PG&E outlines risk-model upgrades, inspection cadence and mitigation priorities in 2026–28 wildfire plan

Pacific Gas and Electric Company presented its 2026–28 base Wildfire Mitigation Plan to an Office of Energy Infrastructure Safety public workshop in May 2025, describing updates to its wildfire risk models, inspection practices and methods for measuring mitigation effectiveness. The company said it will use an updated wildfire distribution risk model (version 4) and a 24‑hour fire consequence simulation to guide where it conducts inspections, vegetation work, equipment replacements and other mitigations.

Why this matters: PG&E serves large parts of Northern and Central California where utility‑caused ignitions can lead to catastrophic wildfires, large power shutoffs and costly liability under California’s inverse‑condemnation legal framework. The company told regulators and stakeholders the new modeling and inspection practices are intended to better target limited resources toward the highest‑risk assets and to reduce ignition counts over time.

PG&E said the updated distribution risk model moves from a pixel‑based approach to an asset‑level probability-of‑failure approach, allowing the company to compute failure probabilities for specific poles, transformers and other named assets rather than only for 100‑meter pixels. “We are anchoring on the wildfire distribution risk model version 4,” said Andrew Obranches, PG&E vice president of wildfire mitigation. The model includes additional years of outage and ignition data, extends the fire‑spread simulation to 24 hours, and incorporates new proxies for egress and suppression and for community vulnerability where feasible. PG&E said the underlying GIS data vintage used for the model training was updated to January 2023.

Inspection cadence and how findings feed operations: PG&E said detailed transmission and distribution inspections operate on a three‑year cycle, with added aerial (drone) inspections and annual or biannual patrols in “extreme,” “severe” and “high” risk categories for distribution. Ground crews perform checklist‑based inspections and a central review team validates aerial data; machine‑learning tools (named SHERLOQ and WALDO in the presentation) assist review but qualified electrical workers make final determinations. Tags generated by inspections are prioritized and bundled for repair work to reduce customer impacts. PG&E said inspection tags are explicitly evaluated when deciding public safety power shutoffs (PSPS): tags within a proposed deenergization polygon can expand or contract the polygon size, and crews may attempt expedited remediation before a PSPS activation.

Measuring mitigation effectiveness: PG&E described a “layers of protection” approach—system hardening, component replacement, vegetation management, and operational mitigations (PSPS, enhanced power‑line safety settings, rapid response)—and said no single mitigation is universally effective. The company explained why attributing avoided ignitions to any single mitigation is difficult when multiple mitigations coexist on a span. For example, PG&E said PSPS effectiveness can be evaluated by checking whether ignitions occur inside or outside a deenergized polygon, but other mitigations applied to the same span complicate attribution. “Each span has multiple mitigations present, and that fact makes it very difficult to determine which mitigation prevented the ignition event,” Obranches said.

Undergrounding and outage proxies: PG&E said it relies on engineering judgment plus historic outage‑to‑ignition ratios as a proxy to estimate undergrounding effectiveness because actual underground ignitions are rare and the company does not yet have long‑term, large‑volume undergrounding experience in high‑fire‑threat districts. When asked about a reported 99% effectiveness figure for fully undergrounded circuits, PG&E staff said the estimate is based on engineering analysis and outage proxies and is subject to revision as real‑world undergrounding data accumulate. The company told the workshop it will continue to refine effectiveness estimates as it gains multi‑year operational data.

Service breakaway connectors and near‑term measures: PG&E said it will standardize use of service breakaway connectors in high‑fire‑threat districts for new construction and will deploy them more widely in the 2026–28 cycle. The company described the connectors as a lower‑cost, front‑line mitigation for service‑drop ignition risk and said it plans to install them at scale (company comment: “probably a hundred thousand or more” over coming years) to evaluate real‑world effectiveness.

Findings from ignition investigations and operational programs: PG&E described several asset‑risk and failure‑analysis programs that feed corrective actions and inspection changes. The SAFE (Strategic Asset Failure Evaluation) and Enhanced Ignition Analysis (EIA) programs drive lab testing, evidence collection, local drone safety assessments, and corrective actions such as revised standards, inspection job aids or targeted replacements. PG&E displayed operational tools used to track failures, including a Palantir Foundry instance and a beta “REACT” tool for locating repeated failure modes. Daily ignition intake emails and monthly deep‑dive meetings were cited as part of the company’s learning loop.

Selected operational statistics and observations discussed during the workshop: PG&E said its ignition investigations found about half of distribution ignitions land within a 10‑foot radius of a pole; on transmission the company said roughly 80% of ignitions fall within a 50‑foot radius of a tower or pole. PG&E reported clearing an additional 50,000 poles last year (including poles that previously had exempt equipment) and said analysis showed long distribution spans over 600 feet make up roughly 1% of spans but accounted for about 9% of recent ignitions. The company also noted that covered‑conductor and undergrounding effectiveness require longitudinal study because new equipment can exhibit early‑life behavior and longer‑term failure modes.

Other tools and partnerships: PG&E said it lends company helicopters to fire‑suppression agencies in season, supports the Alert California wildfire camera network and participates in multi‑agency vegetation‑reduction projects including prescribed burns and large clearing projects with state and local partners.

Quotes and Q&A: During question and answer, PG&E staff said the company expanded its outage dataset through 2022 when building the v4 model, and that certain effectiveness calculations use overhead outage proxies for undergrounded circuits because underground ignitions remain rare in the available record. On a request for tag‑performance data, PG&E agreed to provide a data response on B‑tag completion and failure rates since the company extended certain remediation deadlines and introduced an X‑tag pathway for expedited work.

What was not decided: The workshop was a presentation and Q&A; no regulatory vote or final approval occurred at the session. Energy Safety will accept written comments on the submitted WMPs on published docket deadlines (PG&E opening comments due May 23 for the 2026–28 base plan) and will post slides and recordings to the WMP docket.

Next steps and how the public can comment: Energy Safety will post presentation slides and the recording to the WMP docket and publish deadlines for opening comments. PG&E said it will keep refining model inputs and mitigation‑effectiveness estimates as more operational and undergrounding data become available.