Anomaly Detection of Gas Flow Data Based on DBSCAN Algorithm
The abnormal flow rate in gas flow data can reduce the accuracy of data analysis and pre-diction.To address the difficulty in detecting individu-al outliers,taking the actual flow data of a gate station for nearly 90 days as the research object,the outlier de-tection method of gas flow data based on DBSCAN al-gorithm is proposed.The gas flow data of nearly 90 day is divided into 88 daily flow samples,and the DBSCAN algorithm is used to detect outliers in the daily flow samples.5％of the total number of samples is set as the abnormal number threshold,and the abnormal sam-ples are detected(regarded as abnormal flow condi-tions).The mean-square error is used as the evalua-tion index to find the most similar sample(the normal sample with the smallest mean-square error),which is used as the reference for outlier detection of hourly flow rate.Taking 5％of the hourly flow rate of the most similar samples as the difference threshold,individual outliers are detected.The results indicate that the out-lier detection method of gas flow data based on DB-SCAN algorithm is feasible,which includes individual outliers in hourly flow rate into abnormal flow condi-tions for detection
万方数据
维普
