Battery Energy Storage for Peak Demand Reduction: Sizing, Dispatch, and Economic Analysis

Published: June 2026 Technical Level: Advanced Category: Power Systems Design

Abstract

Peak demand reduction through battery energy storage is the most economically mature commercial BESS application in most utility service territories, driven by demand charges that represent 30 to 50 percent of a commercial customer's total electricity bill and that can be reduced by 20 to 35 percent with a properly sized and dispatched BESS. The engineering design problem is to determine the battery power and energy ratings that minimize the present value of the total electricity cost over the system's useful life, subject to battery physical constraints (SOC limits, C-rate limits, round-trip efficiency) and dispatch strategy constraints (forecast uncertainty, multiple daily peaks, coincident demand windows). This paper develops the load profile analysis methodology for characterizing a facility's peak demand structure, the deterministic and stochastic sizing methods, the model predictive control dispatch framework, and the financial analysis that translates physical performance into net present value. The interaction between peak shaving and secondary value streams (time-of-use energy arbitrage, demand response program participation) is addressed as a stacked revenue optimization problem with shared battery capacity constraints.

1. Introduction

A commercial electricity rate with a demand charge imposes a monthly fee proportional to the customer's maximum 15-minute or 30-minute average power demand during the billing period. Demand charges typically range from 8to8 to 25 per kW-month for commercial customers and from 15to15 to 35 per kW-month for industrial customers. For a commercial building with a 500 kW peak demand and a 15/kW−monthdemandcharge,theannualdemandchargeis15/kW-month demand charge, the annual demand charge is 90,000, a substantial fraction of the total electricity cost that can be reduced by flattening the demand peak.

The BESS peak shaving strategy is conceptually straightforward: charge the battery during periods when demand is low (typically overnight), then discharge the battery during periods when demand peaks (typically weekday afternoons in summer), reducing the net demand measured at the utility meter. The engineering challenge is executing this strategy reliably enough that the demand reduction is realized in every billing period, not just in some months. A BESS that is designed for the average peak event but undersized for the worst-case peak (a hot August afternoon with all HVAC systems running simultaneously, higher-than-expected occupancy, and a cloud-induced loss of onsite solar generation) will fail to shave the peak in the months when the demand charge is highest.

2. Load Profile Analysis

2.1 Peak Demand Characterization

The sizing analysis begins with at least 12 months of 15-minute interval load data from the facility's utility meter. The statistical characterization of this data produces three outputs: the distribution of daily peak demand values across all billing periods, the typical diurnal load profile for peak versus non-peak days, and the correlation between peak demand and identifiable drivers (outdoor temperature, day of week, occupancy level).

From this analysis, the engineer identifies the design peak demand: the demand level that the BESS must sustain the facility at or below, in all billing periods, to realize the target demand charge reduction. The design peak is not the historical maximum peak but rather a percentile of the peak demand distribution that balances the value of reducing that additional peak against the cost of the additional battery capacity required. For most commercial facilities, sizing to the 95th percentile of historical monthly peaks provides a practical design basis: the battery will successfully limit the demand charge in approximately 95 percent of billing periods, with occasional exceedances in the most extreme months.

The required demand reduction is:

ΔP=Pdesign−Ptarget\Delta P = P_{design} - P_{target}

Where PdesignP_{design} is the 95th percentile monthly peak demand and PtargetP_{target} is the target demand level at the utility meter, typically the historical 75th percentile or the customer's contracted demand level.

3. Battery Sizing

3.1 Deterministic Method

The deterministic sizing method selects the battery power and energy ratings based on the worst-case peak demand profile from the historical dataset. The battery power rating equals ΔP\Delta P in kW. The energy rating is determined by the duration for which the battery must sustain ΔP\Delta P of discharge power, which is the maximum duration over which the facility's demand exceeds PtargetP_{target} in any historical billing period.

From the 15-minute interval data, the engineer identifies the longest continuous period above PtargetP_{target} in any month of the dataset. If this maximum overshoot duration is Tpeak,maxT_{peak,max}, the required energy capacity is:

Ereq=ΔP⋅Tpeak,maxηdE_{req} = \frac{\Delta P \cdot T_{peak,max}}{\eta_d}

Where ηd\eta_d is the one-way discharge efficiency (0.92 to 0.95 for modern lithium-ion systems).

The deterministic method is conservative: it sizes for the worst observed peak and accepts that this worst-case sizing may provide excess capacity for typical peak events. The conservatism is appropriate for facilities where a single month of failed peak shaving would result in a demand ratchet charge (a penalty that locks in the higher demand charge for the following 11 months) that substantially erodes the annual savings.

As a worked case, suppose the 15-minute interval data for a manufacturing facility shows a target demand ceiling PtargetP_{target} set 150 kW below the historical monthly peak, so the required discharge power is ΔP=150\Delta P = 150 kW, and the longest continuous excursion above PtargetP_{target} in the three-year record is Tpeak,max=2.5T_{peak,max} = 2.5 hours. With a discharge efficiency of ηd=0.93\eta_d = 0.93, the required usable energy is:

Ereq=150⋅2.50.93=403 kWhE_{req} = \frac{150 \cdot 2.5}{0.93} = 403 \ \text{kWh}

The battery is therefore specified at 150 kW / 403 kWh: a power rating set by the peak reduction itself and an energy rating set by how long that reduction must be sustained. On a $20/kW-month demand tariff, the 150 kW reduction is worth 150×20×12=36,000150 \times 20 \times 12 = 36{,}000 dollars, that is $36,000 per year, and because the energy rating is fixed by the single worst excursion, sizing a few tenths of an hour short of Tpeak,maxT_{peak,max} is what causes a battery to exhaust mid-peak and forfeit the entire month's saving to a demand ratchet.

3.2 Stochastic Method

The stochastic sizing method treats the peak demand level and peak duration as random variables with distributions estimated from the historical data, and selects the battery ratings that maximize the expected annual savings net of battery cost at a target confidence level. This method produces a less conservative result than the deterministic method when the worst-case historical event is an outlier (a single extreme event that is unlikely to recur) and the cost of sizing for that event is disproportionate to the probability-weighted benefit.

The stochastic formulation requires a probability model for the monthly peak demand that captures the seasonal variation (summer peaks are typically higher than winter peaks for cooling-dominated facilities), the temperature sensitivity, and the inter-annual variability. For a 3-year historical dataset, fitting a normal distribution to the monthly peak demands by season is sufficient; for shorter datasets, the engineer should apply a conservative uncertainty margin to account for the limited sample size.

4. MPC Dispatch Framework

The model predictive control (MPC) dispatch framework solves an optimization problem at each 15-minute interval to determine the battery charge/discharge schedule for the next 24 hours, based on a load forecast for that period:

min⁡Pc(k),Pd(k)[Cdemand⋅max⁡k∈BPgrid(k)+∑k=196cTOU(k)⋅Pgrid(k)⋅Δt]\min_{P_c(k), P_d(k)} \left[ C_{demand} \cdot \max_{k \in \mathcal{B}} P_{grid}(k) + \sum_{k=1}^{96} c_{TOU}(k) \cdot P_{grid}(k) \cdot \Delta t \right]

subject to: Pgrid(k)=Pload(k)−Psolar(k)+Pc(k)−Pd(k)P_{grid}(k) = P_{load}(k) - P_{solar}(k) + P_c(k) - P_d(k) SOC(k+1)=SOC(k)+ΔtEcap[ηcPc(k)−Pd(k)ηd]SOC(k+1) = SOC(k) + \frac{\Delta t}{E_{cap}}\left[\eta_c P_c(k) - \frac{P_d(k)}{\eta_d}\right] SOCmin≤SOC(k)≤SOCmax,0≤Pc(k)≤Pc,max,0≤Pd(k)≤Pd,maxSOC_{min} \leq SOC(k) \leq SOC_{max}, \quad 0 \leq P_c(k) \leq P_{c,max}, \quad 0 \leq P_d(k) \leq P_{d,max}

Where: CdemandC_{demand} is the demand charge rate in $/kW-month.

B\mathcal{B} is the set of intervals in the demand billing window (typically 8 AM to 8 PM on weekdays).

cTOU(k)c_{TOU}(k) is the time-of-use energy rate at interval kk in $/kWh.

EcapE_{cap} is the battery energy capacity in kWh.

ηc,ηd\eta_c, \eta_d are the charging and discharging efficiencies.

The demand charge term drives the optimizer to keep max⁡kPgrid(k)\max_k P_{grid}(k) as low as possible; the TOU energy term drives the optimizer to charge during low-rate periods and discharge (or avoid charging) during high-rate periods. The two objectives are complementary in most rate structures but can conflict when the on-peak energy period does not coincide with the demand billing window.

The dispatch behavior that produces the demand-charge savings is illustrated in Figure 1, which shows the original facility load and the net load after battery discharge against the target demand limit.

Peak shaving dispatch over a 24-hour period. The horizontal axis is time of day in hours and the vertical axis is demand in kilowatts. The battery discharges during the afternoon peak to hold the net load at the demand limit.

Figure 1. Peak shaving dispatch over a 24-hour period. The horizontal axis is time of day in hours and the vertical axis is demand in kilowatts. The battery discharges during the afternoon peak to hold the net load at the demand limit, with the shaded area representing the energy delivered from storage. The engineer should observe that the required battery energy is the area between the original load and the demand limit during the peak window, and that an undersized battery is revealed by the net-load curve breaching the limit before the peak ends.

5. Economic Analysis

The net present value of a peak shaving BESS installation is:

NPV=−Cinstalled+∑y=1LΔCdemand(y)+ΔCenergy(y)(1+r)y−Creplacement(1+r)L/2NPV = -C_{installed} + \sum_{y=1}^{L} \frac{\Delta C_{demand}(y) + \Delta C_{energy}(y)}{(1+r)^y} - \frac{C_{replacement}}{(1+r)^{L/2}}

Where: CinstalledC_{installed} is the total installed system cost in dollars (typically $300 to $600/kWh for commercial systems in 2025–2026).

ΔCdemand(y)\Delta C_{demand}(y) is the annual demand charge savings in year yy, accounting for load growth and utility rate escalation.

ΔCenergy(y)\Delta C_{energy}(y) is the annual energy cost savings from TOU arbitrage in year yy.

rr is the discount rate.

LL is the system useful life in years (typically 10 to 15 years for LFP chemistry).

CreplacementC_{replacement} is the mid-life battery replacement cost if required.

For a commercial building with 500 kW peak demand, $15/kW-month demand charge, 25% demand reduction ($22,500/year savings), and a 500 kWh LFP BESS at $500/kWh installed ($250,000): the simple payback is 11.1 years. With investment tax credit (30% under current IRA provisions), the net installed cost is $175,000 and the payback is 7.8 years. At a 7 percent discount rate over a 15-year life, the NPV is approximately $28,000, positive but modest. Adding TOU energy savings of $8,000/year improves the NPV to approximately $80,000 and the payback to 5.8 years.

Related Work

The analysis in this paper connects to several companion studies in this library. Readers concerned with the upstream and downstream engineering will find Battery Energy Storage for Backup Power develops a closely related aspect of the same problem, while Battery Energy Storage Systems extends the treatment into an adjacent domain. For the broader methodological context, Battery Energy Storage Systems Integrated with Solar PV provides complementary depth.

Conclusion

Peak demand reduction is the most economically mature commercial BESS application because demand charges represent 30 to 50 percent of a typical commercial bill and a properly sized and dispatched battery can reduce them by 20 to 35 percent, and the sizing framework developed in this paper shows that the design problem is fundamentally one of minimizing the present value of total electricity cost subject to battery physical and dispatch constraints. The comparison between the deterministic and stochastic sizing methods establishes the central engineering conclusion: deterministic sizing against a historical peak underestimates the battery energy required because forecast uncertainty and the occurrence of multiple non-coincident daily peaks demand reserve capacity that a single worst-case day does not reveal. The model predictive control dispatch framework is what realizes the demand-charge savings in practice, because peak shaving requires anticipatory discharge timed against a demand interval that has not yet completed. For the practicing engineer, the operative takeaway is that peak-shaving BESS sizing must account for forecast risk explicitly, and that the economic result is governed by dispatch quality as much as by installed capacity.

References

[1] NFPA 70, National Electrical Code, Article 706, 2023 edition, NFPA, 2023.

[2] IEEE Standard 1547-2018, Standard for Interconnection and Interoperability of Distributed Energy Resources, IEEE, 2018.

[3] EPRI, The Economics of Battery Energy Storage for Commercial Customers, EPRI Technical Report 3002019123, 2022.

[4] Rocky Mountain Institute, The Economics of Battery Energy Storage, RMI, 2015.

[5] J. Rawlings, D. Mayne, and M. Diehl, Model Predictive Control: Theory, Computation, and Design, 2nd ed., Nob Hill Publishing, 2017.

[6] D. Levi and M. Noy, "Battery Storage Dispatch Optimization for Commercial Buildings," Energy and Buildings, vol. 182, pp. 112–123, 2019.

[7] Lawrence Berkeley National Laboratory, Tracking the Sun: Pricing and Design Trends for Distributed Photovoltaic Systems in the United States, LBNL, 2024.

[8] NFPA 855, Standard for the Installation of Stationary Energy Storage Systems, 2023 edition, NFPA, 2023.