search

zaphiro-technologies / protobuf

A repository including protocol buffer used in Zaphiro's platform
Apache License 2.0
0 stars 0 forks source link

[Feature] how to support uncertainty of fault #13

Closed chicco785 closed 3 months ago

chicco785 commented 11 months ago

Is your feature request related to a problem? Please describe the problem.

@kdevelleZ wrote:

I am not sure if it is the write place to put a comment about the fault.proto file @MarcoPignati and @chicco785 : I noticed one things for the fault protobuffer: the 2 followings are elements:

  • optional float lengthFromTerminal1 = 2; //The length to the place where the fault is located starting from terminal with sequence number 1 of the faulted line segment.
  • optional string acLineSegmentID = 3; //The line segment of this line fault.

But in some case the location algorithm is not able find a unique answer and will return : fault is 200m from terminal A on line 1 or fault is 300m from terminal B on line 2

Can you modify it?

Describe the solution you'd like

A solution to tackle the expressed need

Additional context

No response

chicco785 commented 11 months ago

I don't have time today to follow up on this, so this will have to wait a week when I am back from vacations.

If I am not wrong this would break the alignment with CIM structure (see: https://github.com/zaphiro-technologies/architecture/blob/main/features/31-fault-data-storage.md#data-structures)

The thing is that most probably CIM models "confirmed" faults, uncertainty is not part of the model.

That's why in discussions with @MarcoPignati i raised the importance of attaching a number that defines the likelihood of the fault.

A potential solution is to not change anything in the model and treat the fault as two potentially different fault with very close likelihood, and thus only modifying CIM fault model by adding an accuracy or likelihood field.

Feelings?

kdevelleZ commented 11 months ago

no worries, I have not start to code anything yet so enjoy vacation with the kids!

concerning the likelihood of the fault, I think we ll need Mayank feedback.

MarcoPignati commented 11 months ago

As of now Mayank told me that the algorithm cannot be quickly modified to return the overall likelihood. I'll give it a look with him once he'll be back. Late next week

MarcoPignati commented 3 months ago

there will be some work by mayank or students in the coming months on this. Both to give uncertainty in the distance (on the provided fault) and to associate a probability to each provided fault location (if more than 1)

chicco785 commented 3 months ago

@MarcoPignati I think this is now covered, wdyt?

MarcoPignati commented 3 months ago

Yes now it is covered as far as the model is concerned. We can close