مدل شبکه بیزی برای پیش بینی بار خنک کننده ساختمان های تجاری A Bayesian Network model for predicting cooling load of commercial buildings

1 Introduction In the U.S., building sector accounted for the largest portion of the primary energy consumption in 2010 (U.S. DOE 2014). Furthermore, building energy use is expected to rise by ~31% from 2010 to 2030 (U.S. DOE 2014; EIA 2016). Thus, even small reductions in building energy used can bring great positive benefits to U.S.’s primary energy use. In fact, many studies have been reported in the literature, which reduce the primary energy use through building energy efficiency measures (Xue et al. 2014; Hughes et al. 2015; Hao et al. 2016; Alajmi 2012; Krati 2016; Corbin et al. 2013; Široky et al. 2011; Ma et al. 2012). Those methods include demand response strategies (Xue et al. 2014; Hughes et al. 2015; Hao et al. 2016), energy audit strategies (Alajmi 2012; Krati 2016), and advanced control strategies (Wetter et al. 2016; Huang et al. 2016a, 2017; Ma et al. 2012; Huang and Zuo 2014). To assure the successful application of all these proposed strategies, or for the verification of their implementation (Walter and Sohn 2016), an accurate prediction of building cooling load is necessary (Li and Huang 2013). For example, in the demand response strategy proposed by Hao et al. (2016), the predicted cooling load is required to determine the set points for the temperature of each thermal zones. In Krati (2016)’s energy audit study, the cooling load is necessary in predicting the energy saving from different energy saving methods. The predicted cooling load is also a critical input for the model predictive control strategy proposed by Huang et al. (2016a). This strategy can generate the optimal set points for the future time horizons. Predicting the building cooling load, however, can be difficult. The challenges come from two aspects: First, building cooling load can be affected by countless factors, including weather, internal activities, and occupant preferences (Kim 2011). Considering all those factors simultaneously requires a lot of detailed information regarding buildings. However, this information may not be assessable or is hard to quantify. Second, the relationship between the factors and the cooling load is a complicated non-linear function which is difficult to be described by the commonly used linear regression (Hou et al. 2006). The complexity of the relationship is mainly due to the highly non-linear nature of the building system. For example, the heat transfer between the ambient environment and the building via radiation is governed by the Stefan-Boltzmann law which is described by a non-linear equation.