【作者】 刘璐；

【导师】 俞准； 胡宏敬；

【作者基本信息】 湖南大学 ， 建筑与土木工程（专业学位）， 2019， 硕士

【摘要】 基于实际建筑在室人员数量对暖通空调系统进行实时控制可以有效降低建筑能耗。现有在室人数预测研究中二氧化碳浓度和wifi信号个数分别是应用非常广泛的直接和间接相关参数,但其优点与局限性也较为明显,分别表现在二氧化碳浓度预测结果受延迟性影响但对高人数水平情况判断准确,wifi信号个数易受信号缺失影响但数据具有实时性,而将两者进行耦合则可实现互补。基于此,本文主要研究目标为如何通过有效参数耦合进一步提升在室人数水平预测精度。为实现该目标,本文首先建立了结合hobo门开关状态对wifi数据进行预处理的新方法,在此基础上建立了基于二氧化碳浓度和wifi信号个数的参数动态耦合预测模型。为验证所建方法的有效性,本文基于实测数据建立了预测模型并进行了验证,并进一步对该模型在工作日与非工作日下的预测结果和模型影响因素进行了分析,主要结论如下:(1)所建wifi数据预处理新方法可以有效地对室内信号与室外信号进行区分并筛选出室内人员wifi信号,为wifi数据用于在室行为预测提供了基础;(2)当二氧化碳浓度增值在500ppm以上和250ppm以下应分别采用二氧化碳浓度和wifi信号个数作为在室人数水平预测参数,二氧化碳浓度增值在250ppm至500ppm之间时应对二氧化碳浓度与wifi信号个数的模型训练结果进行加权线性组合来预测在室人数水平;所建基于二氧化碳浓度及wifi信号个数的参数动态耦合模型对在室人数水平的预测精确度相比于单参数模型下的预测结果提高了约10%;(3)在工作日和非工作日情况下,二氧化碳浓度增值在250ppm至500ppm之间时两参数的加权线性组合中权重取值应根据实验环境在工作日和非工作日情况下的wifi信号缺失程度和二氧化碳延迟程度确定。就模型影响因素而言,转移概率步长变化对预测结果的影响较小,而模拟步长变化对预测结果的影响较大,当模拟步长与模型转移概率步长均取10min时预测精确度最高。未来重点研究方向应集中在对多种类型房间进行实验,确定参数耦合时边界二氧化碳浓度增值的取值规律,并进一步将所建模型与暖通空调系统运行控制相结合,从而评估该方法在实际应用中的效果及节能潜力。更多还原

【Abstract】 The optimal control and operation of HVAC system with real-time occupancy schedules can effectively reduce the building energy consumption.In existing studies,carbon dioxide concentration and WiFi signals are direct and indirect representation of occupant information(e.g.presence or count)which are widely used for occupancy estimation.However,due to its inherent characteristics,they have certain advantages and limitations.Carbon dioxide concentration shows relatively better performance in inferring high level of occupancy while its performance is strongly affected by delay issue.On the contrary,Wi-Fi sensing technologies can easily and timely obtain Wi-Fi signals which is suitable for real-time estimation.The limitation of using Wi-Fi signal is that they are easily influenced by signal loss or noise signals.A suitable combination of using Wi-Fi signals and carbon dioxide concentration can overcome their limitations.Accordingly,the main objective of this study is to develop an effective method to reasonably combine the carbon dioxide and Wi-Fi signals so as to improve the accuracy of occupancy estimation.Specifically,a new pre-processing method based on door opening status was proposed in this study to address the issue of Wi-Fi signals.With the preprocessed data,a dynamic coupling prediction model that optimizes the use of carbon dioxide concentration and Wi-Fi signals was established.In order to demonstrate the effectiveness of the proposed method,35-days on-site measurement of occupancy behavior in an office building was carried out.The model performance was evaluated and verified for weekday and weekend based on the measured data.And the model performance is also compared with single parameter model(i.e.the model only use carbon dioxide concentration or Wi-Fi signals).In addition,to further analyze the performance of the proposed model,the predictive accuracy and respective influencing factors were analyzed for weekday and weekend,respectively.The main conclusions are summarized as follows:(1)The new pre-processing method for Wi-Fi data can distinguish indoor signals from outdoor signals effectively and recognize the Wi-Fi signals that belongs to occupants in a given room,which provides a basis for developing the occupancy estimation models.(2)When the increment of carbon dioxide concentration is above 500 ppm and below 250 ppm,the carbon dioxide concentration and the number of Wi-Fi signals should be used as a stand-alone parameter to infer occupancy pattern,respectively.When the increment of carbon dioxide concentration ranges between 250 ppm and 500 ppm,a weighted linear combination of carbon dioxide concentration and the number of Wi-Fi signals is used to predict the occupancy level.Results indicated that the prediction accuracy of the dynamic coupling model is about 10% higher than that of the single parameter model.(3)Due to the significantly different occupancy patterns in weekday and weekend,the weight value of the weighted linear combination of the two parameters needs to be determined by the degree of signal loss in Wi-Fi data and the delay in carbon dioxide concentration in the building.In terms of the model influencing factors,the time step of transition probability has less influence on the prediction accuracy while the selection of simulation time step significantly affect the prediction accuracy.Results showed that the prediction accuracy is the highest when the simulation step size and the model transition probability were both set as 10 minutes.Future studies need to focus on the identification of appropriate increment value of carbon dioxide concentration which is crucial for the success of model application.Thus more experiments would be carried out for different room types.Also,the integration of the proposed model with different HVAC system operation strategies will be studied in the future so as to evaluate corresponding energy saving potentials in practical application.更多还原