| At present,with the development of “Big data”,“Artificial intelligence” and other popular frontier technologies,the field of building energy has gradually carried out the related research on intelligent control direction.A complete intelligent building energy system has many sensors to monitor the physical variable and huge transmission network and data management platform.Every working sensor shoulders the important task of collecting,transmitting and storing a large amount of data.Besides,the stable network environment also plays a vital role in the control system.However,the sensor itself has the limitation of small power supply and storage capacity,and it is often will disturbed by external factors such as strong electromagnetic environment.In addition,the communication network will fluctuate during the operation of the system and people also make some wrong operations,then the monitoring data may be missing randomly and unpredictably.A complete real-time database is the foundation to ensure the safe operation of the building energy system.The adjustment of each time operating mode depends on the data feedback of the previous time,and the system is in a non-responsive state because of the lack of data at a certain time.It seriously affects the feedback control of building energy system,and even affects the safe operation of the system and the value of database application analysis,leading to unnecessary energy waste,reducing environmental comfort and equipment performance loss,etc..Therefore,it is of great significance to study how to solve the problem of data missing in building energy system and to reconstruct and repair it.In this paper,the monitoring data of all sensors in the system are taken as the research object,and take the primary air return system and cooling station system as examples.And after systematic research and classification,corresponding reconstruction methods are designed for different data missing conditions to repair the data.At the same time,considering the sensor fault and the monitoring data drift,the idea of parallel calibration and reconstruction is put forward,which avoids the error accumulation caused by using historical data many times,and ensures the accuracy of data reconstruction.According to the characteristics of building energy system as a thermal system that the high coupling among physical variables,an analytical model is established,it avoids the high dependence on the data directly by using the method of fitting the monitoring historical data to generate the mathematical model.On the contrary,because all variables in the model are closely related,the mutual restriction improves the reliability and accuracy of the prediction data.The essence of the fill data method is to find the estimation value which is the closest to the true value of the missing data point.The key technique of this research is Bayesian reasoning,which is to reconstruct the data by establishing the distance function of the likelihood function,and to reduce the dependence on the historical data by introducing a small number of historical arrays.In addition,since Bayesian reasoning is composed of three parts: a prior distribution based on experience,a likelihood function with historical data and a posterior distribution representing results.In this paper,the prior distribution based on the precision of monitoring instruments is set up.At the same time,in order to satisfy many independent variables and unknown variables in the model,the EM and MLE methods are used to inversely deduce the normal distribution parameters.This method solves various data missing mechanisms simulated and fills in the data one by one.Finally,it is verified that this method can guarantee the accuracy of the estimation under various conditions,and has high filling efficiency and large filling dimension,as well as the accuracy of the historical data and the integrity of the database,this is of great significance to improve the intelligent building energy consumption system. |
PDF Full Text Request