DEV Community: Tinashe Nedi

I Reverse-Engineered NYC Local Law 97. Here's What 40% of Buildings Are Getting Wrong.

Tinashe Nedi — Fri, 15 May 2026 17:34:27 +0000

LL97 Penalties Aren’t What Most Calculators Show

The $268/Tonne Problem

NYC Local Law 97 is one of the most aggressive building performance standards in the United States. Buildings larger than 25,000 square feet can be fined $268 per metric tonne of CO₂e above their annual emissions limit.

After reviewing hundreds of LL97 estimates, I noticed the same issue repeatedly appearing in reports and calculators: the carbon math is often wrong.

Most tools use the EPA eGRID electricity emissions factor for NYISO (0.000371 tCO₂e/kWh).

LL97 does not use EPA eGRID values.

That difference alone can completely change a building’s projected penalties.

What LL97 Actually Uses

According to 1 RCNY §103-14 Table 1, the official LL97 emissions coefficients are:

Fuel Source	LL97 Coefficient	EPA eGRID (NYISO)	Difference
Electricity	`0.000288962 tCO₂e/kWh`	`0.000371 tCO₂e/kWh`	-22%
Natural Gas	`0.00005311 tCO₂e/kBtu`	`0.0000531 tCO₂e/kBtu`	≈0%
#2 Fuel Oil	`0.00007421 tCO₂e/kBtu`	`0.0000749 tCO₂e/kBtu`	-1%

The biggest discrepancy is electricity. LL97’s electricity coefficient is roughly 22% lower than the EPA value most calculators rely on.

Why the Numbers Are Different

LL97 was written using New York City’s 2018 grid baseline.

EPA eGRID values are updated regularly to reflect the current generation mix. Since 2018, NYC’s electricity supply has shifted, with greater reliance on natural gas and fewer hydroelectric imports. As a result, EPA values now show higher emissions.

LL97 intentionally locked in the earlier baseline so building owners would not be penalized for grid changes outside of their control.

Real Example: 75,000 SF Office Building

Building Information

Annual electricity usage: 1,250,000 kWh
Annual natural gas usage: 8,500 therms
Building occupancy: Office (Group B)
Building size: 75,000 SF

2024–2029 LL97 Limit

0.00536 tCO₂e/sf

Incorrect Calculation Using EPA eGRID

elec_co2 = 1_250_000 * 0.000371
gas_co2 = 850_000 * 0.00005311

total_co2 = elec_co2 + gas_co2  # 508.6 tCO₂e

intensity = total_co2 / 75_000  # 0.00678 tCO₂e/sf

limit = 0.00536

excess = (intensity - limit) * 75_000  # 106.5 tCO₂e

penalty = excess * 268  # $28,542/year

Using EPA coefficients produces a projected penalty of:

$28,542 per year

Correct Calculation Using Official LL97 Coefficients

elec_co2 = 1_250_000 * 0.000288962
gas_co2 = 850_000 * 0.00005311

total_co2 = elec_co2 + gas_co2  # 406.3 tCO₂e

intensity = total_co2 / 75_000  # 0.00542 tCO₂e/sf

limit = 0.00536

excess = (intensity - limit) * 75_000  # 4.5 tCO₂e

penalty = excess * 268  # $1,206/year

Using the actual LL97 coefficients changes the result dramatically:

$1,206 per year

That is a difference of more than:

$27,000 annually

from the exact same building data.

The 2030 Compliance Cliff

LL97 has two major compliance periods:

2024–2029
2030–2034

The second period is where the law becomes significantly more aggressive, with emissions limits dropping by roughly 40–50% depending on occupancy type.

For this same office building:

Compliance Period	Estimated Penalty
2024–2029	`$1,206/year`
2030–2034	`$41,832/year`

Many calculators fail to model this transition correctly.

Getting LL97 Calculations Right

1. Use Official LL97 Coefficients

LL97_COEFFICIENTS = {
    "electricity": 0.000288962,
    "natural_gas_kbtu": 0.00005311,
    "fuel_oil_2": 0.00007421,
    "fuel_oil_4": 0.00007753,
    "district_steam": 0.00004489
}

These values come directly from 1 RCNY §103-14 Table 1.

2. Identify the Correct Occupancy Group

LL97 emissions limits are occupancy-specific.

An office building, multifamily property, hospital, hotel, or retail space will all have different allowable emissions intensities under Table 2.

Incorrect occupancy classification can invalidate the entire calculation.

3. Include DER (Distributed Energy Resource) Deductions

LL97 allows buildings to deduct qualifying renewable generation from their reported emissions.

For example:

der_deduction_tco2e = annual_solar_kwh * 0.000288962

A 50 kW rooftop solar installation generating 65,000 kWh/year would reduce emissions by:

deduction = 65_000 * 0.000288962  # 18.8 tCO₂e/year

penalty_saved = 18.8 * 268  # $5,038/year

Over 25 years, that becomes:

$125,950 in avoided LL97 penalties

before even considering utility savings.

API Example

The Energy Audit Automation API handles LL97 calculations automatically using the correct coefficients and occupancy rules.

Request

curl -X POST https://energy-audit-api.p.rapidapi.com/audit/full \
  -H "X-RapidAPI-Key: YOUR_KEY" \
  -d '{
    "property_type": "office",
    "sq_ft": 75000,
    "annual_kwh": 1250000,
    "annual_therms": 8500,
    "city": "nyc",
    "grid_region": "NYISO"
  }'

Response

{
  "ll97": {
    "emissions_limit_2024": 0.00536,
    "your_emissions": 0.00542,
    "excess_tco2e": 4.5,
    "penalty_2024_usd": 1206,
    "penalty_2030_usd": 41832,
    "compliant_now": false,
    "compliant_2030": false,
    "source": "1 RCNY §103-14 Table 1 & 2"
  }
}

Red Flags in LL97 Proposals

Watch for these signs when reviewing vendor reports or compliance tools:

“EPA emission factor” referenced without LL97 citations
No source documentation for penalty calculations
Identical penalties shown for both 2024 and 2030 periods
No occupancy group classification included
No explanation of DER deductions or renewable offsets

Final Takeaway

A surprising number of LL97 estimates are based on the wrong carbon coefficients.

If a calculator or proposal does not explicitly reference 1 RCNY §103-14, there is a good chance the penalty estimate is inaccurate.

Under LL97, the difference between using EPA values and the actual legal coefficients can mean the difference between compliance and tens of thousands of dollars in annual penalties.

Why I stopped using flat $/kWh to size commercial battery storage.

Tinashe Nedi — Tue, 31 Mar 2026 13:03:58 +0000

I've been building energy APIs for about four years. The thing that kept bothering me wasn't the code — it was watching tools I respected give completely wrong BESS cost estimates because they multiplied system size by a flat dollars-per-kilowatt-hour figure.

That's not how battery storage costs work. And it matters a lot when someone is deciding whether to spend $800,000 on a system.

The problem with flat $/kWh

Most BESS calculators I've seen do something like this:


# What most tools do — this is wrong
system_cost = capacity_kwh * cost_per_kwh  # e.g. 500 kWh * $400 = $200,000

The issue is that battery storage has two fundamentally different cost components that scale differently:

Energy cost — scales with kWh (how much you can store)
Power cost — scales with kW (how fast you can charge/discharge)

A 500 kWh system with a 250 kW inverter has a completely different cost structure than a 500 kWh system with a 500 kW inverter. The flat $/kWh model treats them identically.

The NREL ATB two-component model.

NREL's Annual Technology Baseline (ATB) 2024 separates these correctly:

Total Cost ($/kW) = [BatteryPack ($/kWh) × Duration (h)] + BOS ($/kW)

Where BOS (balance of system) covers the inverter, controls, and
installation — costs that scale with power capacity, not energy capacity.

For LFP at moderate scenario (2024 values from Cole, Ramasamy & Turan 2025, NREL/TP-6A40-93281):

# The correct model
pack_cost_per_kwh = 241   # $/kWh — energy component
bos_cost_per_kw   = 339   # $/kW  — power component
duration_hours    = 4.0   # hours

def calculate_bess_capex(power_kw: float, duration_h: float) -> dict:
    energy_kwh    = power_kw * duration_h
    energy_cost   = energy_kwh * pack_cost_per_kwh
    power_cost    = power_kw * bos_cost_per_kw
    total_cost    = energy_cost + power_cost
    cost_per_kwh  = total_cost / energy_kwh  # effective $/kWh

    return {
        "energy_kwh":     energy_kwh,
        "energy_cost_usd": energy_cost,
        "power_cost_usd":  power_cost,
        "total_cost_usd":  total_cost,
        "effective_per_kwh": cost_per_kwh,  # varies with duration!
    }

# 500 kW, 4-hour system
result = calculate_bess_capex(500, 4.0)
# total: $920,500  |  effective: $460/kWh

# Same energy, 2-hour system (1000 kW inverter)
result2 = calculate_bess_capex(1000, 2.0)  
# total: $1,143,000  |  effective: $571/kWh

Same total energy stored. 24% cost difference. That's the error you make with flat $/kWh.

Why duration changes the effective cost per kWh

Here's the relationship visualized:

The longer the discharge duration, the lower the effective $/kWh — because you're spreading the fixed power costs (BOS) over more stored energy. This is why 6-8 hour systems often pencil out better for large industrial loads even though the upfront cost is higher.

Round-trip efficiency convention — another source of errors

One more thing that trips people up: manufacturer specs often quote DC-DC round-trip efficiency (0.92–0.95 for LFP). But commercial buildings are billed on AC metered output, so you need the AC-AC figure.

# DC-DC (what manufacturers advertise)
rte_dc_dc = 0.93

# AC-AC (what actually matters for your electricity bill)
rte_ac_ac = 0.85  # LFP — from NREL ATB 2024b and PNNL-33283

# The difference matters for savings calculations
annual_cycles = 365
usable_kwh    = 400

dc_dc_throughput = annual_cycles * usable_kwh * rte_dc_dc  # 136,380 kWh
ac_ac_throughput = annual_cycles * usable_kwh * rte_ac_ac  # 124,100 kWh

# At $0.18/kWh, that's a $2,207/year difference in projected savings

What I built

I wrapped all of this into a FastAPI service that also handles:

Load Duration Curve sizing (Neubauer & Simpson 2015, NREL/TP-5400-63162)
LCOS calculation per PNNL-33283
IRA 2022 §48E ITC stacking (base 30% + up to 20% in adders)
Demand charge savings with state-level rates from OpenEI

Every output includes the source equation and the paper it comes from. The API is live on RapidAPI if you want to try it.

GitHub repo |
RapidAPI

Sources: Cole, Ramasamy & Turan (2025) NREL/TP-6A40-93281 ·
Mongird et al. (2022) PNNL-33283 · Neubauer & Simpson (2015)
NREL/TP-5400-63162 · IRA 2022 §48E