UVM Theses and Dissertations
Format:
Print
Author:
Seier, Andrew
Dept./Program:
Electrical Engineering
Year:
2014
Degree:
MS
Abstract:
With the advent ofthe 'smart' power grid, mechanisms that leverage increased computation and communications can be developed to reduce energy costs and grid stressors. This work focuses on the adoption of plug-in electric vehicles (PEVs) and their effect on neighborhood transformers. Because Level 2 charging of PEVs is about 7.2 kVA, 25 kVA residential service transformers may be severely overloaded, even at very low levels of PEV market penetration. In this work, we check the validity of standard transformer heating models, explore the extent to which load management via random access techniques can be realistically employed, and gain new insight by bridging these concepts together.
Working with a Vermont utility, two smart transformers were installed along with ambient temperature sensors and anemometers to collect operational transformer data. Using a genetic program, a hot spot temperature curve produced by the Annex G heating model from IEEEC57.91-1995 standard is re-modeled and tested to determine whether the dynamics of small service transformers can be described in a simpler manner. The predicted hot spot temperature of the Annex G model is compared with the measured temperatures of the experimental transformers and is shown to consistently overestimate the actual heating. The measured heating from experimental loading is also fit using the genetic programming approach. Finally, the method of least mean squares is used to find a best parameter fit for the chosen structures arising from the genetic program's outputs. This work shows that the hot spot temperature in a 25kVA transformer can be estimated by simpler means than the Annex G method and that this new model is both less conservative and more consistent with actual measured data.
In addition, the concept of 'packetized charging' is explored as it applies to load management in networks consisting of service transformers and local load. Building off of previously developed simulation methods, this work shows that even small communication or data delays can cause transformer overloads and large oscillating load profiles. We then present a novel change to the 'packetized charging' scheme to allow stable operation of the control system under larger, more realistic, latencies. This simulation model along with the heating model developed herein are taken together to assess the feasibility of the random access method as it would apply to the neighborhoods where the test transformers were installed.
Throughout the work, the concept of leveraging better models and existing infrastructure to prevent costly grid upgrades is central. The results of this work suggest that if one can accurately estimate the state of a transformer (e.g., hot-spot temp, insulation degradation, etc.) high-load PEVs can, with proper management, operated in the smart grid and meet PEV users' expectations without requiring service transformer upgrades.
Working with a Vermont utility, two smart transformers were installed along with ambient temperature sensors and anemometers to collect operational transformer data. Using a genetic program, a hot spot temperature curve produced by the Annex G heating model from IEEEC57.91-1995 standard is re-modeled and tested to determine whether the dynamics of small service transformers can be described in a simpler manner. The predicted hot spot temperature of the Annex G model is compared with the measured temperatures of the experimental transformers and is shown to consistently overestimate the actual heating. The measured heating from experimental loading is also fit using the genetic programming approach. Finally, the method of least mean squares is used to find a best parameter fit for the chosen structures arising from the genetic program's outputs. This work shows that the hot spot temperature in a 25kVA transformer can be estimated by simpler means than the Annex G method and that this new model is both less conservative and more consistent with actual measured data.
In addition, the concept of 'packetized charging' is explored as it applies to load management in networks consisting of service transformers and local load. Building off of previously developed simulation methods, this work shows that even small communication or data delays can cause transformer overloads and large oscillating load profiles. We then present a novel change to the 'packetized charging' scheme to allow stable operation of the control system under larger, more realistic, latencies. This simulation model along with the heating model developed herein are taken together to assess the feasibility of the random access method as it would apply to the neighborhoods where the test transformers were installed.
Throughout the work, the concept of leveraging better models and existing infrastructure to prevent costly grid upgrades is central. The results of this work suggest that if one can accurately estimate the state of a transformer (e.g., hot-spot temp, insulation degradation, etc.) high-load PEVs can, with proper management, operated in the smart grid and meet PEV users' expectations without requiring service transformer upgrades.