Automated Load Identification and Consumption Prediction for Smart Energy Meters

Smart Energy Meters allow consumers to be aware of their consumption in real time and make smarter decisions based on this information. Due to the pandemic, the energy consumption for every household has continued to rise as people are being forced to stay at home. With this, the need to increase awareness of one's own consumption has also risen. The purpose of this study is to present a system capable of presenting essential appliance information in an easily accessible and understandable form. More specifically, this system is capable of electricity consumption prediction alongside precise automatic load detection that adapts to the user's needs. The system gathers data based on specific appliances used by a user with the intention of comparing data per device. The said system is scalable based on the number of users and implemented using a SONOFF POW R2, a short extension outlet, a virtual machine, and present output features through a mobile phone. The results of the study show that the created models with an accuracy of 64% for the UK-DALE dataset, 97 % for the first model that used the power readings in the created dataset and 77 % for the second model which utilized an implementation of PCA in the created dataset, are capable of identifying appliances connected to the smart meter and giving a consumption prediction.