Intelligent Automation For Real-Time Monitoring And Stability Control Of Power Systems

Abstract

Power system operation has been fundamentally transformed by the integration of artificial intelligence (AI) into 
traditional approaches to fault detection, isolation, and control. The objective of this empirical research is to 
evaluate the performance of AI-enabled open-source automation frameworks namely, deep learning (DL), 
machine learning (ML), and convolutional neural network (CNN)-based systems, in the context of real-time fault 
detection in high-voltage transmission and distribution networks. Twelve thousand four hundred fault event 
records from five regional power utilities across India spanning the duration of 2019–2023 were used to evaluate 
and compare the performance of multiple AI models with conventional relay-based protection systems. 
Quantitative methods such as confusion matrix analysis, receiver operating characteristic (ROC) curves, and 
statistical hypothesis testing are used for evaluating model accuracy, precision, recall, F1-factor, and fault 
clearance time. The results show that the proposed hybrid CNN-LSTM model achieves a fault detection rate of 
98.7% in comparison with Support Vector Machine (SVM) being 94.3%, Random Forest (RF) 95.8% and typical 
systems achieving 87.4%. The critical analysis demonstrates how AI-enabled systems improve average fault 
clearance duration by 62%, as opposed to traditional protection schemes, resulting in considerably enhanced 
system reliability indices such as SAIFI and SAIDI. The results justify the deployment of AI-based automation in 
smart grid infrastructure for Indian utilities rapidly digitalizing under the National Smart Grid Mission (NSGM).