Prototyping and Demonstration of a Cloud-based AI Smart Plug System

In this work, a cloud-based smart plug system integrated with artificial intelligence (AI) technology is prototyped and demonstrated. It exploits extra advanced information from plugged house appliances, compared to conventional smart plug products. An Internet of Things (IoT) smart plug printed circuit board (PCB), which is small enough to fit into a household plug outlet, is designed in-house. There are two main components connected to an ESP8266 microcontroller: a relay to control the power outlet and a non-intrusive current sensor to monitor power load. To take real-world application into consideration, the software is built on Google Cloud Platform (GCP) with event-driven architecture (EDA). Such architecture provides scalability, increases agility, and reduces cost. IoT devices are communicated through MQTT protocol on WiFi. Machine learning (ML) techniques are used to process the collected power load data. Thus, several different pieces of information about the plugged appliances can be acquired, such as malfunction detection, operation status identification, and completion time estimation. Edge AI technique is also experimented by executing the trained neural network (NN) model on a STM32F7 microcontroller. Finally, user interface (UI) of the system can provide functions including remote control and notification for the information of plugged appliances.