Model-predictive control techniques for hydronic systems implemented on wireless sensor and actuator networks

Motivated by the need to control complex hydronic systems, this paper presents and analyzes control strategies for energy efficient thermal load regulation by means of a model-predictive control (MPC) embedded into a low-power wireless controller. Hydronic systems are found in a wide variety of applications such as building controls, industrial plants, and naval vessels. A laboratory test-bed which models a simple hydronic system is used to assess the proposed control solutions. The key design parameters analyzed by this paper are the form of the objective function and the structure of the model used for open-loop (OL) optimization during each step of the MPC solution. Objective functions were evaluated based on complexity and efficacy in meeting the stated goal of efficient thermal load regulation. The hydronics were modeled as first-order linear differential equations with non-linear time-varying constraints on control, and as bi-linear differential equations in which the control variable is multiplied by the state but constrained linearly. To minimize the objective functions associated with OL trajectories, a gradient descent algorithm was selected which balanced real-time execution with microcontroller (MCU) computing power. Parametric studies were performed in simulation and experiment to show that a low-power MCU could be used to efficiently control hydronic plants using MPC.