办公建筑中暖通空调系统(HVAC系统)的运行能耗占总能耗比例较高.随着绿色建筑的大力推广,准确预测绿色办公建筑HVAC系统能耗是建筑运行优化的关键.研究以天津市某绿色办公建筑为研究对象,根据绿色办公建筑G中的HVAC系统——地源热泵系统和空调通风系统能耗的实际监测数据,建立了基于人工神经网络的能耗预测模型.研究结果表明,建立的分类多层感知器神经网络预测模型预测精度最好,仅基于气象参数及时间能够精确的预测建筑HVAC系统的小时能耗,为我国绿色办公建筑的设计和运行优化提供科学支持.%With the development and promotion of green building, by the end of 2015, 3979 green buildings were evaluated in China, and the public buildings accounted for 51. 5%. Public buildings belong to the building of large energy consumption and relatively centralized management, which have a great potential to improve the energy efficiency. The researches of the United States green building council ( NBI ) , have shown that the building energy saving of only 30%of the LEED certification buildings exceeded the expectations, building energy saving of 25% of the LEED certification buildings bellowed the expectations, and building energy saving of part of the LEED certification buildings exceeded the standard reference. Domestic researches on the green building operation are also found that there is still a gap between the operation and design of the energy saving rate of green building energy consumption and energy saving rate, and the actual operation may not fully comply with the design concept. According to the survey analysis on public building energy consumption in China and Tianjin, the proportion of the HVAC system energy consumption of office buildings is more than 50%of the total energy consumption. Therefore, accurate prediction of the energy consumption of HVAC systems is potential to achieve high efficiency and energy-saving green building system. The prediction models of green office building with ground source heat pump system and air conditioning ventilation system, energy consumption model based on artificial neural network are established. Based on main functions of the office building, depending on different system demand in different seasons, we divided a year into heating season, cooling season and transition seasons, and then the forecasting models will be set up respectively. The monitoring data of this study contains two heating seasons of 2014 -2015 and 2015-2016 and one cooling season in 2015. First of all, through a variety of structures of the network model, the system running data is used for training the models. Ultimately, the classification of multilayer perceptron network model is determined to train to obtain the optimal prediction model. Eventually, the errors of the prediction models established on the classification of multilayer perceptron network prediction are small, and the precisions of them are high, which can predict the hours of system energy consumption under the certain meteorological conditions more accurately. The prediction models of ground source heat pump system and air conditioning ventilation system can both accurately forecast the tendency of the hours of system energy consumption. The energy consumption prediction model is a dynamic model, and after the model being trained by many years of data and a variety of network learning method, the model precision will be further improved. The input parameters of the established model of this study are fewer and only based on weather parameters and building system operation data the model can predict the energy consumption of equipment system accurately, which can provide scientific support for the operation optimization of green office building.
展开▼
