AI-Based IoT System for Temperature Forecasting

Server rooms are critical elements of the IT infrastructure in many industries, and the stability and performance of servers directly depend on maintaining optimal temperature conditions. Temperature increases can lead to hardware failures, reduced performance, and even complete system shutdowns. Most existing server room temperature monitoring solutions are reactive, alerting staff only when the temperature has reached an undesirable level. In contrast to traditional reactive monitoring systems, we propose a system that implements a proactive approach, which includes actively predicting and preventing potential problems and failures before they occur. Through proactive monitoring, the system can identify and minimize risks, optimize system performance, and provide more stable and efficient service. The system uses an integrated approach that combines the Internet of Things (IoT) for data collection and artificial intelligence (AI) for data analysis, allowing for real-time monitoring of conditions and accurate prediction of future state. An important part of the system is an LSTM-based fault detection model trained on 5 years of historical temperature data from the server room of a Ukrainian telecommunications operator. The model demonstrates high accuracy in prediction, with a root mean square error (RMSE) value of approximately 0.53, which corresponds to an acceptable error of 1 degree. This accuracy ensures reliable anomaly detection and prevents potential failures. The prototype is integrated with the Zabbix monitoring system and interacts with the Telegram Bot API to send notifications and control the system. This ensures that operators are quickly informed of any temperature anomalies, enabling them to respond to problems immediately.