K-Means Algorithm Implementation for IoT-Based Early Fire Detection in Oil Palm Plantations

Abstract

Oil palm plantation fires continue to be a significant problem, significantly impacting the environment, public health, and economic activity. By combining the K-Means algorithm, processed directly on an ESP32 microcontroller, with an Internet of Things (IoT)-based early detection system, this research has produced an innovation that does not require an external server. To monitor hazardous gases, smoke, and temperature, the system uses thermocouples and MQ-2 and MQ-135 sensors. Conditions are then categorized into Safe, Alert, and Fire. Using 15 test data samples, the evaluation was conducted in the field, specifically in the oil palm plantation area in Banyuasin, South Sumatra. The test results showed that the classification had 100% accuracy. However, the limited amount of data was one of the obstacles to this study, so additional testing is needed to ensure the accuracy of the large-scale study. This system is suitable for remote and limited infrastructure, helping to develop effective and responsive early fire detection technology.