Model Creation Method for Anomaly Detection in Refrigeration and Air Conditioning Systems

Predicting anomalies and diagnosing various machines using IoT technology have been widely studied. Refrigeration and air-conditioning systems are among them. Anomaly detection systems for refrigeration and air-conditioning systems often work with many distributed devices, posing a challenge in creating a learning model. Additionally, the data collection period to create learning models needs to be shortened. In this study, we propose a method to automate anomaly detection model creation for a large number of distributed refrigeration and air-conditioning systems and to improve the efficiency of the operation of the diagnosis system. In addition, we propose a fine-tuning method that creates a learning model for the target device with minimal learning data based on a learning model created by another device, thus reducing the learning model creation period.