Realistic Multi-Scale Modeling of Household Electricity Behaviors

• Published in: IEEE access 2019, Vol.7, p.2467-2489
• Main Authors: Bottaccioli, Lorenzo, Di Cataldo, Santa, Acquaviva, Andrea, Patti, Edoardo
• Format: Article
• Language: English
• Subjects: behavioral modeling; Computational modeling; Electric appliances; Electric power distribution; Energy resolution; Home appliances; Household appliances; Household load profile; Households; Load; Load modeling; Markov processes; Model accuracy; Monte Carlo; Network reliability; Non Homogeneous Semi-Markov Model; Populations; Scale models; Simulation; Time dependence; time use survey; use of energy; Water heating
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2018.2886201

To improve the management and reliability of power distribution networks, there is a strong demand for models simulating energy loads in a realistic way. In this paper, we present a novel multi-scale model to generate realistic residential load profiles at different spatial-temporal resolutions. By taking advantage of the information from census and national surveys, we generate statistically consistent populations of heterogeneous families with their respective appliances. Exploiting a bottom-up approach based on Monte Carlo Non-Homogeneous Semi-Markov, we provide household end-user behaviors and realistic households load profiles on a daily as well as on a weekly basis, for weekdays and weekends. The proposed approach overcomes the limitations of the state-of-the-art solutions that consider neither the time-dependency of the probability of performing specific activities in a house, nor their duration or are limited in the type of probability distributions they can model. On top of that, it provides outcomes that are not limited to a per-day basis. The range of available space and time resolutions span from single household to district and from second to year, respectively, featuring multi-level aggregation of the simulation outcomes. To demonstrate the accuracy of our model, we present experimental results obtained by simulating realistic populations in a period covering a whole calendar year and analyze our model's outcome at different scales. Then, we compare such results with three different data-sets that provide real load consumption at the household, national, and European levels, respectively.