spb/datacenterindex

1# PJM Interconnection — news releases & Inside Lines (`pjm`)23| | |4|---|---|5| kind / priority | `utility` / 2 |6| config | `config/connectors/pjm.yaml` |7| parser | `news_press_release_v1` (`apps/worker/src/connectors/news/press-release-parser.ts`) — PDF headline heuristic; records identical to `news_article_v1`, see `docs/PROJECT-EXTRACTION.md` |8| schedule | daily (seed index, RSS, releases) |9| license / attribution | Corporate press releases and newsroom articles (PJM Interconnection, L.L.C.); facts and links only — "PJM Interconnection (pjm.com, insidelines.pjm.com)" |10| stored per item | text kept (`keepText: true`) |1112## What it is13The grid operator for 13 states + DC (Northern Virginia, the world's largest data center market). Two channels:14- **News releases** — PDF letterhead files linked from `/about-pjm/newsroom/announcements-and-news-releases` (`/-/media/DotCom/about-pjm/newsroom/<year>-releases/<yyyymmdd>-<slug>.pdf`, 47 files 2023–2026): board plans to integrate large loads, capacity auctions, reformed interconnection cycles, Reliability Resource Initiative.15- **Inside Lines** — WordPress newsroom `insidelines.pjm.com` with RSS `/feed/` (10 items): large-load disconnection reliability standards (4,000 MW data center load drop in Northern Virginia, July 2026), operations updates, governance.1617## Quirks18- `www.pjm.com/about-pjm/newsroom.aspx?news=All&rss=1` returns an **empty RSS channel** — releases are discovered by following PDF links from the announcements page (seed, `news_index`).19- PDF metadata titles are present on recent files; when missing, the first text line is letterhead ("Contact: PJMNews@pjm.com …", "NEWS RELEASE", "FOR IMMEDIATE RELEASE") and `news_press_release_v1` picks the first non-boilerplate line as the headline. Dateline `(Valley Forge, PA – Jan. 16, 2026)` gives the date.20- `insidelines.pjm.com` is served at **L1 to the crawler UA** (robots.txt 200, allows everything but `/wp-admin/`); a generic `Mozilla/5.0 (compatible; …)` UA gets a WAF 403 — nothing to do with robots.21- Hot-weather / operations updates mention data center load in passing and are archived as weak events; the classifier may label them `planning_document` / `facility_page` (event only, never a project).2223## Verification24Verified live on 2026-09-12 with `scripts/try-connector.ts` (dry run, nothing persisted): 58 URLs discovered (1 seed, 47 PDF releases, 10 RSS items), 5 fetched at L1 (0 credits): seed sentinel + 4 Inside Lines news_events dated 2026-09-01 … 2026-09-10, 0 errors. `--url` on the PDF *PJM Board Outlines Plans To Integrate Large Loads Reliably* → 1 news_event dated 2026-01-16 (PDF parsed at L1, 168 KB). `--url` on *PJM Proposes Reliability Standards to Manage Large Load Disconnection Events* → 1 news_event dated 2026-09-10. 0 projects (policy releases), 0 validation errors.25