Investigating Weather Inputs for Neural Networks in Home Energy Management Systems

The increase in energy demand within smart homes has generated corresponding demand for smart grids and home energy management systems (HEMSs) aimed at helping consumers construct efficient energy use plans and electricity cost savings. Neural Networks (NNs) are being implemented to enable efficient HEMS operation. Examining the effects of weather on a data set that is used for an applied NN in a HEMS could help to optimize energy efficiency. This study finds that information from weather forecasting can play a significant role in designing NNs to support energy management systems (EMSs). Accurate modeling of the energy demand is critical for a HEMS to be effective. Foregoing weather data inputs altogether when designing NNs leads to less specific outputs, resulting in less efficient network response. And, although reliable forecasting provides many benefits when managing energy distribution, NN algorithms must account for forecasting errors if network inputs require more than 24-hour advance notice. NNs for HEMSs forecasting require weather inputs to accurately predict energy demand by accounting for substantially different and continually changing environmental factors.