Integrated Modeling of Condition Monitoring Data for Predictive Maintenance of Electrical Power Plant Systems

Abstract

This study examined the effectiveness of integrated modeling of condition monitoring data for predictive maintenance in electrical power plant systems using a quantitative, quasi-experimental research design. The analysis was conducted on a dataset comprising 12,480 observations collected over an 18-month monitoring period across turbines, generators, transformers, and auxiliary systems. The study integrated multiple condition monitoring variables, including vibration, temperature, and electrical load, to develop predictive models using machine learning techniques. The results demonstrated that integrated models significantly outperformed single-source models, achieving a classification accuracy of 92.6% compared to 81.4%. Precision and recall values also improved, reaching 91.8% and 93.4%, respectively, indicating enhanced fault detection capability. Regression analysis revealed a substantial reduction in prediction error, with root mean square error decreasing from 0.58 to 0.42 and mean absolute error from 0.45 to 0.31. Additionally, integrated models identified faults approximately 36 hours in advance compared to 22 hours for single-source models, representing a 63.6% improvement in early detection capability. Component-level analysis indicated that turbines and generators achieved the highest predictive accuracy at 94.2% and 92.8%, respectively, while auxiliary systems showed comparatively lower accuracy at 87.3% due to higher operational variability. Statistical testing confirmed that these improvements were significant at a probability level below 0.05, with large effect sizes observed across key performance metrics. The findings also highlighted the importance of data quality, feature engineering, and model optimization in achieving reliable predictive outcomes. Overall, the study demonstrated that integrated condition monitoring significantly enhances predictive maintenance performance by improving accuracy, reducing uncertainty, and enabling proactive maintenance planning. These results provide strong empirical support for the adoption of integrated data-driven maintenance strategies in electrical power plant systems and contribute to the advancement of intelligent industrial maintenance practices.