In this paper, we present a prediction-based dynamic programming (DP) control approach, a nonlinear model predictive control (NMPC) approach, and a linear optimal control (LOC) approach to analyze the minimization of the total energy use of a hybrid ground-coupled heat pump (hp) system (incorporating a ground-coupled hp, a gas boiler, a passive cooler, and an active chiller) under operational constraints. A large-scale emulator model (based on finite-volume method and the equivalent-diameter approach) is used for the borehole system and for the assessment of different control algorithms. A nonlinear autoregressive exogenous model is identified from the inputoutput data generated by the emulator model to be used in a DP-based controller. Since DP is a global optimal control method, it was used as a reference for performance assessment. Next, a state-space reduced-order control-oriented model with a larger sampling time is obtained from the emulator model using the so-called proper orthogonal decomposition model reduction technique. This model is used in an NMPC algorithm to see how much NMPC is suboptimal with respect to the DP in terms of annual energy use minimization. Finally, a series of LOCs based on constant hp coefficients of performance is tested to see how much the system performance deteriorates. The control algorithms are used for the satisfaction of heating-cooling demands of three types of buildings: 1) heating dominated; 2) cooling dominated; and 3) thermally balanced. The effects of constraining thermal buildup/depletion of ground, variable electricity prices, and marginal violation of thermal comfort on the performance of the different controllers applied are also separately analyzed.

No abstract

This paper presents a sensitivity analysis of MPC performance (thermal comfort and energy use) to variations of the implemented parameters, with the final goal to rank the building and system parameters influencing the robustness of the MPC. The parameters of the system (ambient temperature, solar radiation and occupancy prediction values; measurement accuracy of zone and CCA temperatures; cost function values such as cost of the CCA or AHU production units; system constraints such as available thermal power of the production units, ) are varied within typical uncertainty ranges and the change in control performance is investigated. The resulting parameter ranking serves as a priority list for future improvement of MPC robustness.