Behavior Tree based Control Strategies for Resilient Heat Pump Operation in Residential Buildings

Abstract

Behavior trees are a proven concept in the creation of complex task-switching control and artificial intelligence for robotic systems and non-player characters in the computer games industry. Requirements such as flexibility, maintainability, reusability of functionalities or expandability also apply to the control of decentralised energy systems. Despite this, there is a noticeable research gap regarding the application of behavior trees in that sector. Based on a foundational heating system, including thermodynamic modelling of a part-load capable heat pump with TESPy, tree structures for its control are created using the Python library py_trees for implementation. With a view to minimising the annual operational performance indicators electricity price and CO2 emissions, which reflect the optimal use of renewable shares, several control strategies are compared. We identify and illustrate the principal limitations of decision trees, mixed-integer linear optimisation performed with oemof-solph, as well as a classic rule-based approach. The proposed higher-level behavior tree combines the strengths of such approaches whilst pursuing the additional target of reducing the start-up and associated wear of the heat pump without significantly increasing the computation time.