Predictive Maintenance for Al Sabiya Power Plant Using Machine Learning Algorithms

Abstract

This study develops a predictive maintenance framework for the Al Sabia steam power plant in Kuwait, employing Support Vector Machine (SVM) and K-nearest Neighbor (KNN) algorithms. This research focuses on anticipating maintenance needs based on critical operational parameters, including temperature, pressure, flow rate, operational hours, and alert signals. Experimental results indicate that SVM outperforms KNN, achieving an accuracy of 0.95 compared to 0.93 for KNN, along with superior precision, recall, and F1-score, suggesting its suitability for this application. Furthermore, an ensemble model SVM and KNN achieves an accuracy of 0.93. The adoption of this model is expected to markedly reduce downtime, improve storage quality, and enhance overall power plant reliability. Additionally, this paper provides a comparative analysis of a neural network model developed in TensorFlow and its equivalent model implemented in TensorFlow Lite. The analysis evaluates both models on three key performance metrics: accuracy, sample size, and latency. Both the TensorFlow and TensorFlow Lite models attain an accuracy of 0.95, affirming TensorFlow Lite's efficacy in facilitating high-performance machine learning on resource-constrained hardware.