Smart buildings are buildings equipped with numerous sensors and actuators, allowing them to evaluate the state of the environment in or around them and react to it or alter it. Accordingly, one could use this available technology in order to optimize the operation of the buildings in terms of sustainability. One option to approach this challenge is through Hierarchical Task Network (HTN) planning, which allows us to analyze the current state of the environment and create a plan to be automatically executed by the actuators, achieving a certain goal. In this study, we systematically analyze the domain of smart buildings and identify strategies with which we can optimize the energy efficiency. We put primary focus on load balancing and HVAC control techniques and develop a domain model for HTN planning concentrating on them. Special attention is paid to usability and ensuring that the model satisfies the requirements of the occupants. Furthermore, we explore the question of how to represent the uncertain nature of the domain in its model. We choose to realize an approach based on cost-variable operators. For this, we use operators whose cost is non-deterministic. We implement our model and evaluate the implementation in terms of quality and quantity. Our solution proves to lead to a significant decrease in electricity bills, making its usage not only beneficial for the environment but also for its users.