Enhanced Dynamic Line Rating Forecasting Under Cyber-Attacks via Incremental Learning Frameworks

Dynamic line rating (DLR) is essential for optimizing the efficiency and reliability of overhead transmission lines (OTLs), as it provides real-time insights into line capacity influenced by environmental and climatic factors. DLR forecasting is a method designed to predict the maximum capacity of OTLs accurately. Numerous solutions have emerged for DLR forecasting, but many face challenges such as the need for multiple sensors, complex measurement systems, precise calibration, and vulnerability to cyberattacks exist, which can lead to poor operational decisions. To overcome these challenges, this paper introduces an innovative cyber-secure online DLR forecasting method known as incremental learning (IL). Several scenarios are developed to evaluate the robustness and accuracy of this methodology using real-world data from an OTL in Khaf, Iran, both in the presence and absence of cyber-attacks. A comparative analysis was conducted with established techniques, including extreme learning machine (ELM), bidirectional long short-term memory (BiLSTM), and support vector regression with LSTM (SVR-LSTM). The results indicate that the proposed IL approach delivers exceptional performance even in the face of cyber threats, accurately forecasting DLR values with minimal error and surpassing existing advanced methods.