Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ice Calculator topic
No spam. Unsubscribe anytime.
Lawrence Berkeley releases ICE 2 interruption-cost calculator with new surveys and API
Summary
Lawrence Berkeley National Laboratory and partners outlined ICE 2, a substantially revised Interruption Cost Estimator that uses recent, standardized customer surveys, updated models and an API. CPUC stakeholders were shown the methodology, sample sizes, documentation and planned phase releases.
Lawrence Berkeley National Laboratory on June 24 presented the initial public release of ICE 2, an updated Interruption Cost Estimator intended to replace the older ICE calculator used to monetize short-duration power-interruption costs. LBNL and partner Resource Innovations described new survey work, revised customer-damage functions and a programmatic API that utilities and regulators can use to run batch analyses.
The ICE calculator estimates the customer cost of short outages and is used by utilities for reliability planning and benefit–cost analysis. LBNL said ICE 2 replaces multiple, inconsistent older surveys with a coordinated, multi‑phase survey program and updated statistical models; the project includes funding and technical input from sponsoring utilities and federal partners.
Christina Lachomar, program manager at Lawrence Berkeley National Laboratory, told the CPUC workshop: "We're excited to be talking about the updated and newly released ICE calculator." She said phase 1 is complete, phase 2 (which includes all California investor‑owned utilities) is in progress, and a combined release incorporating phase 1 and phase 2 results is expected by the end of the year or early next year.
LBNL senior modeling lead Chris Ramey described the survey approach and modeling choices. "So within residential customers, we used a willingness to pay approach with a hypothetical backup service," he said, and explained that nonresidential customers were modeled using a direct‑cost approach that captures lost production, labor and other interruption‑specific costs.
LBNL reported the phase‑1 effort relied on 11 surveys covering 24 investor‑owned utility distribution territories, producing roughly 3,026 validated residential survey responses and 3,874 validated nonresidential responses used to fit the new customer‑damage functions. Phase 1 plus phase 2 eventually cover about 30 distinct utility service territories; phase 3 will add rural cooperative utilities to fill geographic gaps, the lab said.
The contractors described the modeling approach for the two customer classes: a weighted generalized linear model for residential customers and a two‑part model (probit plus GLM) for nonresidential customers to address the high incidence of zero reported costs in the survey data. Ramey said the final predictor set for residential customers includes variables such as interruption duration, annual usage, season, backup generation, work‑from‑home status and income buckets; for nonresidential customers the model uses duration, usage, advanced warning, day of week and industry mix.
LBNL also demonstrated a new API for ICE 2 to support high‑volume, programmatic model runs. Mike Spears, who helped build the API, provided documentation and examples; the API requires a user account and an API key. LBNL said spreadsheet versions of ICE 2 will be shared on a case‑by‑case basis and are restricted from public re‑distribution while rights and IP are finalized with DOE; the older ICE 1 spreadsheet remains publicly available.
The release includes detailed user documentation, a comparison memo summarizing differences between ICE 1 and ICE 2, the phase‑1 report and survey instruments. LBNL said ICE 2 uses surveys conducted in the last three years to address dated and inconsistent survey vintages in ICE 1 and to broaden geographic representation across U.S. regions.
Stakeholders pressed on technical details in a Q&A session. LBNL described using a LASSO approach to select candidate predictors and cross‑validation for model selection; they confirmed they have not yet completed a formal sensitivity analysis on all predictor variables but plan further work in phase 2. LBNL said the tool interpolates model results between the discrete durations covered in the surveys (momentary, 2, 8 and 24 hours) when users request intermediate durations.
Why this matters: regulators and utilities use ICE outputs to prioritize reliability investments, estimate benefits and support rate filings. LBNL emphasized ICE 2 is a living product — stakeholders should review the documentation and consider the new API for higher‑volume applications.
Next steps and calls to action: LBNL plans to publish the combined phase‑1/phase‑2 report by year‑end or early 2026; phase‑3 surveying of co‑ops is under way to fill remaining geographic gaps. The CPUC Safety Policy Division convened the workshop to gather stakeholder feedback as it finalizes the SB 884 project‑list data template, with a target to finalize that template by 2025‑07‑24.
