LoadProfileGenerator: An Agent-Based Behavior Simulation for Generating Residential Load Profiles

Abstract

A load profile contains a value for each step in time that describes how much energy was consumed in that particular time step. There are load profiles for example for electricity, gas, heat demand and other load types. This is critical input data for both energy system planners and researchers. For instance, car charging load profiles have a very strong influence on the self-consumption of a photovoltaic system. (N. D. Pflugradt & Muntwyler, 2019). This paper describes the LoadProfileGenerator (LPG), an application for creating synthetic residential load profiles. It uses a desire-driven agent simulation to model the behavior of the residents in detail and generate load profiles with the high temporal resolution of 1 minute for residential energy consumption, primarily electricity and domestic hot water. Additionally, it generates behavioral data, travel and locational data. The LPG features 60 predefined households, validated for Germany. These households are fully customizable and can be defined by specifying the people residing in them and their traits. Examples of traits are "alarm at 7am", "sleeps 8h" or "works at an office from 9-5". There are over 400 predefined traits and more can be added by users. To enable the modeling of a large number of households, the LPG contains functions for automatically creating new household definitions on the basis of traits. The LPG is written in C# and comprises about 60.000 lines of code. Unit test code coverage is greater than 70%. The LPG does not currently contain a detailed heating demand simulation, as that field is already covered in detail by other tools such as TRNSYS, Polysun, tsib (Kotzur et al., 2020) or EnergyPlus. Statement of need Load profiles are utilized by researchers, planners and others to model energy systems in buildings, interactions with electricity, gas or district heating grids, when developing new technologies such as energy management systems or smart grid control algorithms and in many other applications. Averaged profiles over a larger group of individual households are not an valid option for applications such as system simulations and analysis with a high spatial resolutions, since they have a significantly different shape than real individual load profiles, as shown in Figure 1. Pflugradt et al. (2022). LoadProfileGenerator: An Agent-Based Behavior Simulation for Generating Residential Load Profiles. Journal of Open Source Software, 7 (71), 3574. https://doi.org/10.21105/joss.03574.