Dr. Mileta Žarković

Due to the rapid digitalization of the energy sector, the amount of data actively monitored and stored in power systems is increasing daily. Various parameters, including meteorological, chemical, thermal, mechanical, and electromagnetic factors, influence this data. In the future, power system experts and engineers will require AI-based support tools that will provide them with a clearer insight into all available data and facilitate decision-making in various planning processes. This lecture explores the application of fuzzy logic for the experiential formation of expert systems in diagnosing the condition of power equipment. Unsupervised machine learning algorithms are employed for clustering data related to power equipment to develop an optimal maintenance plan. The use of artificial neural networks for detecting accelerated aging of power equipment is also discussed. Additionally, the lecture presents the methodology of using autoencoder neural networks to identify anomalies in power systems. Based on such results, risk maps can be utilized to transition from Time-Based Maintenance (TBM) to Predictive Condition-Based Maintenance (CBM) of power equipment. The proposed methodology enables engineers to make well-informed and timely maintenance decisions within the power system. A key takeaway from the lecture is that the fundamental advantage of AI application lies in its ability to learn dependencies between monitoring parameters of critical power system equipment. The implementation of these methods is demonstrated through the Monitoring and Diagnostic Center (MDC) for assessing the condition of generators and power transformers, which are crucial components of the power system. The MDC aims to support maintenance planning for key power equipment in AD EPS and conduct regular diagnostics and condition assessments. Practical examples indicate that existing power system databases serve as “living” knowledge bases, from which timely conclusions and decisions can be derived in the future. Ultimately, applying AI in diagnosing the condition of power equipment leads to increased reliability, security of energy supply, and improved energy efficiency.

Keywords: Power Equipment Diagnostics, Artificial Intelligence, Fuzzy Logic, Autoencoders, Predictive Maintenance