Commercial electricity bills have two components that most building operators conflate into a single line item called "electricity cost." The first is the energy charge — the commodity cost of kilowatt-hours consumed, typically $0.06–$0.14/kWh depending on market and rate class. The second is the demand charge — a monthly fee assessed on the single highest 15-minute average power draw recorded during the billing period, typically $8–$22/kW in mid-Atlantic and Southeast markets, and as high as $28–$35/kW for industrial customers in constrained urban distribution zones.
The demand charge does not scale with energy consumption. A building that uses 200,000 kWh in a month and another building that uses 180,000 kWh in the same month can have identical demand charges if they hit the same 15-minute peak. This asymmetry is intentional: utilities use demand charges to recover the fixed cost of distribution infrastructure that must be sized for peak load conditions, not average load. For commercial building operators, this creates a specific optimization target — not reducing energy, but reducing peak 15-minute demand during the utility's billing measurement windows.
How Demand Charges Actually Work
Most commercial utility rate schedules define demand charges in one of three structures. The simplest is a flat monthly peak: the utility records the highest 15-minute average kW in the billing period, multiplies by the demand charge rate, done. More complex rate structures add time-of-use demand tiers — peak demand occurring between 2–8 PM on weekdays in summer may carry a $22/kW charge while off-peak demand carries $6/kW. A third structure, used in high-density urban markets and for large commercial accounts, applies both a coincident demand charge (your demand as a fraction of the system-wide peak, relevant to utilities with ratchet clauses) and a non-coincident demand charge.
What makes demand charge management difficult in practice is the measurement interval. The utility's revenue-grade meter records power in 15-minute intervals. Within any given 15-minute window, the building's instantaneous power can vary substantially — a single elevator start, a large rooftop unit compressor cycling on, or a commercial kitchen's morning startup sequence can create a brief but real spike that elevates the entire 15-minute average. And the monthly peak is determined by only that highest interval. Eleven months of excellent demand management can be undone by one operational anomaly on the worst possible afternoon.
The 15-Minute Visibility Problem
Consider a typical 180,000 sq ft Class B office building in the PJM territory — roughly 800 kW peak demand under a rate schedule with an $18/kW demand charge. The building's annual demand charge exposure is approximately $172,800. In most of these buildings, the facility manager sees the peak demand number once per month on the utility invoice. There is no real-time visibility into the current 15-minute interval's demand trajectory.
This is the fundamental operational gap. Without sub-interval demand telemetry — ideally 1-minute or better resolution from the building's main electrical panel — the facility manager cannot intervene before a new monthly peak is set. By the time an anomalous demand event is visible on the invoice, it has already happened. The operational response is always after the fact.
The building automation system (BAS) is not a substitute. Most Tridium N4 or Johnson Controls Metasys installations poll BACnet/IP devices on 15–30 second cycles and display trending data for setpoint management and fault detection. They are not architected to perform real-time demand forecasting against a rolling 15-minute window, and they do not generate alerts when the current 15-minute interval is trending toward a new peak.
Load Factor as the Primary Metric
Load factor — the ratio of average demand to peak demand over a billing period — is the metric that reveals how much demand charge optimization headroom a building has. A load factor of 0.85 means the building's average demand is 85% of its peak; a load factor of 0.40 means there's a large spike relative to average consumption. Most commercial office buildings run load factors between 0.45 and 0.65. A low load factor indicates either predictable peak patterns (good — they're optimizable) or chronic spike events from operational irregularities (bad — they need to be identified and suppressed).
Mathematically, every kW of demand reduction from a new monthly peak translates to $18 in annual savings on an $18/kW tariff (one month saved) — or $216 in savings if the peak would have persisted as the monthly maximum across the full 12-month measurement period. In buildings where the peak month is summer and the demand charge applies to coincident peak hours, a 100 kW peak reduction saves $1,800 per month during the 4–5 high-risk summer months: $7,200–$9,000 annualized from a single demand event avoidance.
Practical Demand Shaving Strategies
Demand charge reduction comes from three categories of building response: load shifting, load shedding, and behind-the-meter generation or storage dispatch.
Thermal Load Shifting
HVAC represents 40–60% of commercial building peak demand in climate-controlled office buildings. Chilled water plants, rooftop units, and AHU fan arrays are all significant demand contributors and, critically, they have thermal inertia that enables pre-conditioning strategies. By cooling a building to 70°F before a predicted demand peak window — overriding the standard 72°F setpoint during off-peak hours — the building accumulates thermal mass that allows HVAC to reduce output during the peak window. A well-insulated 180,000 sq ft building can sustain a 1°F setpoint increase for 20–30 minutes before occupant complaints begin, which is precisely the duration needed to let a high-demand 15-minute interval pass without HVAC contributing at full load.
Chilled water storage takes this further. Facilities with chilled water tanks can charge storage during overnight off-peak hours ($0.06/kWh energy, low demand risk) and discharge during afternoon peak windows. A 500-ton-hour chilled water tank feeding a 400-ton chiller plant can defer roughly 400 kW of chiller demand for 75 minutes — enough to flatten a summer afternoon peak across 3–4 consecutive demand intervals.
Coincident Load Coordination
Many peak demand spikes are not caused by any single large load but by the coincidence of multiple moderate loads — a chiller starting, an elevator bank in morning rush, the kitchen exhaust hood activating — that happen to overlap within the same 15-minute interval. Preventing these coincidences requires demand-aware sequencing: staggering the start times of large loads so that no two compressor starts or elevator banks activate simultaneously. This is operationally straightforward once you have real-time demand visibility, but almost impossible to manage manually across a multi-tenant building where tenants control their own equipment starts.
Automated demand sequencing — where the BMS receives a demand advisory signal and delays non-critical load activations when the current interval is already trending above a threshold — reduces peak demand by 8–15% in buildings with high coincidence exposure without any impact on occupant comfort or operational function.
Battery Storage Dispatch
Commercial battery energy storage systems (BESS) have become cost-effective for demand charge reduction at buildings above approximately 500 kW peak demand, where the monthly demand charge savings can justify BESS capital cost within 5–8 years. The dispatch logic is straightforward in concept: charge during low-demand periods (typically overnight, 10 PM–6 AM) and discharge when demand approaches a threshold. The challenge is predicting when that threshold will be reached, which requires load forecasting rather than reactive response.
A reactive BESS that fires when demand hits 700 kW is often too late. If the 15-minute interval started at 680 kW and a sudden load event pushes instantaneous demand to 820 kW, the average for that interval may already be committed to a new peak before the BESS response has time to pull the interval average back below 700 kW. Predictive dispatch — forecasting the likely interval average 5–10 minutes ahead and pre-dispatching the BESS when the forecast exceeds threshold — prevents peak events rather than reacting to them.
We're not saying that reactive BESS dispatch is worthless — it still catches a meaningful fraction of peak events and provides resilience value beyond demand charge reduction. The argument is that predictive dispatch doubles or triples the peak-shaving effectiveness of the same BESS hardware, and that the marginal cost of adding forecasting to a BESS that's already installed is far lower than the value of the avoided peaks.
Rate Schedule Selection and Tariff Engineering
Before deploying any hardware or software, building operators with multiple rate schedule options should confirm they're on the optimal tariff for their load profile. This is frequently overlooked. A building with a highly peaked load profile — low load factor, sharp morning or afternoon spikes — may benefit from a time-of-use demand rate where only on-peak demand is billed at the high rate, rather than a flat monthly peak rate where any interval in the month counts. Conversely, a building with consistent flat load may prefer the flat rate's lower off-peak exposure.
In many mid-Atlantic utility territories (PPL, Dominion, Duke), commercial customers can request a tariff review and rate change once per year. An analysis comparing the building's actual 12 months of 15-minute interval data against all available rate schedules often reveals $15,000–$40,000 in annual savings simply from rate migration, without any operational changes. The interval data required for this analysis is typically available from the utility's customer portal upon request, but requires AMI meters (advanced metering infrastructure) — which are now standard for most commercial accounts above 50 kW.
Building Across the Portfolio
Demand charge optimization at scale — across a portfolio of 10, 50, or 200 buildings — requires data infrastructure that most building operators don't currently have. Each building's peak demand interval may occur at a different time, driven by its specific use type, occupancy pattern, and local weather. Portfolio-level demand management needs per-building interval telemetry, not monthly invoice aggregation.
The operational leverage at portfolio scale comes from cross-building coordination. If buildings A, B, and C are all approaching demand peaks in the same ISO dispatch hour — perhaps because a regional heat wave is driving simultaneous HVAC peak loads across the portfolio — coordinating BESS dispatch and HVAC pre-conditioning across all three buildings simultaneously provides a larger demand response capacity than any single building can deliver. This is the building-blocks logic of virtual power plant (VPP) aggregation, applied to demand charge rather than market revenue.
Portfolio operators who implement interval telemetry and coordinated demand management routinely achieve 20–35% reductions in blended portfolio demand charge costs within 6–12 months of deployment. That's a $172,800/year exposure on a single 800 kW building becoming $112,000–$138,000 — and the return scales with portfolio size because the coordination benefits compound as more buildings are added to the managed pool.