| ## Background |
|
|
| This is level 5 of 5 and continues the work of the earlier levels. The new files for this level are in `workspace/t5/`. The records from the start are in `workspace/t1/`, and the earlier results are in `workspace/t1/results/` to `workspace/t4/results/`. |
|
|
| ## Task |
|
|
| Subject: Fern and the summer heat: what we tell customers |
|
|
| Hi, |
|
|
| Maren has sent the PJM point forecasts to the desks; they now want uncertainty ranges before committing. Those point forecasts need to stay unchanged. The desks ask how PJM's forecast and ours would have held up in Winter Storm Fern (24 to 27 January) and in the heat of 30 June to 2 July. My draft notes are in commercial/. Marketing collected summer Blend bulletins and upgrade notices (weather/nbm/), EIA articles (reading/) and NOAA's storm reports (weather/ncei/stormevents/). |
|
|
| I need a customer-facing review of the two events, with last year's heat of 23 to 25 June 2025 as the comparison: what load did; how PJM's forecast and ours did, above all at the peaks, using the forecasts we could actually have issued on those mornings; how much of our error came from the weather forecast rather than from our model; and how the weather forecasts behind it all did. Tell me which statements in my notes we can keep. |
|
|
| For the desks, please extend the forecaster so it issues an 80% prediction range for each delivery hour alongside its point forecast, under the same scheduler contract. I'd like the range to reflect how forecast uncertainty changes, including with the weather and input availability. Fit and calibrate it on what we had before 2026, then validate the extended system over January to September 2026 with monthly point scores, range coverage, width and interval scores, and show how the ranges held up in both events and at their peaks. The hourly validation results and calibration provenance go in the supporting workbook, and the review should explain to customers what the ranges do and do not promise. The extended forecaster goes in workspace/t5/results/pjm_forecaster/. |
| |
| The review as a PDF in workspace/t5/results/, with the tables behind it as a workbook in the same folder. |
| |
| Sana |
| |
| ## Input |
| |
| - `workspace/t1/`: the records supplied with the work (`company/`, `eia861/`, `eia930/`, `forecasting/`, `github/`, `mail/`, `meetings/`, `projects/`, `reading/`, `weather/`). |
| - `workspace/t2/`: the files that came with level t2 (`ferc714/`, `mail/`, `pudl/`). |
| - `workspace/t3/`: the files that came with level t3 (`notes/`, `weather/`). |
| - `workspace/t4/`: the files that came with level t4 (`eia930/`, `mail/`). |
| - `workspace/t5/`: the files that came with level t5 (`commercial/`, `reading/`, `weather/`). |
| - `workspace/t1/results/`: the results of level t1. |
| - `workspace/t2/results/`: the results of level t2. |
| - `workspace/t3/results/`: the results of level t3. |
| - `workspace/t4/results/`: the results of level t4. |
| |
| ## Output |
| |
| - `workspace/t5/results/pjm_events_review.pdf` |
| - `workspace/t5/results/pjm_events_tables.xlsx` |
| - `workspace/t5/results/pjm_forecaster` |
|
|
