Smart Residential Community Simulator for Developing and Benchmarking Energy Management Systems

TL;DR

This paper introduces a scalable, flexible simulator for residential energy management systems that models diverse configurations of houses and DERs, supporting comprehensive testing and benchmarking in on-grid and off-grid scenarios.

Contribution

The paper presents a novel, scalable simulator capable of modeling any number of houses with heterogeneous DER configurations, integrating real datasets, and supporting multiple control strategies.

Findings

Simulator effectively models diverse house configurations.

Supports benchmarking of control policies across scenarios.

Demonstrates flexibility with community-level simulations.

Abstract

Home Energy Management Systems (HEMS) are being actively developed for both individual houses and communities to support demand response in on-grid operation, and ensure resilience during off-grid scenarios. However, most simulators used for closed-loop HEMS testing are tailored to a specific distributed energy resource (DER) configuration with a fixed number of houses, limiting flexibility and scalability. This leads to additional development efforts to support diverse DER configurations across any number of houses and to integrate appropriate weather and load data pipelines. To address these limitations, we present a scalable simulator capable of modeling any number of houses in both on-grid and off-grid modes as a Gymnasium environment. Each house can have a unique DER configuration - Rooftop Solar Photovoltaics (PV), Battery-only, PV-only, or no DER - and includes models for air-conditioning and eight grouped circuit-level loads. The simulator integrates National Solar Radiation Database (NSRDB) weather and Pecan Street load datasets, supports three default controllers (two for off-grid, and one for on-grid scenarios), and includes performance metrics and visualization tools. We demonstrate its flexibility through simulations on individual houses and a four-house community with heterogeneous DERs, benchmarking the controllers across built-in metrics and computation time. The results highlight the simulator's capability to systematically evaluate control policy performance under varying system configurations.