Forecasting Device Specific Energy Consumption from Wireless Network Traffic in Smart Home Networks

Abstract

This thesis suggests a method to forecast the energy consumption of devices from the wireless network traffic in smart home networks. Device specific wireless network packets are captured and the corresponding electrical energy consumption is measured in an experimental smart home network setting. The correlation between wireless network traffic and energy consumption is examined by applying multiple linear regression. The wireless network traffic is forecasted with the Autoregressive Integrated Moving Average (ARIMA) model, and energy consumption is derived from the forecasted wireless network traffic. We compare the feasibility of this indirect method to a more direct method of building a forecasting model with energy consumption itself. We evaluate the methods using root mean squared error (RMSE) and misclassification rate (MCR).