Analyzing household charging patterns of Plug-in electric vehicles (PEVs): A data mining approach

摘要

Plug-in electric vehicles (PEVs) have now become ubiquitous. Due to its high-power consumption, the utility industry has strong motivation to provide this extra demand on off peak hours. It is anticipated that the potential stresses on power delivery systems can be mitigated through asset management, system design practices, controlled charging of PEVs, or some combination of the three. Due to high variability in customers' EV choices, car types, charging patterns, charging preferences, and participation in utility-centric TOU (time of usage) charging options, utility will not be able to manage this risk in an ex post fashion. In many cases, it is likely the utility will not be notified or aware of an electric vehicle (EV) addition. Such inefficiency calls for big data-centric approach that can not only detect households with PEVs, but can also provide some customized PEV charging incentives to managing imbalance in the grid system. Present paper developed a data-driven approach detecting households charging PEVs. The analysis performed here is based on high frequency advanced metering infrastructure (AMI) data obtained from a large-scale utility company in Midwest. The data-driven models developed here achieved an accuracy of 80% or more in successfully detecting households with PEVs.