Abstract
The current state of urban medium-voltage cable networks is characterised by an increase in electrical loads, which is caused by the growing level of electrification in urban environments, the development of electric transport, the expansion of charging station infrastructure, and the implementation of digital energy technologies. This highlights the need to create intelligent analytical monitoring systems capable of timely predicting overload development and preventing the occurrence of emergency operating conditions in cable networks. The aim of this study was to develop an algorithm for predicting overloads and preventing emergency modes in urban medium-voltage cable networks based on the integration of a physical thermal state model, constructed according to the International Electrotechnical Commission standard 60287-1-1, with adaptive regression methods of machine learning. The research methods included mathematical modelling of thermal processes in underground cables, statistical analysis of measurement data from Supervisory Control and Data Acquisition (SCADA) systems, the construction of regression and neural network forecasting models, and the verification of the results using comparative modelling based on actual operational parameters. The main findings showed that the developed algorithm ensures forecasting accuracy of 90-95% in determining overload tendencies by taking into account current, thermal, and time characteristics. The system is capable of identifying the risk of critical conditions 1.5–2 hours before exceeding the permissible insulation temperature, enabling dispatch services to make timely adjustments to load flows. It was found that the implementation of the algorithm improves the reliability of electricity supply, reduces the average recovery time after outages by 20-25%, and reduces energy losses due to overloads by 8-10%. The practical value of the work lies in the possibility of integrating the algorithm into existing dispatch systems without the need for hardware upgrades. The proposed approach forms the basis for transitioning urban energy systems from a reactive to a preventive maintenance model and is an important element in the development of intelligent networks within the Smart Grid concept