近年来，模型预测控制（MPC）因其在节约暖通空调运行能耗和改善热舒适性方面的潜力而在建筑中得到广泛应用。暖通空调系统的一种MPC是基于EnergyPlus模型的预测控制（EPMPC），其中EnergyPlus模型被集成到MPC算法中，以预测未来的建筑能耗性能。EPMPC可以通过重用通常在建筑项目设计阶段开发的EnergyPlus模型来减少开发MPC算法的工作量。然而，与传统的基于规则的暖通空调控制逻辑相比，MPC，尤其是EPMPC，更复杂，计算量更大。它还需要不断获取更新的预测数据作为计算输入，例如天气预测数据和入住计划预测数据。因此，将MPC应用于实际的暖通空调系统运行是一个挑战。在这项研究中，开发了一个用于暖通空调系统的MPC软件框架，以便于其实现。 EPMPC被部署在宾夕法尼亚州匹兹堡的可持续景观建设中心（CSL），使用该框架作为案例研究。该框架使用客户机-服务器结构构建。在这种结构中，轻型客户端程序在与楼宇自动化系统（BAS）相连的本地计算机中运行，以将控制值从MPC计算写入HVAC系统，而重型服务器程序可在远程计算机中运行，以进行密集的MPC计算。因此，MPC的计算密集型工作隐藏在幕后，MPC成为BAS的简单“插件”算法。通过在服务器程序中提供软件接口，该框架还将MPC算法与用于为MPC计算提供输入的预测模型解耦。这项研究表明，通过使用该框架，EPMPC算法可以在CSL大楼的暖通空调系统中成功实施，而不会对大楼的现有BAS进行重大更改。 文中还对EPMPC算法进行了评估，并对其可扩展性、灵活性和暖通空调控制性等实际问题进行了讨论。引用:2017年年度会议，加利福尼亚州长滩，会议论文

Model predictive control (MPC) has become popular in buildings in recent years due to its potential to save HVAC operation energy and improve thermal comfort. One type of MPC for HVAC systems is EnergyPlus Model-based Predictive Control (EPMPC), where an EnergyPlus model is integrated into a MPC algorithm to predict future building energy performance. EPMPC could reduce the effort of developing a MPC algorithm by reusing the EnergyPlus model that is commonly developed during the design phase of a building project. However, MPC, especially EPMPC, is more complex and computationally-intensive compared to traditional rule-based HVAC control logic. It also needs to constantly acquire updated forecast data as inputs for computation, such as weather forecast data and occupancy schedule forecast data. Therefore, implementation of MPC to real HVAC systems operation is challenging. In this study, a software framework of MPC for HVAC systems was developed to facilitate its implementation. EPMPC was deployed in the Center for Sustainable Landscape building (CSL) in Pittsburgh, PA by using the framework as a case study. The framework is constructed using a client-server structure. In this structure, a light client program runs in the local computer connected to the building automation system (BAS) to write control values from MPC computation to HVAC systems, while a heavier server program can run in a remote computer to conduct intensive MPC computation. Hence, the computation-intensive work of MPC is hidden behind the scene and MPC becomes a simple "plug-in" algorithm for BAS. Through providing software interfaces in the server program, the framework also decouples MPC algorithm from forecast models that is used to provide inputs for MPC computation. This study demonstrated that, by using the framework, an EPMPC algorithm can be successfully implemented in the CSL building's HVAC systems without major changes to the building's existing BAS. The EPMPC algorithm is also evaluated and the practical issues, such as scalability, flexibility and HVAC controllability, are discussed.