What goes into a substation capacity figure

Aggregation

Seven feeders feed PG&E’s Hollister substation, and their available capacity adds up to 13.3 MW in the snapshot I pulled. However, the part that’s easy to get wrong is that not all seven feeders update on the same schedule. Some had fresh readings this cycle. Some hadn’t changed in weeks. If I were to drop the stale ones because they “didn’t update,” I would’ve deleted capacity from the total.

This issue turns up with any figure made from multiple underlying records. Utilities update pieces of their own datasets on their own timelines, not all at once, and a snapshot can easily catch two feeders mid-change while five sit exactly where they were last time. While this could be interpreted as staleness, it’s just how utility data ages.

PG&E publishes data at the feeder level, and I group those feeders by substation. I then take each one’s latest available value, whenever that value was last touched. Hollister’s capacity comes out to 13.3 MW. Trimble, with the same math, comes out to 11.4 MW.

Neither number shows a complete picture, because whether a substation can serve additional demand depends on its operating limits and whatever constraints sit along the connection path. Feeders sharing upstream equipment can’t just have their capacity numbers added together and called a substation capacity. The sum is a proxy that uses the available information to draw up a useful metric that can be tracked. If there’s zero capacity available at the feeder level, then we know the same can be said of the substation upstream.

Also, a capacity figure made by grouping isn’t just vulnerable to individual feeders going stale. The grouping itself can change. If the label that says “these seven feeders belong to Hollister” gets computed from a stored, previous version of the data, a substation can lose feeders it still has, if nothing about those feeders’ values changed enough to trigger a recheck. An unchanged value and an unchanged group membership are two different claims.

Age

A file that I download today can still be full of old measurements. The date I collected a record and the date the utility last performed the assessment are two different dates, and they should not be conflated.

For PG&E specifically, I checked the utility’s own timestamps rather than trusting row count as a proxy for freshness. Records get added, regrouped, or republished without anyone re-running the underlying capacity assessment, so an increase in rows doesn’t necessarily mean newer information.

An unchanged value deserves the same care in the other direction. Just because a number sat still while everything around it moved doesn’t mean it’s wrong or abandoned. It might still be exactly what the utility believes to be true. Reconstructing an aggregate means keeping that value in place while being honest about its assessment date.

Data provenance

Zooming out from PG&E, across the utilities in the pipeline, some utilities provide substation-level figures with equipment ratings and connected load attached, while others provide feeder estimates and leave the substation number to be derived. Historical depth varies as well, which alters whether trends can be determined.

A useful comparison across sites has to pay attention to those differences. A developer looking at two candidate sites is able to tell through data notes what’s being measured, what equipment it describes, and how the number was calculated.

The transparency is apparent in the tools and API because it’s important that everything can be traced back to raw numbers. I store the raw numbers from every week’s run in the cloud to ensure that the source files stay intact even if the utility writes over them. That way, even if the logic or circuits change, the correct information can be backfilled. Utilities will also sometimes publish a corrupted export, with a batch of rows tagged to the wrong substation. Keeping the raw file allows us to make a correction later. If I ever have a question about an outlier, I can go back and check.