The Capacity
Ladder

A constrained site often begins by asking how to buy more power. The better first question is whether the existing service is truly full. Billing demand, electrical equipment ratings, operating peaks, and utility-authorized capacity are different facts. Until they are reconciled, a capacity shortfall is still a hypothesis.

Section 01Capacity is not one number

Four documents can display four different values for the same site: the utility service agreement, the main switchboard nameplate, the highest interval on the bill, and the facilities team's connected-load list. None is automatically the site's usable headroom. The utility governs what it will serve. Customer-owned transformers, conductors, switchgear, protection, and distribution equipment govern what can be carried safely. Operations govern which loads actually coincide. The tariff governs how demand is measured and billed.

Current PG&E Schedules B-19 and B-20 define maximum demand for affected accounts as the highest average over a 15-minute interval in the billing month, with specified exceptions for severe fluctuations and certain equipment.^1,2 That interval is commercially important, but it is not an engineering study and it is not permission to add load. A site that has never crossed a particular billing level may still be limited by a transformer, protective device, conductor, voltage condition, or utility operating constraint. Conversely, a site with one ugly billing peak may have more controllable headroom than its monthly bill suggests.

PG&E Rule 2 sets service characteristics and general load limitations by voltage and arrangement, and requires prior utility approval for installations outside stated limits or with larger configurations.³ Rule 16 separately defines the utility's service facilities and the service-related equipment required on the applicant's premises.⁴ Those boundaries matter. A spare breaker position is not spare utility capacity. A transformer nameplate is not an operating entitlement. A low bill is not proof that the owner can energize the next process line.

Section 02Build the baseline before climbing

The first rung is evidence, not equipment. Assemble at least one complete operating cycle of utility bills and interval data; reconcile every meter to a service agreement and physical location; obtain the current one-line diagram; record transformer, switchgear, conductor, protection, and major-load ratings; and align those records with production, weather, maintenance, and shift logs. If a planned load has not arrived, build a separate forecast rather than folding it into history.

PG&E states that electric interval data is generally available at 15-minute resolution for most commercial customers through Share My Data, with billing and account information available when authorized. It also warns that reads can occasionally be missing or corrected through its validation process.⁵ That is enough to start, but not enough to finish. Revenue-meter data shows the net service profile. It does not reveal which chiller, compressor, charger, pump, furnace, or tenant caused the peak. Temporary submetering and operating logs are what turn a meter trace into an action plan.

Two views of the same data are especially useful. A chronological profile shows when a peak happened and what else was happening. A load-duration curve sorts all intervals from highest to lowest and shows whether the constraint is a narrow spike or a broad plateau. Load factor, defined by the U.S. Energy Information Administration as average load divided by peak load over a specified interval, provides another compact view.⁶ A low load factor can indicate a peaky site with shaping opportunity. It can also describe a legitimately intermittent process. The metric identifies a question; it does not supply the answer.

The baseline should end with a signed fact table: each service and meter; the responsible utility account holder; the verified peak and when it occurred; the customer equipment that carries it; the critical loads that cannot move; the loads that may move; and every uncertainty. This table prevents the most common capacity error, which is comparing a utility number from one boundary with an equipment number from another.

Section 03Rung one: eliminate false scarcity

False scarcity is capacity that appears unavailable because the records, controls, or operating sequence are wrong. It is recovered through correction rather than capital.

Start with the account. Confirm that the rate schedule, meter mapping, demand interval, service voltage, and power-factor treatment match the actual service. Current B-19 and B-20 bills can include separate maximum-demand, time-period demand, energy, and power-factor components.^1,2 A lower charge is not the same as more physical capacity, but a correct tariff model shows which intervals actually matter commercially.

Then inspect coincidence. Large motors, compressors, refrigeration racks, electric boilers, vehicle charging, and process heating are often started by independent local schedules. Each schedule may be sensible alone while their coincidence creates the monthly maximum. Staggering starts can reduce the integrated peak, but the sequence must also be checked for instantaneous current, voltage drop, protection coordination, process safety, and recovery behavior. A smoother 15-minute trace can still hide a damaging short event.

Reactive demand deserves the same discipline. Correcting poor power factor can reduce current through customer equipment and can change bills under applicable tariffs. It does not reduce the real power required by the process, and it does not by itself enlarge the utility's authorized service. Treat it as an electrical and tariff measure, with a qualified engineer reviewing resonance, harmonics, switching, and protection before equipment is selected.

False scarcity can also run the other direction. A demand-limiting controller may keep the revenue meter below a setpoint during ordinary operation, but a failed sensor, communications fault, maintenance bypass, or simultaneous restart can remove that protection. Any capacity credited to control must therefore have a defined failure state, operator response, alarm, and tested fallback.

Section 04Rung two: remove demand the site does not need

Efficiency is the most durable form of recovered capacity because the avoided load does not need to be shifted, stored, or generated later. Candidate measures include correcting simultaneous heating and cooling, resetting air and water temperatures, repairing compressed-air leaks, tuning refrigeration controls, matching pumps and fans to actual flow, replacing persistently loaded inefficient equipment, and shutting down idle auxiliaries.

The case for efficiency is strong when it lowers the site's coincident peak while preserving production, comfort, safety, and product quality. The case against is equally important. Some measures save annual energy but miss the binding peak. Others reduce redundancy, narrow an environmental tolerance, add maintenance, or produce savings only under modeled behavior that operators cannot sustain. The Department of Energy's guidance on grid-interactive efficient buildings emphasizes both reducing waste and managing demand against operational constraints.⁹ A capacity study should therefore report the measure's effect at the constrained interval, not only annual consumption.

Commissioning comes before replacement. A sensor that reads incorrectly can command both heat and cooling. A failed damper can make a fan retrofit look necessary. An override left in place after maintenance can erase years of control logic. Correct the operating system, document the new sequence, and measure a representative period before treating the recovered demand as available for a new load.

Section 05Rung three: schedule what can move

Some demand is necessary but not time-critical. Vehicle charging may need a departure state rather than an immediate start. A water tank may need a level band rather than a fixed pump hour. Refrigeration can sometimes pre-cool within product limits. Batch equipment may have a completion deadline but flexibility inside a shift. Thermal processing, material handling, defrost, cleaning, and some computing work can have similar operating envelopes.

California's demand-flexibility proceeding treats load management as a resource that includes price response, demand response, and coordinated end-use control.⁷ Lawrence Berkeley National Laboratory organizes demand flexibility into four functions: shape the normal profile, shift energy in time, shed load during defined events, and provide faster modulation.⁸ For an owner trying to admit new load behind an existing service, shape and shift usually come first. They change routine coincidence without depending on an emergency dispatch.

The honest case against scheduling is operational. A delayed batch can miss shipping. A compressed maintenance window can raise labor cost. Pre-cooling can affect humidity or product conditions. Deferred charging can leave a vehicle unavailable. Several deferred loads can rebound at once and recreate the peak later. Every flexible load needs six written fields: earliest start, latest finish, minimum run, maximum interruption, recovery profile, and protected operating condition.

External demand-response programs can add value after the internal envelope is proven, but enrollment obligations, baselines, dispatch frequency, and current compensation must be checked directly against the applicable program. Do not count program revenue, bill savings, or service capacity until the relevant terms and site performance are verified.

Section 06Rung four: store the service

Storage does not create energy. It moves demand from one interval to another. That distinction makes it useful for a narrow service constraint and dangerous when it is presented as unlimited capacity.

Thermal storage is often the first storage screen where cooling, heating, hot water, or refrigeration drives the peak. The Department of Energy describes thermal storage as charging a thermal medium when power is available and discharging it later with a lower electrical requirement at the constrained time.¹⁰ Its case for is direct: it addresses the thermal load without converting stored electricity back into heat or cooling at the peak. Its case against is site specificity. Space, piping, temperature requirements, losses, plant integration, water treatment, and control quality can defeat an otherwise attractive model.

Battery storage is faster and can serve electrical loads across the site. It can discharge into a short peak, support a controlled demand ceiling, and coordinate with solar. Its counterweight is duration, recharge demand, conversion loss, degradation, fire and electrical requirements, replacement planning, and the possibility that several long peaks exhaust the stored energy. The dispatch model must preserve any reserve assigned to resilience rather than quietly spending it on bill management.

Solar earns a separate screen. Daylight production can reduce net service demand when output coincides with the site's binding interval. It cannot be credited as firm headroom when the peak occurs after sunset, under poor weather, or during a maintenance outage. The Department of Energy notes both solar's time and weather variability and storage's ability to shift some production to a later demand period.¹¹ The case for solar is mature, fuel-free energy at the site. The case against is noncoincidence and variability. Pairing it with storage may improve the profile, but the pair still requires an explicit worst-day dispatch and interconnection case.

Section 07Rung five: reshape supply without choosing a favorite

If controls, efficiency, scheduling, and storage do not close the gap, on-site generation may reduce net utility demand. It also changes the project. Fuel, emissions, interconnection, protection, ventilation, noise, maintenance, staffing, standby treatment, and outage behavior enter the decision. The U.S. Environmental Protection Agency's technology catalog documents materially different operating characteristics for reciprocating engines, combustion turbines, microturbines, and fuel cells.¹² California's public research record also includes field demonstrations of linear generation, while its younger operating base still warrants configuration-specific evidence.¹³

California's Air Resources Board maintains a distributed-generation certification program for certain technologies, but local permit applicability remains a project-specific question.¹⁴ Grid-connected generation also enters the applicable interconnection process. PG&E's current generator application materials require a one-line diagram, site plans, equipment ratings, transfer-scheme information, and protective-relay documentation where applicable.¹⁵ Non-export controls can change the technical screen; they do not erase the need to establish the correct interconnection route.

The table does not identify a winner. It identifies the proof burden. A technology earns capacity credit only after its net output, duty cycle, downtime, degradation, maintenance, fuel, permit, and interconnection conditions are represented. Nameplate output is not delivered headroom.

Section 08Rung six: obtain the operating envelope in writing

After the internal ladder has been climbed, return to the utility with a narrower, better-supported request. Provide the service agreement and meter identifiers; current one-line; verified interval profile; planned loads and ramp dates; control and storage sequence; expected maximum import; failure behavior; and any generation configuration. Ask the utility to identify the service arrangement it recognizes, the facilities and study assumptions that govern the request, and any condition that would require an upgrade or new application.

This is not a negotiation trick. It makes the utility question answerable. “How much power do we have?” is ambiguous. “Will the documented service and utility facilities support this dated import profile under this operating sequence?” is a defined engineering and commercial question.

A written answer can still be conditional. The utility may require studies, field verification, protection changes, a service modification, or construction. The customer's engineer may identify a lower limit in customer-owned equipment. Record each condition separately. The capacity available for planning is the lowest verified limit across the utility boundary, customer electrical system, and operating control case.

Do not convert a successful month into a guarantee. Weather, production mix, maintenance, tenant behavior, control failure, and new loads can move the peak. The operating envelope needs alarms, ownership, periodic testing, and a revalidation trigger whenever a major load, service component, tariff, control sequence, or utility condition changes.

Section 09The capacity-ladder test

A study is ready to support a decision only when the following questions can be answered without inference.

1. What is the exact constraint?State whether it is utility authorization, customer electrical equipment, operating coincidence, bill exposure, or a combination. Attach the evidence for each.
2. Are all meters and boundaries reconciled?Map each account, meter, transformer, switchboard, tenant, and major load to the physical one-line and responsible owner.
3. What happened during the binding intervals?Join interval data to production, weather, maintenance, shift, and control records. A timestamp without an operating cause is not a plan.
4. What demand can disappear rather than move?Measure commissioning and efficiency measures at the constrained interval before assigning recovered capacity.
5. What load can move, and what is protected?Document start and finish limits, interruption, recovery, quality, safety, and the consequence of a missed schedule.
6. How does storage fail?Model duration, recharge, reserve, loss, degradation, repeated peaks, communication loss, and the control fallback.
7. Has every supply path received an honest screen?Grid service, solar, batteries, engines, turbines, microturbines, fuel cells, and linear generators each need a case for, a case against, and a proof requirement.
8. Which limit is confirmed in writing?Keep utility conclusions, engineering conclusions, and operating estimates separate. Use the lowest verified limit for the decision.

Section 10Buy capacity only after recovering the decision

The capacity ladder will not always avoid a service upgrade or generating project. It is not supposed to. Its purpose is to prevent the owner from buying the wrong amount, at the wrong boundary, for a peak that could have been measured, removed, or moved.

Sometimes the answer will be commissioning and a revised operating sequence. Sometimes it will be thermal storage, a battery, solar, generation, a utility modification, or a portfolio. Sometimes the existing service is genuinely full and the upgrade case becomes stronger because the alternatives have been tested. Each is an acceptable outcome when the evidence leads there.

The owner's advantage is sequence. Measure before modeling. Remove before shifting. Shift before storing. Store before supplying. Verify every interface before capital moves. That sequence does not promise more capacity. It produces a bounded, written decision about how much usable headroom exists, what it depends on, and what must happen next.