Models ~ phaseId
Meter phases can change due to transformer maintenance and load balancing. If the phase changes aren't recorded correctly, it makes future load balancing difficult. This model identifies the true phase of AMI meters based on clustering the meters based on voltage correlation. While power line carrier meter reading systems can typically identify meter phase automatically, radio frequency meter reading systems, which are now being widely deployed, cannot.
You can create and run a new instance of the model via our web interface here: http://omf.coop/newModel/phaseId/fromWiki
The model is based on code and results developed by Logan Blakely and Matthew Reno at Sandia National Laboratory. For a full overview of the method, please see the published methodology and the accuracy results or access the open source code.
A prior version of this model relied The model is based on code and results developed by Jeremy Keen. Those results are expansions of earlier results by Tom Short [Short]. That model was validated against a meter dataset collected from an African utility, and a set of synthetic meter readings generated from GridLAB-D and the PNNL Taxonomic Feeders augmented with an NREL AMI meter data set. In each case, the model identified 100% of true phases on meters whose labels were changed as part of the test.
Large amounts of distributed solar generation could interfere with this technique. It might be possible to correct for that by zero'ing out the meter readings for the daylight hours and only running the model during times with zero solar output. There has also been some work on alternative approaches by [Padullaparti et al.].
The model requires one main input file:
- A .csv file with readings from AMI meters. The first row should contain the names you would like to use to identity each meter. The second row should contain the current best-known phase, expressed as 1, 2 or 3. The rest of the rows should include voltage readings for each of the meters over time. The read resolution For an example of the format, please see this example AMI input.
A confusion matrix, showing any meters whose label did not match the predicted true phase in the off-diagonal entries:
A list of all meters with their input phases and identified true phases:
[Short] Short, Tom. (2013). Advanced Metering for Phase Identification, Transformer Identification, and Secondary Modeling. IEEE Transactions on Smart Grid. 4. 651-8. 10.1109/TSG.2012.2219081.
[Padullaparti et al.] Padullaparti, Harsha, Santosh Veda, Surya Dhulipala, Murali Baggu, Tom Bialek, and Martha Symko-Davies. 2019. Considerations for AMI-Based Operations for Distribution Feeders: Preprint. Golden, CO: National Renewable Energy Laboratory. NREL/CP-5D00-72773.