A comparative study of predicting the availability of power line communication nodes using machine learning

Area of Science:

• Electrical Engineering
• Computer Science
• Data Science

Background:

• Power Line Communication (PLC) technology transmits data over existing power cables.
• Predictive maintenance for PLC nodes can save significant time and resources by avoiding manual testing.
• Identifying optimal node performance beforehand is crucial for network reliability.

Purpose of the Study:

• To develop and evaluate a predictive model for determining the operational status of PLC nodes.
• To compare the performance of various machine learning algorithms for PLC node health assessment.
• To identify key features influencing PLC node performance.

Main Methods:

• A dataset of 1000 PLC node readings was utilized, with 90% for training and 10% for testing.
• The model was trained using dominant features: Signal to Noise Ratio (SNR), Received Signal Strength Indicator (RSSI), and Carrier to Interference plus Noise Ratio (CINR).
• Algorithms including Multi-layer Perception, K-Nearest Neighbors, Support Vector Machine (SVM), Random Forest, and Adaptive Boosting (ADA) were evaluated.

Main Results:

• Adaptive Boosting (ADA) demonstrated superior performance among the evaluated algorithms.
• ADA achieved an accuracy of 87%, an F-score of 0.86613, precision of 0.9, and recall of 0.8646.
• The model effectively distinguished between optimum and non-optimum PLC node readings.

Conclusions:

• The developed predictive model, particularly using Adaptive Boosting, is effective for assessing PLC node performance.
• Feature selection (SNR, RSSI, CINR) is critical for accurate PLC node status prediction.
• This approach offers a viable method for proactive maintenance in PLC networks.