The Bill After You Build:
Standby Rates and
Departing-Load Charges

A self-generation business case is the difference between two utility bills: the one the site pays today and the one it pays after the machine energizes. The first is known to the penny. The second has line items most models have never met, and every one of them is published.

Section 01Two bills, one model

When an owner evaluates on-site generation, the seller's model almost always prices the wrong bill. It takes the site's current invoice, multiplies the kilowatt-hours the new machine would displace by something close to the current average rate, and calls the product savings. The arithmetic is correct. The premise is not, because the bill that arrives after energization is a different document, assembled under different tariff schedules, with lines that did not exist before the project and lines that survive it.

Those new and surviving lines come in two families. The first family bills you for the grid you keep: standby service, which prices the utility capacity held in reserve for the hours when your generator is down, plus the ordinary charges of remaining a connected customer. The second family bills you for the grid you leave: departing-load and cost-responsibility charges, which recover costs the utility incurred on your behalf while you were a full-requirements customer, and program charges that regulators have made nonbypassable by design.

Neither family is obscure. In California each investor-owned utility maintains a standby schedule and departing-load provisions in a public tariff book, and the state's regulator has been ruling on the framework for more than two decades. This paper reads one utility's published documents as the worked example; the other utilities' versions differ in numbers and details, and those differences are themselves a study finding for any specific site. The purpose here is not to recite rates, which change, but to name the mechanisms, which persist, so that an owner can recognize a model that has quietly omitted them.

Section 02Standby service: the grid you keep

Standby service exists because of a physical fact: the transformer, feeder, and upstream capacity serving a site are sized for the site's full load whether or not a generator is running behind the meter. A customer whose machine serves nearly all of its load on most days still expects the full load to flow from the grid on the day the machine is down. Someone pays to keep that capacity standing ready. Standby tariffs make that someone the customer who benefits, which is defensible cost causation, not a penalty for self-generating.

The mechanics are visible in Pacific Gas and Electric Company's standby schedules. The legacy schedule, Electric Schedule S, is closed to new enrollment; its successor, Schedule SB, is the standby schedule a newly interconnecting customer encounters today.^1,2 The structure has four working parts.

The reservation charge. The customer contracts for a reservation capacity, the amount of utility capacity held in its name, and pays a monthly charge per kilowatt of that reservation whether or not any backup energy flows. On the current PG&E sheets the charge is, in the tariff's own words, "applied to 85 percent of the Reservation Capacity," a built-in diversity allowance acknowledging that standby customers do not all fail at once.¹

The unbundled component table on the same sheets explains the gap between those two numbers: at secondary and primary voltage, $13.97 of the $16.29 is distribution; at transmission voltage there is almost no distribution wire being reserved, and the charge falls to $2.08.¹ Where a site connects is therefore one of the largest levers on its post-build bill, and it is a lever most technology proposals never mention. Arithmetic on the published sheet, not a forecast: a secondary-voltage customer reserving capacity at that rate pays on the order of $166 per reserved kilowatt per year until the sheet changes.

Energy when you actually lean on the grid. When the generator is down and the site draws standby power, that energy is billed at time-differentiated rates. An outage in a summer peak window and an outage on a mild winter night are very different invoices. The schedules also distinguish scheduled maintenance, which can be arranged for favorable windows, from forced outages, which arrive when they arrive.

The charges that never left. Customer charges, meter charges, and reactive demand charges continue. Under the PG&E schedules, billed maximum demand is the highest fifteen-minute average of the month, which means a single unplanned trip during a peak interval can set the month's demand consequences on the applicable schedule.¹

The federal backstop. Standby pricing is not lawless. For qualifying cogeneration and small power production facilities under federal law, rates for backup and maintenance power must be just, reasonable, and nondiscriminatory, and may not rest on an assumption, unless factual data supports it, that all self-generators' outages "will occur simultaneously, or during the system peak, or both."⁵ That principle, written in 1980, is why diversity allowances like the 85 percent factor exist in print. An owner who believes a standby rate assumes the impossible has a regulatory argument, not just a grievance.

Section 03The exemption lattice: real, narrow, conditional

Standby tariffs come with exemptions, and the exemptions are where models most often help themselves. Three examples sit on the face of the PG&E schedules, each with a different scope.^1,2

• A technology-criteria exemption. A customer on a time-of-use schedule whose generation technology meets the criteria defined in the utility's Electric Rule 1 for distributed energy resources is exempt from the standby reservation charges, but continues to pay the schedule's energy, reactive demand, customer, and meter charges. Relief from one line is not relief from the schedule.
• A statutory solar exemption. Solar customers on the net energy metering or net billing tariffs, or solar facilities at or under one megawatt serving on-site load with no more than incidental export, are exempt from standby charges under the tariff, with the statute doing the work. The exemption attaches to a configuration defined in law, not to the general virtue of any technology.
• A performance-based microgrid suspension. Customers operating a microgrid interconnected under Rule 21 whose generation and storage resources meet stated variability, emissions, and operational standards qualify for a suspension of standby reservation capacity charges, as directed by the regulator. A suspension conditioned on measured performance is not a repeal, and it can be lost.

Read as a set, these clauses teach the honest lesson. Combustion engines, turbines, fuel cells, linear generators, solar, and storage each map differently onto the lattice, and none escapes it by brand. Each technology has configurations that qualify for real relief and configurations that do not; the same machine can be exempt in one wiring diagram and fully charged in another. A model that assumes an exemption without naming the clause, and without verifying that the project as actually configured qualifies, has not found savings. It has borrowed someone else's tariff.

The most common flattery in a self-generation model is not an invented number. It is a missing line.

Section 04Departing load: the grid you left

The second family of charges answers a different question: what do you still owe for the years when the utility planned, procured, and built on the assumption that your load was staying. California's answer is the indifference principle. Customers who remain should be no worse off because you departed, and you should carry the costs incurred on your behalf before you left. Serving your own load with on-site generation is a form of departure, and the tariff books say so explicitly: PG&E's published departing-load provisions define customer generation departing load as load that replaces utility purchases with electricity from customer generation, alongside the municipal category for customers who move to a publicly owned utility.³

The framework decision is old enough to be settled law in spirit. In 2003 the state's regulator adopted cost-responsibility surcharge mechanisms for customer generation departing load, requiring qualifying departed load to carry its share of bond-related and transition costs from the energy crisis era, while adopting a lattice of exemptions: net-metered load was excused entirely, cleaner and smaller systems received partial relief subject to megawatt caps, and specific categories were grandfathered.⁴ The particular charges of 2003 have largely run their course. The architecture they established has not: departure triggers a determination, the determination assigns cost responsibility by category and vintage, and exemptions exist but must be individually earned.

What does the modern version look like? PG&E's current departing-load page lists the nonbypassable categories a departing customer may still be billed for: competition transition charges, the energy cost recovery amount, nuclear decommissioning, the wildfire fund charge, public purpose programs, and the power charge indifference adjustment, the last ensuring departed customers pay their share of generation costs contracted before they left.³ The same names appear inside the standby schedule's own unbundled rate tables, together with newer entries such as the new system generation charge, the wildfire hardening charge, and the recovery bond charge and its offsetting credit, each riding on delivered kilowatt-hours at whatever value the current sheet carries.¹ Some are fractions of a cent; some are credits in a given year. The sheet, not the narrative, carries the sign.

Two consequences matter for modeling. First, every kilowatt-hour the site still takes from the grid, including standby energy, carries the nonbypassable components of the day. Second, and less intuitively, a cost-responsibility determination can attach charges to kilowatt-hours the utility no longer delivers, because that is precisely what the departing-load mechanism is for. Whether a given project owes such charges, in which category, at which vintage, with which exemption, is a determination made under tariff rules and regulatory decisions. It is not a cell a seller's spreadsheet is entitled to set to zero.

Section 05Four flatteries

With the mechanisms named, the ways models ignore them become easy to spot. Four appear constantly.

Flattery one: blended-rate arithmetic. The model values every displaced kilowatt-hour at the site's current average all-in rate. But that average includes fixed delivery costs that do not scale down with displaced energy, nonbypassable components that continue on every kilowatt-hour still delivered, and, where a determination attaches, cost-responsibility charges that follow the departed load itself. A displaced kilowatt-hour avoids only its avoidable components. The honest version of this line requires decomposing the bill into what the project truly avoids and what it does not, using the current unbundled sheets, and the honest answer is smaller than the blended one. How much smaller is a site-specific finding, which is exactly why sellers prefer the blend.

Flattery two: the missing reservation line. The model contains no standby contract at all, or silently assumes an exemption it never names. If the site intends to lean on the grid when the machine is down, the reservation capacity, its voltage class, and the current charge belong in the pro forma as a named, sourced line. If the model instead relies on an exemption, the clause, its conditions, and the project's verified eligibility belong in the pro forma with it. Either is defensible. Neither can be implicit.

Flattery three: the perfect machine. The model runs the generator at its contractual availability every hour of every year, so no standby energy is ever purchased, no forced outage ever lands in a summer peak window, and no fifteen-minute interval ever sets a demand consequence. Real fleets have forced outage rates, and the difference between a scheduled October maintenance window and an unscheduled July afternoon trip is exactly the difference the standby schedule's time-differentiated rates are built to price. A defensible study models the worst plausible month, not the average one, and says which assumptions drive it.

Flattery four: the frozen tariff. The model treats today's bill as the after bill, or today's after-rates as permanent. The sheets cited in this paper took effect on March 1, 2026, and they will be revised again by the ordinary machinery of advice letters and decisions.¹ This cuts in both directions, and no one should pretend to know which way. The discipline is not prediction; it is dating. Every tariff-derived number in a study carries the effective date of its sheet, and the sensitivity section shows what happens to the decision, not just the returns, if the after-bill lines move.

One counterpoint belongs on the record, because neutrality requires it. Sometimes the after-bill really is small. A transmission-voltage site reserves almost no distribution and sees a reservation charge roughly an eighth of the secondary-voltage rate on the current sheets. A configuration that genuinely qualifies for the statutory or performance-based exemptions carries real, lawful relief. The error is not assuming the after-bill is large; it is assuming anything. The line must exist, sourced and dated, even when its value turns out to be modest. And a site that departs entirely, taking no utility service at all, has not escaped the question either: reliability then rests wholly on the machine and its maintenance plan, and departure itself is the very event the cost-responsibility framework was built to address.

Section 06Six questions for any post-build bill model

An owner does not need to be a tariff analyst to audit a model. Six questions, answerable in writing, separate a decision document from a brochure.

1. Which standby schedule applies, by name and effective date?If the answer is none, the model must say why: the named exemption clause, full departure, or a documented alternative arrangement. "It won't apply to us" is not a clause.
2. What reservation capacity is assumed, and who chose it?Contract reservation demand is a negotiated number with a monthly price. A model that never states it has either omitted the charge or reserved zero, and only one of those is a plan.
3. What does the worst month look like?A forced outage in a peak window, priced with the schedule's own time-differentiated rates and interval-demand definitions. If the model only shows the average year, ask for the bad July.
4. Which per-kilowatt-hour components survive displacement?Enumerate the nonbypassable components from the current unbundled sheets and value displaced energy net of what continues. The blend flatters; the decomposition informs.
5. Has a departing-load determination been modeled or requested?Category, vintage, and exemption eligibility, by clause. The utility makes this determination under its tariffs; the model's job is to anticipate it, not to wish it away.
6. Which numbers are firm, which are illustrative, and as of when?Tariff sheets carry effective dates. Every tariff-derived line in the model should carry one too, and every estimate should be labeled as an estimate.

Section 07The bill is published

Nothing in this paper is hidden knowledge. The standby schedules, the exemption clauses, the departing-load provisions, and the regulator's framework decisions are public documents, maintained online, revised on the record. The reason they are missing from so many models is structural rather than sinister: the party selling the machine is paid for the machine, and the after-bill is the part of the economics that machine revenue never touches.

That is also why the after-bill is a fair test of any analysis put in front of you, including ours. A study prepared on the owner's side of the table names the schedules, states the reservation assumption, decomposes the blended rate, models the bad month, and dates every number. The distance between the model and the meter is where self-generation projects go quietly wrong. Reading the tariff before you build costs attention. Discovering it after costs the difference.