The “Machine Learning-Based Three-Phase Smart Power Monitoring and Electrical Anomaly Detection System for Industry” is an innovation designed to monitor three-phase electrical power distribution in industrial environments. It was developed to overcome the limitations of manual electrical inspections, which fail to provide continuous parameter recording or the automatic identification of deviations in operational patterns.
The system utilizes three PZEM-004T V3.0 sensor modules to measure electrical parameters—such as voltage, current, active power, power factor, frequency, and electrical energy—across the R, S, and T phases. Data processing is handled by two ESP32-S3 microcontrollers with distinct functions: the first ESP32-S3 manages data acquisition and validation, while the second performs feature processing and local anomaly detection. Both devices communicate via the ESP-NOW protocol.
A key feature of this innovation is the implementation of the Isolation Forest machine learning algorithm, which identifies unusual electrical patterns by analyzing the interrelationships between multiple parameters simultaneously. The system employs a feature engineering approach using three configurations: raw measurement features, derived features (based on inter-phase relationships), and temporal features (based on changes over time). This approach allows for the evaluation of the system’s detection capabilities across various electrical characteristics.
Model training is conducted using Python, while inference is designed to run directly on the ESP32-S3 via an edge machine learning approach. Consequently, the anomaly identification process does not rely on an external server. The system generates an “anomaly score” and classifies conditions as either normal or anomalous; this data can be transmitted via a Wi-Fi network to be stored and visualized on a web-based dashboard. Development was completed on October 3, 2026, taking a total of 10 days with a project cost of IDR 2,000,000. The innovation team was led by Vania Rapita Augesta, with Irfan Ashari and Muhammad Syauqi Azis Fikri Amrulloh serving as team members.
This innovation is intended for implementation at the Sub Distribution Panel (SDP) for Line 14 at PT Roda Pasifik Mandiri, serving as a tool for electrical monitoring and maintenance support. Through this system, operators can access historical measurement data, monitor changes in electrical conditions, and identify signs of irregularities early, allowing for timely follow-up via technical inspections.
The development of this innovation aims to improve the availability of information regarding electrical conditions and support data-driven maintenance. It offers an alternative, intelligent monitoring system that leverages the Internet of Things (IoT), machine learning, and embedded computing devices—all at a relatively affordable cost. The system functions as a monitoring and anomaly-detection tool; it is not intended to replace electrical protection equipment or serve as a direct diagnostic tool for physical hardware faults. This innovation was also showcased at the Capstone EXPO held on November 11, 2026, in the E11 building of the Faculty of Engineering at UNNES.
