Improving Demand Response Implementation in Distribution Network by Data Mining in Advanced Metering Infrastructure

Abstract

Advanced metering infrastructure (AMI) is an enabling technology improving the quality of metering and transferring the energy consumption data in the electric energy network. Data with such quality acquired from a bidirectional infrastructure is a highly instrumental in increasing energy awareness in smart cities. The bidirectional communication infrastructure provided by AMI increases the customers' awareness about their consumption behavior, too. One of the issues affected by considering AMI in distribution network is demand side management (DSM). The accurate customer characterization using AMI data is a practical point in increasing energy awareness in the smart cities. The role of AMI in DSM programs is embossed when customers are regarded as influencing participants in these programs. This paper proposes a procedure investigating the role of AMI in improving demand response (DR) by costumer characterization in distribution network. Authors believe that DR is suitably implemented in distribution network if the analysis of customers' consumption behavior is conducted accurately. Besides, customers' capacities for participating in DR are different which is unfold by analyzing their consumption behaviors. Customers' consumption that are recorded by smart meters in different time slots compose time series on which pattern recognition procedures are applied. Different time series are clustered on the basis of their similarities and DR is implemented on each cluster separately. In This study, DR is implemented in two years in one of which capacities of each cluster for participation in DR is estimated and in another one these capacities are compared with the observed capacities. The results of this study confirm the accuracy of the estimated capacities and the proposed method. This knowledge helps the networks' operators to get a more accurate overview on the total capacity of the distribution network for implementing DR. The proposed method is implemented on a real dataset provided by Irish social science data archive (ISSDA).