Author(s):
1. Jovan Škundrić, Faculty of Mechanical Engineering, Republic of Srpska, Bosnia and Herzegovina
2. Indir Mujanić, Elektrane Stanari d.o.o., Republic of Srpska, Bosnia and Herzegovina
3. Darko Knežević, Faculty of Mechanical Engineering, Republic of Srpska, Bosnia and Herzegovina
4. Saša Laloš, Faculty of Mechanical Engineering, Republic of Srpska, Bosnia and Herzegovina
5. Marko Lazarević, Elektrane Stanari d.o.o., Serbia
6. Danilo Đurica, Elektrane Stanari d.o.o., Republic of Srpska, Bosnia and Herzegovina

Abstract:
For the reliable operation of any complex system, including large-scale thermal power plants, real-time monitoring of operating parameters is essential. A high-quality and efficient monitoring system enables timely intervention when any of the parameters begins to significantly deviate from the optimal or calculated value. However, there are situations in which all parameters remain within acceptable limits, yet the plant does not operate entirely properly. In such cases, it is often difficult to detect and localize the problem, which typically manifests only through reduced efficiency or compromised operational stability. The application of machine learning methods, such as linear regression models, can be a valuable tool under such conditions. The core idea is that a well-trained regression model can timely detect and indicate even subtle anomalies in the operation of individual subsystems. This paper presents a possible approach to modeling the behavior of a properly functioning condensation system within a large thermal power plant, based on available process data. The study was conducted using data from the Stanari Thermal Power Plant, and linear regression was applied as the selected machine learning model.

Key words:
Thermal Power Plant,Fault Detection,Machine Learning,Linear Regression

Date of abstract submission:
25.07.2025.

Conference:
Contemporary Materials 2025 - Savremeni Materijali