Load shape performance
I’m logging the issue that was discussed in the email thread initiated by Henry Liu (PG&E) on 6/30 regarding load shape processing performance. The load shapes uploaded on 6/30 remain in the processing queue as of 7/8. This delay is also impacting the ex-ante measure package review schedule and the CPUC/ex-ante team’s ability to determine next steps.
The processing time was initially estimated at approximately 30 hours, with an additional 25 hours later identified. However, the actual processing duration has significantly exceeded those estimates. Notably, this issue pertains to only a single measure package, while several additional packages must be processed to support the DEER 2028 update.
Please prioritize the resolution of this issue as soon as possible. Thank you.
- Select the module in CEDARS
Log in to comment and vote
Comments4
Roopa Reddy
Aug 6
@cedars team
PG&E has uploaded approximately 15 load shape files over the past two days, with about 67% of the files ranging between 80 MB and 180 MB in size. While some files completed processing within 24 hours, others were processed within 48 hours. As of now, nearly all files uploaded on 8/3 and 8/4 have completed processing successfully.
Based on this improvement, you may close this ticket. Should any additional issues arise, we will submit a new ticket.
Thank you.
Also attaching my update from August 4.
cedars team
Jul 30
We’ve made considerable improvements and are pausing additional work while we monitor processing times for the next week, particularly for larger files closer to the 250MB limit.
Roopa Reddy
Jul 28
@cedars team Could you please provide an update on the parallel processing enhancement and let us know where things currently stand? Thank you.
spencer.lipp@futee.biz
Jul 10
Discussion on the potential improvements occurred in the 7/9/26 FutEE led deemed load shapes issues coordination call. The following three solutions were discussed.
1) Queue management process has capacity to do jobs concurrently to increase speeds. In testing to make sure it is robust. Likely deployed tonight (7/9).
2) Arithmetic operations could be done more efficiently with parallel computations. No timeline given.
3) Pull data from ACCs only once rather than many pulls during the process. Monday (7/13) at the earliest.