Anomalous Electricity Consumption Detection

Have you ever thought about the scenario to get an in-time alert about your electric energy usage? Residential and commercial buildings consume around 40% of the total energy generated in the US. There have been developments to mitigate these inefficiencies but they are still very basic and there is a lot of room for improvement. The current energy consumption market lacks dynamic control systems to manage consumption, and existing benchmarking tools are too basic. Inefficient consumption habit in urban life has worsened this scenario.

In UC Berkeley Engineering School IEOR Department, a team of 5 students used high-frequency electricity consumption data from houses to predict energy consumption and detect anomalous consumption for households through the means of training energy usage dataset with rolling ARIMA, SARIMA (seasonal autoregressive integrated moving average) and RNN. This project would enable the households to understand their electricity consumption behavior and to be aware of unexpected power consumption values. Thus, households can perform energy-saving actions. With this project, they believed effective energy consumption feedback mechanism can reduce energy consumption, help economic savings and achieve sustainability.