MQTT based Time-Series Anomaly Detection System for Smart Grid

Abstract

Developing smart grid system is a crucial thing and relies heavily on modern technology. Powered by cloud computing and artificial intelligence, electrical grid can make a generational leap towards a more efficient and smart future. Meanwhile, machine learning can get great result in fields like anomaly detection and load forecasting, However, power grid equipment is usually quite underpowered, can't perform deep learning algorithm locally. In this paper, we proposed a new stream processing architecture based on MQTT protocol, therefore time-series anomaly detection algorithms running on the cloud. Allowing any anomaly event can be detected fast and accurately.