Greenhouse Gas Emission Data Anomaly Detection Method Research And Application

With the rapid development of social economy,more and more greenhouse gas emissions have brought more and more serious global climate problems,which seriously threaten the living environment of mankind.In order to achieve the goal of energy-saving,emission-reduction and low-carbon development,our government is preparing to carry out carbon trading.In 2016,the National Development and Reform Commission of Henan Province(NDRC)commissioned us to develop the Data Reporting Platform for Greenhouse Gas Emission in Henan Province.It is used to collect the greenhouse gas emission data generated in the production process of large and medium-sized enterprises in various industries in Henan province for nearly 10 years,and provide a data support for smooth carbon trading.There may be a small number of enterprises are overstating or exaggerating the data in order to gain substantial benefits in the future carbon trading,so it is very important and necessary to detect the emissions data reported by the enterprises.Because of the low efficiency of manual audit and the erroneous judgement caused by the uncontrollable factors in the process,this paper studied the abnormal detection of greenhouse gas emission data,and proposes the use of clustering method to quickly identify the data set,review the key data according to the clustering result,and then construct an anomaly detection model based on the identified data.This paper first describes the traditional anomaly detection methods,and then,this paper introduces the method of fast identification of data with clustering idea.Based on the classical fuzzy C-means(FCM)algorithm and KNN-RFCM algorithm,the adaptive equalization weight K-nearest fuzzy rough set C-means algorithm(SKNNRFCM)is improved.The FCM algorithm is combined with K-nearest membership degree and rough set and the equalization weight function is improved.The GHG data set is divided.According to the characteristics of the data set,the samples belonging to the large class are judged as normal data,and the small class is abnormal data set..The BP network is trained by the identified samples of the training set,and the particle swarm optimization(PSO)algorithm and BP are combined to construct the anomaly detection model,which makes up for the problem that the BP network fails to ensure the convergence result and improves accuracy of anomaly detection.In this paper,the emission data of the enterprises involved in the power generation industry in Henan Province in the last 10 years are regarded as the object of study.Moreover,the methods and classical algorithms are compared with our experimental results.Our results show that the method in this paper can detect the abnormal data more accurately when compared with other detection methods.Finally,the anomaly detection module is designed,implemented and applied to the greenhouse gas emission platform of Henan Province.