Anomaly Detection of Pollution Control Equipment Based on AMI Data Analytics

Abstract

In order to reduce industrial pollution, various policies have been issued in many countries and regions encouraging businesses to install pollution control equipment (PCE). However, some businesses do not use PCE as required, which is currently a blind spot for the regulatory authorities. Advanced metering infrastructure (AMI) provides large amount of electrical data which can well characterize the operating states of PCE. To this end, the anomaly detection of pollution control equipment (ADPCE) based on AMI data analytics is proposed in this paper. Clustering and Gaussian models (CGM) are used as the basic tools to identify the abnormal conditions of PCE, and convolutional neural network (CNN) is further introduced to detect the anomalies when the PCE is suspected to be replaced by a similar process equipment. The case study using realistic and simulated data demonstrates the validity of the proposed methods. The achievements made in this paper can effectively help the government supervise the implementation of pollution control by relevant businesses.