Dynamic Consumer Classification for Dynamic Consumer Billing in Nigeria

Abstract

Electricity consumer classification plays a vital role in power system network planning and electricity service delivery. In Nigeria, consumers are pre-classified based on their type of activity and their envisaged load demand as provided in the consumer data at the point of registration. Unfortunately, many consumers falsify the data they provide and some change activity without notifying electricity service provider. Consumer behavior is dynamic. As such, network planning, billing and service delivery issues occur. A better approach to consumer classification is the utilization of dynamic consumer load pattern. In this system, consumer consumption data is collected and analyzed by a software. Decisions are made based on result. The approach used in this projects applies data mining clustering algorithm to classify consumers. In this approach, Expectation Maximization (EM) clustering algorithm is applied to 386 validated monthly consumer consumption data. The result is analyzed and then used to propose a suitable dynamic consumer billing model. The cluster result generated only four (4) cluster as against the seven (7) classes in the pre-classification system. A ranking of cluster of consumers by consumption magnitude from lowest to highest shows that residential consumers fall into the lowest rank while industrial consumers fall into the highest rank. Most commercial consumers fall in the mid ranks. Some consumers moved across ranks over the two month under investigation signifying a net shift in consumption. The billing model evolved uses the consumer rank and activity class. The four (4) clusters obtained from the result evolves a simplified consumer classification system that eases network planning process. The result also shows the net effect consumers have on the network vis-à-vis the consumer cluster. The result obtained can easily be used to track changes in consumer behavior as to detect irregularities or to make decisions as regard network planning and electricity service delivery. The process can be automated in real-time.