根据EIA的数据，2019年，如果将电气系统的能量损失计算在内，住宅和商业建筑消耗的能源约占美国最终使用能源总量的39%。因此，建筑能源系统的先进和优化运行将减少能源使用，改善室内环境条件。为了在建筑系统中实现最佳性能，准确的组件模型对于预测系统性能并最终优化它至关重要。本文提出了一种综合的两种方法- 水平优化方法，用于优化冷冻水VAV系统和成套DX VAV系统的性能。第一级优化是使用遗传算法（GA）的组件模型优化（CMO）。优化算法将确定精确组件建模的最佳模型结构。此外，还进行了参数研究，以选择最佳模型结构，并验证优化过程的结果。第二级优化是系统性能优化（SPO）。 SPO是通过创建一个优化算法来执行的，该算法将集成所有创建的组件模型的输出。数据驱动模型的输出将被优化，最终用于优化整个系统的设定点。本文选择的优化设定点是送风温度（SA）和静态风管压力（PS）。测试结果表明，所提出的模型能够准确预测系统性能。此外，两者- 与本研究选择的基线情况相比，水平优化过程在风机功耗和再热能量方面分别节省了25%和22%。此外，本文还验证了此类数据驱动模型的使用，这些模型可以帮助开发行业的几个方面。引文:2021篇虚拟会议论文

In 2019, when electrical system energy losses are included, about 39% of the total U.S. end-use energy was consumed by the residential and commercial buildings according to EIA. Thus, advanced and optimal operations in building energy systems will reduce their energy uses and improve indoor environmental conditions. To achieve optimal performance in the building systems, accurate components models became crucial to predict the system's performance and finally optimize it. This paper proposes an integrated two-level optimization approach to optimize the performance of the chilled water VAV system and a packaged DX VAV system. The first level of optimization is a Component Model Optimization (CMO) using genetic algorithm (GA). The optimization algorithm will determine the optimal model structures for accurate component modeling. Also, a parametric study was conducted to select the best model structure and validate the optimization process results. The second level of optimization is System Performance Optimization (SPO). The SPO was conducted by creating an optimization algorithm that will integrate the output of all the created component models. The output of the data-driven models will be optimized to be eventually used to optimize the set points of the whole system. The setpoints that were chosen to be optimized for this paper are the supply air temperature (SA) and the static duct pressure (PS). The testing results show that the proposed models can accurately predict system performance. Moreover, the two-level optimization process has resulted in a 25% and 22% savings in the fan power consumption and the reheat energy, respectively, when compared against the baseline case that was selected for this study. Also, this paper has validated the use of such data-driven models that can help in developing several aspects of the industry.