Energy use in buildings contributes a large part in global energy demand. To reduce energy use in this group of consumers, specially in cold seasons, an automatic control technique is proposed. In this paper, a model predictive controller (MPC) is employed to minimize the boiler activation time. The method uses the building model and incorporates the weather forecast data to act on the actuator in an optimal fashion while treating the user comfort constraints. This technique, as a part, can be embedded into the building energy management system. The building model parameters are obtained via an online identification process using unscented kalman filter (UKF). This identification is performed on-the-fly so the model of a building is continuously updated. The results of the system identification as well as the control performance are shown via Monte Carlo simulations, and compared with the results of a conventional control law. The comparison shows that the proposed method saves $\text{\%}$13 energy consumption in the boiler activation.

建筑中的能源使用量占全球能源需求的很大一部分。为了减少这类消费者的能量消耗，特别是在寒冷季节，提出了一种自动控制技术。在本文中，使用模型预测控制器（MPC）来最小化锅炉的激活时间。该方法使用建筑模型，并结合天气预报数据以最佳方式作用在执行器上，同时处理用户的舒适性约束。作为一部分，该技术可以嵌入到建筑能源管理系统中。建筑模型参数是通过使用无味卡尔曼滤波器（UKF）的在线识别过程获得的。这种识别是即时进行的，因此建筑物的模型会不断更新。系统识别的结果以及控制性能通过蒙特卡洛模拟显示，并与常规控制律的结果进行比较。比较表明，该方法节省了成本。$\text{％}$13锅炉活化中的能耗。