branches exchange

[karimsebaa-PRO]

I hope you are doint well

Is it possible to start from the previous voltage solution , in order the speed up the power flow

regards
karim

[mgovers]

Hi @karimsebaa-PRO,

At this moment, that is not yet supported, and given the complexity of the interface, I also do not foresee such a feature in the short term.

However, in practice, we rarely see that that is the bottleneck. Do you notice a bottleneck in your calculations?

• It may be worth reviewing https://power-grid-model.readthedocs.io/en/stable/user_manual/calculations.html#power-flow-algorithms to check whether the algorithm you use to run your calculations suits your needs (especially the performance v.s. accuracy tradeoff).
• Otherwise, it may be worth trying to provide a "minimal reproducible case" (i.e., a grid that you think should be more performant than it is), and we can see if we can find a bottleneck in the calculations.
• Otherwise, can you maybe explain the way you would see the interface to such a warm start? We are open to new ideas, so maybe there's something or some usecase we haven't thought of yet.

[nitbharambe]

While it is theoretically a performance improvement, practically at PGM, we find being able to run parallel calculations independently to be bring in more performance benefits. Note that we begin each calculation with a solution of a linear power flow calculation for newton_raphson.

Im curious about the title though? What does branches exchange mean?

[karimsebaa-PRO]

Branch exchange is a network reconfiguration operation that simultaneously changes the status of two branches: one branch is opened (switched OFF) while another branch is closed (switched ON). The exchange is accepted only if the resulting network topology remains feasible, connected, and radial (arborescent). In other words, after the branch exchange, the distribution network must still form a spanning tree with no isolated nodes and no loops.

[mgovers]

Is it correct that you are more so looking towards the graph analysis than the actual calculation results? If so, it seems to me like maybe power-grid-model-ds provides some functionality you are looking for in that case. It provides some graph utilities that could solve that for you.

Regarding the PGM package itself, if you really are interested in the actual calculations itself, it is more likely that the topology step is the bottleneck than that the warm start would be it. If not, the linear calculation method may help you out - although often less accurate, it does not use an iterative process and is completely linear, so flat start or not does not have any impact on the result.