Adaptive FEM-BPNN model for predicting underground cable temperature considering varied soil composition
Abdullah Ahmed Al-Dulaimi,Muhammet Tahir Guneser,Alaa Ali Hameed,Fausto Pedro García Márquez,Osama E. Gouda
DOI: https://doi.org/10.1016/j.jestch.2024.101658
IF: 5.155
2024-03-10
Engineering Science and Technology an International Journal
Abstract:In underground cables of power systems, the maximum temperature of the cable is a crucial factor in determining its capacity. According to standards, the permissible operating temperature for the XLPE cable conductor under steady-state conditions is 90 °C – a limit that should not be exceeded. Exceeding this temperature may lead to a thermal breakdown in the cable insulation, thereby resulting in interruption of the electrical power supply. Many factors affect the cable temperature, particularly through the processes of heat dissipation and diffusion from the cable into its surroundings. These factors include soil types and compositions, cable installation configuration, and thermo physical properties; therefore, accurate analysis of these factors is crucial for cable loading. In this study, the finite element method (FEM) is employed to predict the cable temperature considering different soil compositions and to present a new approach for the thermal analysis of an underground cable system. The novel approach considers various environmental conditions including single-layer and multi-layer soil types, homogeneous and non-homogeneous soil compositions, two configuration types – flat and trefoil – as well as two types of backfill materials, specifically sand-cement mixture backfill (SCMB) and fluidized thermal backfill (FTB), and dry zones to offer deeper insight into a thermal analysis. Given that the FEM requires the construction of a complex geometric model within an optimal operating condition to obtain results with high accuracy—a process that can often be complex as well as not adaptable because it depends on constant mathematical calculation—This paper presents a novel approach FEM-BPNN that uses an adaptive Backpropagation neural networks (BPNN) model as its mainstay. The proposed BPNN model exploits historical data from FEM to refine its predictive power, therefore, increasing its efficiency and accuracy. Furthermore, the model is subject to an optimization process, adjusting and refining its internal parameters in response to new data, with the ultimate goal of improving the predictive model capabilities for the temperature of underground power cables. The results underscored the high performance of FEM in the simulation, and it was observed that FEM yielded results closely aligned with those of the IEC standard. Moreover, the proposed FEM-BPNN demonstrated exceptional accuracy, achieving a low RMSE score of 0.008. It also exhibited impressive performance in the linear regression analysis, with an R2 value of 0.99. These metrics collectively signify the robustness and efficacy of the model.
engineering, multidisciplinary