TRNSYS-AgentControl: A Python-based framework for agent-driven supervisory control workflows in TRNSYS

Abstract

This software provides a Python-based workflow for agent-driven supervisory control in TRNSYS. It supports baseline data preparation from TRNSYS-exported datasets, control-environment construction, external agent training, and runtime inference through the TRNSYS Python interface. The framework is intended for researchers and engineers performing energy simulation and analysis in TRNSYS, enabling agent-based optimization in place of conventional rule-based supervisory logic while remaining configurable for project-specific states, actions, rewards, and operational constraints. Although the implementation demonstrates a Deep Q-Network (DQN) controller for building-scale photovoltaic-battery energy management, the workflow is not restricted to a single algorithm and can be extended to Python-based decision agents.