Energy Consumption Prediction System Based on Deep Learning with Edge Computing

Governments around the world are actively pursuing research on smart cities as an attempt to pursue the city's sophistication and sustainability. In order to create a smart city, it is important to reduce energy and manage it efficiently. With the development of smart city, it is expected that a large number of Internet of Things (IoT) devices will be installed in various buildings in the city, and a lot of energy usage information will be measured. When many IoT devices are connected to the internet, a large amount of data is generated. Therefore, when the existing cloud service is used, data is delayed in analyzing and transmitting data, and it becomes impossible to receive the analysis result quickly. With edge computing, you can process data directly in the edge environment where data is collected and apply the results quickly to the field. It can respond to data much faster than it can deliver data and wait for the results to be analyzed and reduce the load on the data. Therefore, in this paper, we propose a system for predicting energy consumption using a deep learning (DL) algorithm in an edge computing environment. We have applied the proposed system to office environment by building testbed. We used the long short-term memory (LSTM) network which shows high accuracy in time series data analysis and obtained the energy prediction result per day. We check the state of the electronics through the energy consumption prediction data, detect the abnormality and provide the energy consumption pattern analysis service respectively.