Appliances Energy Prediction Using Random Forest Classifier

Abstract

The objective of this work is to simplify the decision-making process that energy providers must go through in order to decide whether or not to provide a range of residential buildings with energy depending on the demand for that energy. Using the Multi-layer Perceptron model and the Random Forest model, respectively, we classify residential structures in this article according to the amount of energy that they use.

In order to keep track of the temperature and humidity levels within the house, a network of wireless sensors powered by ZigBee was used. There was an interval of about 3.3 minutes between the transmission of the temperature and humidity measurements by each wireless node. After that, an average was calculated using the wireless data based on intervals of ten minutes. Every ten minutes, the data on the energy consumption was recorded using m-bus energy meters. Using the date and time column, the weather information from the weather station at the closest airport, which was Chèvres Airport in Belgium, was obtained from Reliable Prognosis (rp5.ru).

The retrieval of the data, the extraction of the features, and the prediction are the three processes that make up the prediction. During the data retrieval process, the database is queried to obtain the data that has been utilized hourly and on a daily basis. Statistical characteristics such as "mean," "standard deviation," "skewness," and "kurtosis" are derived from the data that was obtained through the use of feature extraction. At this point in the process, the prediction stage, Multi-layer Perceptron and Random Forest have been used to anticipate which buildings would have high or low electricity usage.