning of PINN model. Finally the PINN model will be used to train the neural network with the available experimental data.
**Your Background**
We think you have an education and interest in theoretical physics, applied mathematics, machine learning or similar field. You have taken courses in machine learning and data analysis.
NKT is committed to fostering a diverse organization and a culture where people from different backgrounds can thrive and are inspired to perform at their best. We believe that a diverse organization enables sustainable performance, and that an inclusive and welcoming culture makes for a better place to work.
This thesis will be done at NKT HV Cables AB, Technology Consulting in Västerås, Sweden. The thesis correspond to 30 hp at full speed and is expected to be run from January 2025.
**Union representatives**
LI-NKT
**About NKT**:
- Join a diverse and international team of experts developing the power cable technology of the future with focus on deeper sea, lower losses and higher performance. Technology is leading the corporate R&D program including material development and operating some of the most advanced high-voltage test centers in the industry. NKT also operates a Technology Consulting center in Sweden, where technical experts and scientists supports industries worldwide in cross-disciplinary R&D projects and technical investigations._
- NKT connects a greener world with high-quality power cable technology and takes centre stage as the world moves towards green energy. NKT designs, manufactures and installs low-, medium