Webinar-Mapping Low Voltage Networks Using AMI Data - Shared screen with speaker view
Who can see your viewing activity?
To find out more about Data Access Service, please click on the link below:https://c4net.com.au/services/data-access-services/
For EV data service:https://c4net.com.au/services/ev-data-warehouse/
We encourage you to pop your questions in the Q&A section (or this chat box)
have you used micro PMUs
@ Abhinab, No, we only use smart meter data
Where do we calculate the line impedance? within the smart meter?
No, they are calculated at a central point
@ Jian, it is done centrally using the smart meter measurements from all customers in a circuit
Try Rate of Change of Frequency (RoCoF) methods with cheaper microPMU applied
Thanks, if I understand correctly, it is delta voltage over change in current. What is the time interval in polling V & I?
Smart meters provided only 5 min interval data of voltage, current and power factor for every meter accros the distribution sub/transformer.
How do you know which phase is supplying?
We have developed a clustering algorithm to identify customer's phase
what pre-processing is done before applying TSNE?
We filter the voltage time series by removing outliers
how do you calcuate the distance between two residential customers? As you don't have the real GIS toplogy as well as you don't know the cable types
We calculate 'electrical distance', from voltage sensitivity coefficients that we estimate
"electrical distance" means impedance (or actually, resistance)
Hi there, good presentation! Just wondering whether you use constrained or unconstrained k-means clustering for the phase grouping ?Also, the algorithm for the impedances ("the black box") is based on some AI/Machine Learning method?Thank you
Very good presentation. will you share the recording please?
Great presentation - do your assumptions take into account reverse power flow conditions?
@Robert, yes it works with reverse power flow.
@Angela, we just used regular k-means. What sort of constraints did you have in mind? Also the calculating of impedances was an optimization problem. It wasn't a typical ML algorithm, but ML is basically just optimization.
@Andrew, thank you for your answer.In terms of constrained k-means or not, the question is whether there is an information for some of the customer's phase groups, making it constrained. Or, no information at all for any phase group of any customer making it unconstrained.
Great stuff Reza and Team! Many thanks for the preso!
Thank you for this presentation, I'm pretty new at this domain so please excuse my very naïve questions ?? have we got this models tested with the real time data ? i.e. not only applying constrained dataset in controlled environment and is the efficiency of this model better than already processes working with Distributors prod system i.e. phase detection, meter impendence, topology, fault detection, if yes, then how was the difference of effectiveness identified
Thanks for joining us.We would highly appreciate your feedback, please take a moment to complete this survey.https://forms.office.com/r/q4CE9bKfn7
Thanks for this great presentation.
Yes, the recording will be available on our website soon.
how does the service line interfere with the accuracy of data? some conductors are very old and may have a different gauge compared to new one..