Reading a utility tariff sheet before modeling a site

Every number in ForgeAsset's tariff library started as a filed sheet — a PDF on a utility's regulatory page, formatted for rate-case lawyers rather than site hosts. Building and maintaining 86 of them has made the failure modes familiar: the sheets are readable, but they are not readable the way a price list is. The price is assembled from pieces that live pages apart, and skipping a piece produces a number that looks precise and is wrong.

This post lists what the model extracts from a sheet and where each piece hides. It describes the process; the methodology page carries the full derivation for every schedule.

First: are you even on the right schedule?#

Rate schedules gate by customer class, voltage, and size. A DC fast-charging site is a commercial customer with an unusually spiky load, and utilities increasingly file EV-specific schedules for exactly that profile — Georgia Power's TOU-EVC-2, FPL's GSLD-1EV, Xcel Minnesota's EV Charging A90. Where an EV schedule exists it usually replaces a general schedule that would bill the same site very differently. Where none exists, the site lands on general service, and size tiers matter: Austin Energy's sheet, for instance, breaks at 300 kW, which a multi-stall site crosses easily.

The eligibility paragraph — customer class, voltage level, minimum or maximum demand — is the first thing worth reading, because everything after it is conditional on being on the right sheet at all.

The demand section is the economics#

Demand charges decide DC fast-charging outcomes, and the library's schedules bill them in at least four structurally different ways:

• Per-kW of monthly peak — the classic form. Duke Energy Carolinas LGS bills a filed rate on the highest 15-minute draw of the month.
• Subscription blocks — the California EV rates sell capacity in advance (PG&E BEV-2-S in 50 kW blocks).
• Load-factor tiers — the Massachusetts EV rates scale the demand rate by how flat the site's usage is, so identical peaks can bill differently.
• Capped per-kW — FPL's GSLD-1EV bills the lesser of the metered peak or the month's energy divided by 75 hours, a cap that shelters low-utilization sites in their worst months.

Two sheets can quote similar headline rates and produce monthly bills thousands of dollars apart on the same site, purely through structure. The utility ranking holds one reference site constant across all 86 schedules to make that spread visible.

The energy rate is a sum, not a number#

The kWh price on the front page is almost never the kWh price. Riders — separately-filed adders for fuel, storm recovery, efficiency programs, capacity — stack on top, and some are percentages of the whole bill rather than per-kWh amounts. Real examples from the library's derivation notes:

• Georgia Power's base TOU-EVC-2 rates carry three rider sheets on top; two of them are percentage riders that fold into the demand charge as well as energy.
• Entergy New Orleans' schedule picked up a storm-recovery rider stack that moved the effective rate by double digits — invisible to anyone reading only the base sheet.
• Several utilities reset rider values quarterly, which is why the library records an effective date and re-verifies on revision.

The digit-check post walks five states' worth of these finds in detail.

Seasons and clocks change the answer#

Time-of-use windows rarely match across utilities — peak might be 4–9 p.m. everywhere on the West Coast and 2–7 p.m. summer weekdays only in Georgia. Seasonal rates split the year into summer and winter blocks with different prices in each. A model needs one annualized rate per TOU bucket, so seasonal filed rates get month-weighted — five summer months at one price, seven winter months at another, averaged by calendar. Every entry in the library that does this shows the arithmetic in its derivation note.

The stamps at the bottom#

Filed sheets carry an effective date and a revision number. Both matter: rates revise mid-year, sometimes on a schedule (fuel riders), sometimes by rate case. The library records the effective date per schedule and updates within two weeks of a revision — because a model calibrated to last year's sheet is a model of a site that no longer exists.

What this buys a scenario#

The payoff of reading the sheet properly is that electricity — typically the largest operating line in the cost stack — enters the model as filed fact rather than a regional average. From there the wizard auto-selects the territory's schedule from the address and prices the full 15-year statement on it; each schedule's rates, structure, and source sheet are on its utility page.