Using Bayesian Networks for the Prediction of Domestic Appliances Power Consumptions

In this paper, we deal with the short term prediction of the domestic electrical appliances power consumption. We go further than the prediction of on/off activities of appliances to the prediction of a bounded interval of consumed power. The short term prediction of appliances consumptions is an unavoidable problem to solve when constructing optimal schedules of appliance consumptions. For reliable predictions, the appliances inter-dependencies and the cyclic characteristic of appliances activities should definitely be considered. Short term prediction is performed in the current work by graphical modeling dependencies between appliances use in a particular period of time. Obtained dependencies and short time history of appliances states are then used to infer future appliances states. We demonstrate through testing on real-world data that our model provides a promising result, nearly 90% of correct rate of classification.