Prediction of Energy Consumption in a Smart Home Using Deepened K-Means Clustering ARIMA Model

Abstract

In this technological era everything is made to act smart. So is a home. A smart home is the place where all the devices are designed to act smartly and it can be programmed in such a way that the maximum benefit is extracted out of it. This helps the mankind in multiple ways and one of the majoradvantages of smart devices in a home is the management of electric energy. When energy management is done in an efficient way, it helps in reducing the price to be paid and scarcity of energy can be avoided at the time of outages. To efficiently manage the energy, forecasting plays a vital role. Based on the forecast, energy production can be planned ahead and energy consumption by various smart devices can also be handled in a better way. In this paper a new model Deepened K-Means Clustering ARIMA (DKMCA) is proposed to predict the amount of energy consumed in a smart home from a smart grid during different seasons in a year. This proposed model removes the ambiguity in the K-Means clustering algorithm and from the
clusters obtained, the forecasting model is built using the Model-Based forecasting method ARIMA. The data from a single smart home from the Pecan Project, Texas, USA is taken for this work. The average amount of energy consumed by each smart device in a month from the smart grid for seven years is taken and prediction is done on the average amount of energy that will be consumed from the smart grid during various seasons in a year. The performance of the proposed model (DKMCA) is compared with the ARIMA model. From the result obtained it is found that the RMSE, MAPE, AIC,AICC and MAE of the proposed model is less compared to the ARIMA model. The loglikelihood of the proposed model is also high compared to the ARIMA model. Hence the accuracy of the proposed model is better compared to the standard Model-Based method ARIMA.