Research On Cleaning Methods Of Building Energy Saving Climate Data

The design of building energy-saving can not be separated from the simulation of energy consumption,and the simulation of building energy consumption is closely related to outdoor climate data.The change of climate data directly affects the results of building energy consumption simulation.Due to the poor quality of climate data used for energy simulation during the measurement of climate data,the reliability of energy simulation is poor.Therefore,it is particularly necessary to clean the climate data.So far,researchers have proposed many different methods of data cleansing.These methods can effectively clean up the anomalous data and missing values.However,most of the abnormal data cleansing in climate data are simply statistical calculations and remove abnormal data.The need for excessive manual intervention and the correction of abnormal data lacks certain grounds.This paper aims at the deficiency of the existing anomalous data cleaning method in the processing of climate data,studies the anomalous data cleaning method from different angles and puts forward a method based on clustering neural network to clean the climate data.The main work of the dissertation is as follows:First,the quality problems of the data and the basic principles and basic processes of data cleaning were studied.In particular,anomalous data cleaning methods were studied in depth.Based on the existing method of cleaning anomalous data,the climate data was classified using a method based on K-means clustering.Second,because the K-means clustering algorithm does not deal with the abnormal data,this paper uses the least square method to set the threshold to improve the clustering effect.First,analyze the factors that have important influence on the energy consumption simulation in the climate element.Then,use the least square method to set the threshold and detect the abnormal data.Finally,store the abnormal data in the data set D.Thirdly,the data in the classified data set and data set D is used as a training sample of the BP neural network to carry out network design and training,establish a mapping relationship and obtain a network model.By inputting the data to be measured,the network output value and the actual value are compared,and the error between the network output value and the true value is calculated.If the error is within the set range,it is normal data.Otherwise,it is the abnormal data and is corrected using the network value.Climate data cleaning.Experiments show that the clustering neural network method proposed in this paper can improve the quality of climate data reasonably and effectively,and the effective cleaning rate of climate data is as high as 95.1%,so as to increase the credibility of subsequent building energy conservation design and building energy consumption simulation results.Provide some theoretical support for climate data research.