Modelling and Simulation of Residential Load Profiles as an Approach for Data-Driven Prediction

Rapid growth of buildings energy consumption encourages to take measures to improve energy efficiency by actors involved in the field. One of the approaches developed last decades consists in energy management through energy prediction. These approaches engage machine learning algorithms, which focus on predicting energy consumption based on past-observed data. But there are also cases when this information is missing so in this paper, we focus on solving the problem when measured data are not available. Initially, we develop an electrical home appliance simulator, which reflects their energy consumption and occupant behavior. Each of the considered device is modelled using an electrical circuit analogy. Then aggregating single appliance energy consumption from simulator, total power consumption data is generated. Synthetic data are feed to an Artificial Neural Network algorithm to learn consumption pattern and to predict next hour energy consumption.