Date of Award

5-16-2025

Publication Type

Thesis

Degree Name

M.Sc.

Department

Electrical and Computer Engineering

Keywords

Combined Cycle Power Plant (CCPP); Machine Learning; Artifical Neural network; Support Vector Regression

Supervisor

Mohammad Hassanzadeh

Supervisor

Majid Ahmadi

Rights

info:eu-repo/semantics/embargoedAccess

Abstract

This thesis presents a comprehensive, data-driven framework for enhancing the monitoring and operational efficiency of Combined Cycle Power Plants (CCPPs) through the integration of advanced machine learning and deep learning techniques. The research encompasses three major components: power output forecasting, classification, and anomaly detection. In the first phase, multiple regression models—including Random Forest, Support Vector Regression, and deep neural networks—were evaluated for predicting power output using environmental and operational parameters. The Artificial Neural Network (ANN) model achieved the highest accuracy, demonstrating its ability to capture complex nonlinear relationships. The second phase introduced classification schemes that categorize power output into discrete levels using quartile-based and equal-bandwidth approaches. Both traditional machine learning models (Random Forest, LightGBM, Gradient Boosting) and a Convolutional Neural Network (CNN) were employed, with CNNs delivering competitive performance and strong generalization capabilities. In the final phase, an Anomaly Detection system using Autoencoders system was developed to identify operational faults and performance deviations. Simulated anomalies were injected into the dataset, and detection thresholds were optimized using the F2 score to prioritize recall. The proposed method achieved a 100% anomaly detection rate with minimal false positives, outperforming conventional techniques. Collectively, the results demonstrate the potential of combining predictive modeling, classification, and anomaly detection to support real-time decision-making, improve reliability, and facilitate predictive maintenance in CCPP operations.

Available for download on Friday, May 15, 2026

Share

COinS