Abstract:
The invent of advanced metering infrastructure (AMI) opens the door for a comprehensive analysis of consumers con-sumption patterns, which were not possible beforehand. The AMI can be used to address the major problem of power utilities suchas energy theft. This paper proposes a fraud detection methodology using data mining techniques such as hierarchical clusteringand decision tree classification to identify abnormalities in consumer consumption patterns and further classify the abnormalitytype into anomaly, fraud, high or low power consumption based on rule-base learning. The focus has been on generalizing thealgorithm for varied practical cases to make it adaptive to non-malicious changes in consumer profile and provide a clear bifur-cation between theft and anomaly. Moreover, the performance of the algorithm is evaluated when subjected to various types oftheft and abnormality level. This paper proposes a novel validation technique used for validation of preliminary targets derivedfrom classification block. The validation technique utilizes predicted profiles to ensure the accurate bifurcation between anomalyand theft targets. The proposed algorithm is implemented on real dataset of Nana Kajaliyala village, Gujarat, India which showsits high detection ratio and low false positive ratio due to application of appropriate validation block. The proposed methodologyis also investigated from point of view of privacy preservation and is found to be relatively secure owing to low sampling rates aswell as minimal usage of metadata and communication layer. The proposed algorithm has an edge over state of art theft detectionalgorithms in detection accuracy and robustness towards outliers.