Energy usage and human behavior modeling for residential bottom-up energy simulation

We introduce a probabilistic modeling for a disaggregated Bottom-up simulation of residential energy usage. Parametric probability distributions are modeled with parameters that have a natural explanation in terms of usage and appliance power. Human behavior such as sleep and home occupancy variables are considered too, with its corresponding trained probabilistic Models. Model parameters are adjusted by the minimization of the Kullback-Leibler divergence from known appliance and behavior usage data. Self-generated photovoltaic Energy is included in the simulation with a battery for storage and electric vehicle usage. Simulations match individual and aggregated usage load profiles in the European REMODECE and RSE Italian load data sets. Obtained Models are useful for residential disaggregated sim-ulations allowing individual appliances to change from house to house. Probabilistic distributions can be used as prior knowledge for energy management systems, risk management, and grid failure prediction and can be adapted based on non-stationary real-time house behavior and appliance usage.(c) 2022 Elsevier B.V. All rights reserved.

Automated summary

This paper presents a probabilistic modeling approach for simulating residential energy usage at a disaggregated level. Parametric probability distributions are used to model appliance power and human behavior variables, which are adjusted based on known data. The simulation also incorporates self-generated photovoltaic energy and electric vehicle usage. The models obtained from this approach can be used in residential disaggregated simulations, allowing for individual appliances to vary across houses. The probabilistic distributions can provide prior knowledge for energy management systems, risk management, and grid failure prediction, and can be adapted based on real-time changes in house behavior and appliance usage.