Year: 2018
Type: article
Abstract: summary:In our paper we investigate the applicability of independent and identically distributed random sequences, first order Markov and higher order Markov chains as well as semi-Markov processes for bottom-up electricity load modeling. We use appliance time series from publicly available data sets containing fine grained power measurements. The comparison of models are based on metrics which are supposed to be important in power systems like Load Factor, Loss of Load Probability. Furthermore, we characterize the interdependence structure of the models with autocorrelation function as well. The aim of the investigation is to choose the most appropriate and the most parsimonious models for Smart Grid simulation purposes and applications like Demand Side Management and load scheduling. According to our results most appliance types can be modeled adequately with two states (on/off model) and the semi-Markov process can reproduce the properties of an aggregate load well compared to the original time series. With the price of more parameters of the semi-Markov model compared to identically distributed random sequence and first order Markov chain, it gives better results when the autocorrelation function, Loss of Load Probability and Load Factor are considered.
Source: Kybernetika
Authors: Rajmund Drenyovszki, Lóránt Kovács, Kálmán Tornai, András Oláh, István Pintér
Language: English
Topic: Smart Grid Energy Management
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2783194331