The optimal operation of high-voltage underground power cables is crucial for powering our communities, and it hinges on the intricate dynamics of insulation temperature around the conductor, primarily influenced by joule heating. This temperature responsiveness is further molded by seasonal and diurnal fluctuations in power demand, as well as the moisture content in the surrounding soil. Past research concentrated on theoretical analyses and experiments under dry conditions, but our study expands this scope. Through extensive laboratory tests exploring static and cyclic thermal loads in both dry and saturated sand environments, we uncovered valuable insights. Cyclic thermal loads in dry sand demonstrated a significant thermal charging effect, especially with shorter relaxation times. In static thermal loading, utilizing saturated sand enhanced heat dissipation due to higher thermal conductivity. However, it also revealed a noteworthy observation: a robust convection cell formed after three days of continuous heating, presenting challenges for cables under crop fields despite facilitating efficient cooling. Highlighting the importance of high-voltage power cable infrastructure, our study delves into the critical intersection between infrastructure and the underground soil. Understanding these interactions becomes imperative for the sustainable development of clean energy initiatives. As the world transitions to cleaner energy practices, optimizing the performance of underground power cable systems becomes pivotal in realizing their full potential and aligning with broader clean energy goals. This research contributes essential knowledge to enhance the safety, efficiency, and sustainability of high-voltage underground power cable systems in support of a cleaner and more sustainable energy future.

Power transmission covering long-distances has shifted from overhead high voltage cables to underground power cable systems due to numerous failures under severe weather conditions and electromagnetic pollution. The underground power cable systems are limited by the melting point of the insulator around the conductor, which depends on the surrounding soils’ heat transfer capacity or the thermal conductivity. In the past, numerical and theoretical studies have been conducted based on the mechanistic heat and mass transfer model. However, limited experimental evidence has been provided. Therefore, in this study, we performed a series of experiments for static and cyclic thermal loads with a cylindrical heater embedded in the sand. The results suggest thermal charging of the surrounding dry sand and natural convection within the wet sand. A comparison of heat transfer for dry, unsaturated and fully saturated sand is presented with graphs and colour maps which provide valuable information and insight of heat and mass transfer around an underground power cable. Furthermore, the measurements of thermal conductivity against density, moisture and temperature are presented showing positive nonlinear dependence.

Physics-Informed Neural Network (PINN) is a data-driven solver for partial and ordinary differential equations (ODEs/PDEs). It provides a unified framework to address both forward and inverse problems. However, the complexity of the objective function often leads to training failures. This issue is particularly prominent when solving high-frequency and multi-scale problems. We proposed using transfer learning to boost the robustness and convergence of training PINN, starting training from low-frequency problems and gradually approaching high-frequency problems through fine-tuning. Through two case studies, we discovered that transfer learning can effectively train PINNs to approximate solutions from low-frequency problems to high-frequency problems without increasing network parameters. Furthermore, it requires fewer data points and less training time. We compare the PINN results using direct differences and $L_2$ relative error showing the advantage of using transfer learning techniques. We describe our training strategy in detail, including optimizer selection, and suggest guidelines for using transfer learning to train neural networks to solve more complex problems.