Predicting Wasteful Energy Consumption via a LSTM Recurrent Neural Network

According to figures from the State of California Department of Finance, the State of California is projected to reach a population of 45 million by the 2030s. Meanwhile, whether at home or at work, technology will continue to become a more integral part of everyday life. As a result, loads on energy grids will worsen and the increase in population in the United States’ center of technology will amplify this effect. Especially after switching to time-of-use (TOU) pricing, energy service providers may be forced to charge customers, including residential homes, higher prices for energy use during peak hours due to an increase in energy demand and production. Its impact on the environment would involve the potential rise in carbon emissions depending on the availability of renewable energy sources. For both residential and business consumers, it may be difficult to determine specific areas or devices for which to charge usage patterns for financial and environmental benefit as energy bills may only provide a summary of total unit energy usage. Nowadays, energy efficiency has become a popular topic that attracts scientists' attention. How can we reduce wasteful energy consumption? How to maximize energy usage efficiency? Besides creating more clean energy, we need to figure out a way on the user-side as well to improve the energy utilization rate.

Our goal is to reduce wasteful energy consumption on plug loads through predictive modeling and provide an actionable user experience to better manage their energy consumption. By using the Long Short Term Memory RNN model, we can predict users' energy usage patterns base on usage history. The model will automatically turn on/off the electronic devices if they are not in actual use. This solution will ultimately give users better control of their energy usage and costs, minimize the load on the grid, and promote the wellbeing of the environment.