Gray box modeling of temperature dynamics and power consumption in a swine barn

Controlling indoor environments in animal facilities is essential for animal health and comfort, and productivity of farms. This is most often mitigated by use of multiple fans moving large quantities of air through the barn, displacing heated and pollutant-laden air. One possible means of reducing the energy and costs associated with operating barns is through smart control of these ventilation fans. In order to prototype control strategies for these fans, succinct models of the environment they control must be developed. Physics-based models use more computational power and information than is feasible for rapid controls development, while empirical models don't actionable information. A solution to this problem, “low-order gray-box modeling”, has been developed for commercial buildings and residences but not yet been applied to the problem of modeling livestock barns. Such succinct models of swine and cattle barns can be useful for optimizing control strategies for ventilation fans. The current study uses a gray box model structure to predict thermal dynamics of a swine barn in Ohio. The models use operational data collected from the barn over four years and are used to predict the power usage of the ventilation fans. The models were trained with a genetic algorithm and the continuous time stochastic modeling toolbox and have been tested with operational data. Results show 2-state models perform substantially better than 1-state and that more than six months of operational data is needed to obtain the same confidence in model parameters as that of a single day in which the barn is intentionally perturbed thermally in a pseudo-random fashion. The trained models will be used to analyze and optimize control strategies of ventilation fans to reduce the overall energy consumption of the barn while increasing the comfort for livestock inside.