Realtime Forecasting for Smart Agriculture System using Simple Moving Average Method

Abstract

This study proposes a real-time smart agriculture monitoring system that integrates Internet of Things (IoT) technology with a SMA  forecasting method. The system utilizes soil moisture, temperature–humidity, and light intensity sensors connected to an ESP32 microcontroller to monitor environmental conditions in agricultural fields. Sensor data are transmitted to a cloud database and visualized through a mobile application to support remote monitoring. This study obtains reliable environmental measurements during field testing in a rice field environment. The system harvests temperature values ranging from 27.5 °C to 35.6 °C, relative humidity between 48.1% and 69.4%, and light intensity values between 21,580 and 54,612 lux, while soil moisture measurements consistently reflect variations between dry and wet soil conditions. This study also implements a lightweight forecasting module using the Simple Moving Average method with a bounded trend adjustment to provide short-term environmental predictions. We obtain stable prediction performance using a moving window of ten observations, which effectively smooths sensor noise while maintaining responsiveness to environmental changes. Forecasting evaluation using the MAPE indicates low prediction error values, demonstrating the reliability of the proposed prediction mechanism. By executing the forecasting module externally from the embedded ESP32 device, the system maintains efficient real-time monitoring while still delivering predictive insights to support early agricultural decision-making. These findings confirm that the proposed IoT monitoring system provides reliable environmental measurements while offering additional advantages.