Calculation Methodology

Complete reference for every algorithm, constant, and formula used across the Ener.co platform — the Estimator, the Proposer, and the Dashboard.

Part I — Financial Model (Estimator & Proposer)

E1. ASHRAE Efficiency Standards — kW/ton nameplate schedule by install year
E2. Field kW/ton Degradation — compound degradation + 1.5× cap
E3. M&V Savings Estimation — log-curve model, condition, equipment type
E4. Max Savings Cap — 95% nameplate floor
E5. Regional Defaults & Condition Adjustments — life, EFLH, degradation by region
E6. RUL / EUL Calculation — remaining & extended useful life
E7. Sinking Fund — ASHRAE Uniform Sinking Fund formula
E8. Year-by-Year Energy Savings — escalation, degradation gap, efficiency cap
E9. Year-by-Year R&M Savings — dual degradation curves
E10. Capital Reserve Calculation — windowed sinking funds by scenario
E11. ATCO — Year 1 Annual Total Cost of Ownership
E12. CO₂ / Carbon Savings — EPA grid-average emission factor
E13. TCO — Run to Fail — accelerated degradation, emergency replacement
E14. TCO — Buy New — planned replacement at RUL
E15. TCO — Performance Partnership — shared savings, extended life
E16. TCO — Fee for Service — upfront cost, full savings retained
E17. Sensitivity Analysis — 5-point range ±30%
E18. Constants & Parameters — master reference table
E19. HVAC Replacement Cost Reference — multi-factor model, location index, inflation

Part II — Dashboard Savings (Sensor-Based)

D1. Energy Savings — amperage-based kW, degradation escalation
D2. R&M Savings — EER-based improvement factor
D3. CapEx Savings — sinking fund, RUL vs EUL
D4. Cumulative Savings — real-time accrual counter
D5. ATCO — annual total cost of ownership

Part III — Sensor & Live Reading Calculations

S1. Amperage → kW / kWh — Monnit CT sensor conversion
S2. Enthalpy Delta — psychrometric coil performance
S3. Discharge → Outlet Delta — condenser health proxy
S4. EER — Energy Efficiency Ratio — real-time BTU/W from sensors
S5. EFLH — Runtime Method (A) — amperage ON/OFF classification
S6. EFLH — Energy Method (B) — enthalpy-based cooling output
S7. EFLH Annualization (CDH) — Cooling Degree Hour scaling
S8. Latent vs Sensible Decomposition — dehumidification effectiveness
S9. EER Scatter — Regression — EER vs outdoor temp, R²
S10. Hourly Load Factor — load vs temp, CDH validation
S11. Per-Unit Summary Metrics — avg EER, avg kW, subcool ΔT

Part IV — IPMVP Option B Baseline Readiness

V1. Overview & Scope — why Option B, what applies to coil-coating M&V
V2. OAT Bin Coverage — 5°F bins across cooling range
V3. Min Readings per Bin — statistical power per temperature bin
V4. Total Readings — dataset size for regression confidence
V5. Data Completeness — missing interval detection
V6. Monitoring Window — calendar span requirement
V7. kW-vs-OAT R² — regression quality metric
V8. Sensor Channel Completeness — all configured channels reporting
V9. Overall Readiness Score — composite scoring & status thresholds
V10. Informational Metrics — CV-RMSE, NMBE (non-gating)
V11. Energy Savings Approach Selection — A through F rubric, decomposition, per-project methodology
V12. Outdoor Air Fraction Estimation — mixed-air diagnostic for high-OA units

Part I — Financial Model (Estimator & Proposer)

These algorithms power the Refinery Estimator and Proposer tools. They use ASHRAE standards, equipment age, condition, and region to model savings before sensors are installed. Source: shared/calculations.js and estimator/index.html.

E1. ASHRAE Efficiency Standards

Every piece of HVAC equipment manufactured in the US must meet a minimum efficiency standard set by ASHRAE 90.1 (adopted by DOE). We use these as the nameplate kW/ton starting point for each unit based on its install year.

-- federalMinKwTon(install_year) --
-- Returns ASHRAE minimum kW/ton for the standard in effect --

2023+ → 1.07 kW/ton   ASHRAE 90.1-2022
2016–2022 → 1.09 kW/ton   ASHRAE 90.1-2016/2019
2013–2015 → 1.11 kW/ton   ASHRAE 90.1-2013
2010–2012 → 1.19 kW/ton   ASHRAE 90.1-2010
2005–2009 → 1.26 kW/ton   ASHRAE 90.1-2004
2000–2004 → 1.30 kW/ton   ASHRAE 90.1-1999
pre-2000 → 1.41 kW/ton   ASHRAE 90.1-1989

Lower kW/ton = more efficient. A 2023 unit at 1.07 kW/ton uses 24% less energy per ton than a pre-2000 unit at 1.41 kW/ton. These are nameplate ratings — actual field performance degrades over time (see E2).

E2. Field kW/ton Degradation Model

Equipment efficiency degrades over time due to coil fouling, refrigerant loss, and mechanical wear. We model this as compound annual degradation from the ASHRAE nameplate, varying by equipment condition. A hard cap at 1.5× nameplate prevents unrealistic values.

-- derivedKwTon(install_year, condition) --

nameplate = federalMinKwTon(install_year)
age = max(0, current_year − install_year)

-- Condition-based annual degradation rate --
deg_rate = { poor: 3.5%, avg: 2.0%, good: 1.0% }

-- Compound degradation --
degraded = nameplate × (1 + deg_rate)^age

-- Hard cap: equipment cannot degrade beyond 150% of nameplate --
cap = nameplate × 1.5

field_kwton = min(degraded, cap)

Condition	Rate	Rationale
good	1.0%/yr	Well-maintained, regular coil cleaning
avg	2.0%/yr	Typical commercial maintenance schedule
poor	3.5%/yr	Neglected coils, heavy fouling, deferred maintenance

Why the 1.5× cap? A compressor drawing 50% more power than nameplate would realistically have failed or triggered fault protection before reaching that point. The cap prevents the model from projecting equipment states that wouldn't exist in the real world. Example: a 2010 unit (1.19 kW/ton) in poor condition hits the 1.785 cap at age 12 — without the cap, it would be at 2.06 kW/ton by age 16.

New unit field penalty: When modeling a replacement unit in the Buy New scenario, nameplate efficiency is derated by 10% to reflect real-world installation losses (duct leakage, refrigerant charge, controls integration). Source: PNNL field studies.
fieldKwTon(year) = federalMinKwTon(year) × 1.10

E3. M&V Savings Estimation

Pre-sensor estimate of EER improvement from EnerCoat treatment. Uses a logarithmic curve fit to M&V field data — older equipment has greater optimization potential but with diminishing returns. Adjusted for condition and equipment type.

-- Base M&V estimate (log-curve fit of field data) --
base_pct = 5.58 × ln(age + 1) + 2

-- Condition adjustment --
cond_mult = { poor: 1.10, avg: 1.00, good: 0.90 }
cond_adjusted = base_pct × cond_mult

-- Equipment type adjustment --
evap_only → 7% fixed (evaporator-only systems)
cond_only → cond_adjusted
full (both) → cond_adjusted + 7% (evap bonus added)

-- Final: capped at max achievable savings (see E4) --
savings_pct = min(mv_estimate, max_savings_pct)

Log-curve rationale: Younger equipment has less fouling to remove, so gains flatten quickly. At age 5 the base is ~11.0%; at age 15 it's ~17.4%; at age 25 it's ~20.2%. The 7% evaporator bonus reflects the additional EER improvement from treating the indoor coil — supported by NYBC and NYT monitored data.

E4. Max Savings Cap (95% Nameplate Floor)

Savings cannot bring field efficiency below 95% of the ASHRAE nameplate minimum. This prevents claiming that treatment makes a unit better than new — we can only recapture lost efficiency, not create it.

floor = federalMinKwTon(install_year) × 0.95

-- If field kW/ton is already at or below the floor, no savings --
if kwton ≤ floor → 0%

-- Otherwise, max possible improvement --
max_savings_pct = (1 − floor / kwton) × 100

The 95% factor gives 5% headroom below nameplate. Example: a 2016 unit (nameplate 1.09) running at 1.33 kW/ton → floor = 1.035 → max savings = 22.2%. The M&V estimate is capped at this value.

E5. Regional Defaults & Condition Adjustments

Climate and operating environment determine equipment life, cooling hours, and degradation rate. These defaults can be overridden per-unit.

Region	Base Life (yr)	EFLH (hrs/yr)	Deg Without (%/yr)
northeast	18	2,500	3.0
southeast	16	4,000	3.5
midwest	18	2,500	3.0
southwest	20	4,000	2.5
west	22	2,500	2.0

Condition life offsets — added to/subtracted from base life:

Condition	Offset	Effect on 18-yr Base
poor	−3 years	15 years
avg	0 years	18 years
good	+2 years	20 years

Condition mapping from external data: EXCELLENT/GOOD → good, FAIR → avg, POOR → poor. The mapping is case-insensitive and applied during spreadsheet import in the Proposer.

E6. RUL / EUL Calculation

Remaining Useful Life (RUL) estimates how many years until the equipment needs replacement. Extended Useful Life (EUL) is the projected life with EnerCoat treatment — always RUL + 10 years.

condition_life = regionDefaults[region].life + conditionLifeOffset[condition]

RUL = max(3, install_year + condition_life − current_year)

EUL = RUL + 10

Minimum 3-year clamp: Even end-of-life equipment gets at least 3 years of RUL. This prevents degenerate sinking fund calculations and reflects that equipment rarely fails exactly on schedule. The 10-year EUL extension is supported by 13+ years of monitored data at NYBC and 9+ years at NYT showing near-zero degradation post-treatment.

E7. Sinking Fund (ASHRAE USF)

Calculates the annual deposit needed to accumulate a future replacement cost over N years at a given discount rate. Used for capital reserve comparisons across all TCO scenarios.

-- Uniform Sinking Fund --
A = FV × r / ((1 + r)^N − 1)

-- When r = 0 (no discount) --
A = FV / N

Variable	Meaning	Default
FV	Future replacement cost (inflation-adjusted)	tons × cost_per_ton × (1.03)^N
r	Discount rate	7%
N	Years to accumulate	RUL or EUL

E8. Year-by-Year Energy Savings

Energy costs rise from two forces: equipment degradation (needs more kW per ton) and utility rate escalation (electricity costs more). Savings grow as the gap between treated and untreated widens each year.

-- Base annual energy cost (Year 1) --
base_energy_yr1 = tons × kwton × eflh × kwh_rate

-- For each year y = 1 to horizon: --
esc = (1 + 0.02)^y−1   energy rate escalation (2%/yr)
df = min(deg_cap, (1 + deg_without)^y−1)   degradation factor
deg_cap = (nameplate × 1.5) / kwton   max degradation multiplier

-- Efficiency improvement widens each year --
max_eff = max(0, 1 − (nameplate × 0.95) / (kwton × df))
eff_pct = min(sav_pct + (y−1) × (deg_without − deg_with), max_eff)

energy_without = base_energy_yr1 × df × esc
energy_with = energy_without × (1 − eff_pct)
energy_saving = energy_without − energy_with

Key insight: The savings percentage grows each year because untreated equipment degrades faster (2–3.5%/yr) than treated equipment (0.25%/yr). In Year 1 you might save 15%; by Year 10 the gap has widened and you're saving 30%+. This is the "escalating degradation gap" method.

E9. Year-by-Year R&M Savings

Maintenance costs grow as equipment degrades. Treated equipment's R&M baseline is reduced by the initial savings percentage, then grows at the much slower treated degradation rate.

-- Base annual R&M (ASHRAE RP-1237) --
base_rm = tons × cost_per_ton × rm_factor default 4%

-- For each year y: --
rm_without = base_rm × min(deg_cap, (1 + deg_without)^y−1)
rm_with = (base_rm × (1 − sav_pct)) × (1 + deg_with)^y−1

rm_saving = rm_without − rm_with

R&M savings often exceed energy savings in later years because maintenance costs compound with equipment degradation. The default 4% R&M factor (ASHRAE RP-1237) means a 100-ton unit at $5,500/ton has a $22,000/yr maintenance baseline.

E10. Capital Reserve Calculation — NPV / EUL Linear Amortization

The CapEx benefit of EnerCoat is the present value of deferring a replacement from RUL to EUL. We compute that NPV once, then recognize it as a flat annual rate over the equipment's extended useful life. The result is a defensible, conservative number that respects two realities: (1) the decision genuinely creates value the moment it's made, and (2) that value is only earned over time.

-- Step 1: PV of each replacement event --
pv_without = replace_cost × ((1 + infl) / (1 + r))^RUL
pv_with = replace_cost × ((1 + infl) / (1 + r))^EUL

-- Step 2: NPV of deferral (the lump-sum value created today) --
npv_deferral = pv_without − pv_with

-- Step 3: Recognition window --
Forward proposals: cap at horizon so full NPV lands in the contract
recognition_yrs = min(EUL, project_horizon)
M&V (looking back): always EUL — never claim what time hasn't earned
recognition_yrs = EUL

-- Step 4: Flat annual savings recognized in years 1..recognition_yrs --
annual_rate = npv_deferral / recognition_yrs
cap_saving(y) = annual_rate if y ≤ recognition_yrs, else 0

Why this approach: Single-point sinking-fund deltas understate the deal because they ignore the fact that the decision shifts a real cash event from year RUL to year EUL. Multi-cycle NPV models go the other way and introduce artifacts at cycle boundaries. NPV / EUL amortization captures the full present-value benefit and spreads it linearly — the same number is shown each year, the cumulative line is straight, and total recognized savings exactly equal the deal NPV.

Buy-New (BN) drag: When BN replaces earlier than RUL (e.g., proactive turnover), the same formula yields a small negative NPV, which is amortized over the new unit's life. This shows up as a modest ongoing capital drag on the BN scenario.

Sinking-fund mode (legacy): A toggle is preserved that uses windowed sinking funds for projects/users that prefer the older display. Both methods are documented in the engine.

E11. ATCO (Annual Total Cost of Ownership)

Year 1 snapshot comparing the full annual cost of ownership with and without treatment. Used as the headline metric in proposals.

ATCO_without = base_energy_yr1 + base_rm + sinkingFund(inflated_rul, r, RUL)
ATCO_with = (base_energy_yr1 × (1 − sav_pct)) + (base_rm × (1 − sav_pct)) + sinkingFund(inflated_eul, r, EUL)

ATCO_reduction = ATCO_without − ATCO_with

E12. CO₂ / Carbon Savings

co2_factor = getCO2Factor(region) regional metric tons CO₂ per kWh (EPA eGRID 2022)

kwh_saved_yr1 = tons × kwton × eflh × sav_pct
co2_total = (total_energy_savings / kwh_rate) × co2_factor

CO₂ factor is now region-specific using EPA eGRID 2022 data (see E18 constants table). Falls back to national average (0.00037 MT/kWh) if region is unknown. Total energy savings (in dollars) is converted back to kWh by dividing by the energy rate, then multiplied by the regional CO₂ factor. This captures the full horizon cumulative impact including escalation.

E13. TCO — Run to Fail

Models the worst-case scenario: running equipment past its useful life until emergency failure. No capital reserves — purely reactive. This is the "do nothing and hope" approach.

-- Overshoot: equipment runs 5 years past RUL --
RTF_OVERSHOOT = 5 years
rtf_replace_yr = RUL + 5

-- Post-RUL: degradation accelerates 1.5× --
rtf_deg = deg_without × 1.5

-- Post-RUL: R&M scales from 2× to 3× over 5 years --
rm_mult(overshoot_yr) = 2.0 + (overshoot_yr − 1) × (1 / 4)

-- Emergency replacement: 25% premium + 2× disruption --
emergency_cost = inflated_cost_rtf × 1.25
disruption = emergency_cost × disruption_pct × 2

Three phases:
• Years 1–RUL: Normal degradation, normal R&M
• Years RUL+1 to RUL+5: 1.5× degradation, R&M spikes 2→3× (ASHRAE RP-1237 end-of-life spike)
• After RUL+5: Emergency replacement at 25% premium (DOE FEMP), 2× disruption (unplanned), new unit with 10% field penalty

E14. TCO — Buy New

Planned replacement at end of RUL. Capital reserves built via sinking fund from year 1. Includes disruption at replacement and a new sinking fund for the next replacement cycle. Baseline comparison scenario.

-- Phase 1: old unit (y = 1 to RUL) --
energy += base_energy_yr1 × df × esc
rm += base_rm × df
reserve += sf_bn_annual

-- Replacement event at RUL --
disruption = inflated_cost_rul × disruption_pct

-- Phase 2: new unit (y = RUL+1 to horizon) --
new_kwton = federalMinKwTon(replacement_year) × 1.10 10% field penalty
energy += new_base_energy × (1.01)^ny−1 × esc
reserve += sf_new_annual new sinking fund for next cycle

New replacement units degrade at only 1%/yr (good condition baseline) since they're brand new. The 10% field penalty (PNNL) reflects real-world installation losses vs. lab-rated nameplate efficiency.

E15. TCO — Performance Partnership

Ener.co treats and monitors equipment. Savings are shared — client keeps their split, Ener.co's share covers ongoing service. Equipment life extends to EUL.

-- Savings split (default 75/25) --
client_pct = 75% enerco_pct = 25%

-- For each year y: --
yr_savings = energy_sav + rm_sav + cap_sav (streams toggleable)
enerco_payment = yr_savings × enerco_pct
monthly_payment = enerco_payment / 12

-- Client total cost = reduced energy + reduced R&M + EUL reserves + Enerco payments --
pp_total = energy_with + rm_with + ec_reserves + total_enerco_payments

Capital savings in PP compare the Buy New sinking fund vs. the EUL-based sinking fund at each year. The deferral advantage (spreading replacement cost over more years) is the primary CapEx benefit. Streams (energy, R&M, CapEx) can be individually toggled on/off for the deal structure.

E16. TCO — Fee for Service

One-time upfront project fee. Client keeps 100% of all savings — no ongoing payments. Same performance and life extension benefits as PP.

project_cost = tons × price_per_ton

ffs_total = energy_with + rm_with + ec_reserves + project_cost

-- Key metrics --
payback = project_cost / yr1_total_savings
ROI = (total_savings − project_cost) / project_cost × 100

Pricing	Condenser	Evap	Both
Default $/ton	$750	$300	$1,050

E17. Sensitivity Analysis

Shows how key metrics change if actual EER improvement differs from the M&V estimate. Five scenarios span ±30% around baseline.

-- Scenarios --
Conservative      → baseline × 0.70   (−30%)
Slightly Below → baseline × 0.85   (−15%)
Baseline          → baseline × 1.00
Above Baseline → baseline × 1.15   (+15%)
Strong            → baseline × 1.30   (+30%)

-- Each scenario recalculates independently --
s_sav = sav_pct × multiplier
-- Energy & R&M recalculated at adjusted savings --
-- Capital scales proportionally (capped at 1.0× for conservative) --

E18. Constants & Parameters Reference

Parameter	Value	Source / Rationale
current_year	2026	Updated annually
energy_escalation	2%/yr	EIA long-term reference case for commercial electricity
inflation_rate	3.5%/yr	FRED PPI HVAC Equipment (PCU3334133341); above pre-COVID trend for R-454B transition + labor
discount_rate	7%	DOE uses 9% (7% real + 2% inflation); we use 7% conservative
co2_factor (national)	0.00037 MT/kWh	EPA eGRID 2022 national average (delivered basis)
co2_factor (northeast)	0.00030 MT/kWh	NEWE+NYCW+NYLI+RFCE avg — hydro/nuclear/gas
co2_factor (southeast)	0.00040 MT/kWh	SRSO+FRCC+SRVC avg — gas + some coal
co2_factor (midwest)	0.00050 MT/kWh	RFCW+MROE+MROW avg — coal heavy
co2_factor (southwest)	0.00035 MT/kWh	ERCT+AZNM avg — gas + growing renewables
co2_factor (west)	0.00025 MT/kWh	CAMX+NWPP avg — hydro/solar/wind
rm_factor	4%	ASHRAE RP-1237 baseline (% of replacement value/yr)
field_penalty	10%	PNNL: new unit field derate vs. nameplate
evap_fixed_pct	7%	Fixed evaporator-only savings component
deg_with	0.25%/yr	NYBC 13+ yr, NYT 9+ yr monitored data
deg_without (poor)	3.5%/yr	NREL/DOE Building America (2006)
deg_without (avg)	2.0%/yr	NREL/DOE Building America (2006)
deg_without (good)	1.0%/yr	NREL/DOE Building America (2006)
degradation_cap	1.5× nameplate	Equipment fails before exceeding 50% over nameplate draw
max_savings_floor	95% nameplate	Cannot claim savings below near-nameplate efficiency
eul_extension	+10 years	EUL = RUL + 10; NYBC/NYT long-term data
rul_minimum	3 years	Floor on remaining useful life
rtf_overshoot	5 years	Run to Fail runs 5 years past RUL
rtf_deg_mult	1.5×	Degradation acceleration during overshoot
rtf_rm_range	2.0–3.0×	R&M spike during overshoot (ASHRAE RP-1237)
rtf_emergency_premium	25%	DOE FEMP unplanned replacement cost premium
rtf_disruption_mult	2×	Unplanned replacement causes double disruption
mv_log_a	5.58	Log-curve coefficient: 5.58 × ln(age+1)
mv_log_b	2	Log-curve intercept (base 2% at age 0)
cond_life_poor	−3 yr	Equipment life offset for poor condition
cond_life_good	+2 yr	Equipment life offset for good condition
cond_mv_poor	1.10×	M&V multiplier for poor condition (more fouling)
cond_mv_good	0.90×	M&V multiplier for good condition (less fouling)

E19. HVAC Replacement Cost Reference

Multi-factor model for total installed replacement cost. Calibrated against Google SF Bay Area facilities data and contractor quotes (2024). Includes equipment, labor, crane, piping, controls integration, startup, and commissioning.

cost_per_ton = min((base × location_index × facility_mult) + access_adder, $12,000)
total_cost = cost_per_ton × tonnage

Base Cost by Equipment Type × Tonnage Tier ($/ton)

Equipment Type	Small (≤10T)	Mid (11–25T)	Large (26–50T)	XL (51+T)
split	$5,500	$4,000	—	—
ptac	$4,500	—	—	—
rtu	$3,500	$2,800	$2,200	$2,500
chiller_air	$4,000	$3,000	$2,500	$2,800
chiller_water	—	$3,500	$2,800	$3,200
crac	$8,000	$7,000	$6,000	—
ahu	$2,800	$2,200	$1,800	$2,000

Location Cost Index (national avg = 1.00)

Metro	Index	Metro	Index
nyc_metro	1.42	dallas	0.92
sf_bay	1.38	houston	0.90
boston	1.28	atlanta	0.92
la_metro	1.22	phoenix	0.88
seattle	1.18	rural_northeast	0.95
dc_metro	1.15	rural_southeast	0.82
chicago	1.12	rural_midwest	0.85
philadelphia	1.08	rural_west	0.90
miami	1.05

Regional fallbacks: northeast=1.12, southeast=0.92, midwest=1.05, southwest=0.90, west=1.15, tropical=1.10

Facility Type Multiplier

Facility	Mult	Facility	Mult
office	1.00	healthcare	1.40
retail	1.00	hospital	1.50
warehouse	0.90	data_center	1.70
education	1.00	pharma / lab	1.50
hospitality	1.15	food_processing	1.30
multifamily	1.00	clean_room	1.60
industrial	1.10	mission_critical	1.65

Access Complexity Adder ($/ton)

Access Level	Adder	Notes
standard	$0	Ground level, easy access
rooftop	+$250	Crane required
mechanical_room	+$350	Interior rigging
high_rise	+$500	>10 floors
confined	+$450	Tight spaces

Equipment Cost Inflation

Period	CAGR	Source
Pre-COVID 2015–2019	3.1%/yr	FRED PPI PCU3334133341
Post-COVID 2020–2025	8.0%/yr	FRED PPI PCU3334133341
Recent 2024–2025	2.2–3.3%/yr	ACHR News PPI Report
Full period 2015–2025	5.3%/yr	FRED PPI PCU3334133341
Forward rate (model)	3.5%/yr	Above pre-COVID trend; R-454B transition, labor, materials

Validation — Google SF Bay Area Facilities Data

Scenario	Calculation	Result	Google Range
1T split, SF, office	$5,500 × 1.38 × 1.0 + $0	$7,590/ton	$3K–$7.5K ✓
100T chiller_air, SF, DC, roof	$2,800 × 1.38 × 1.70 + $250	$6,818/ton ($682K)	$250K–$600K+ ✓
100T chiller_air, SF, office	$2,800 × 1.38 × 1.0 + $0	$3,864/ton ($386K)	✓
30T CRAC, NYC, data center	$7,000 × 1.42 × 1.70 + $0	$12,000/ton (capped)	✓

$12,000/ton hard cap prevents unrealistic stacking of multipliers (e.g. NYC × data center × CRAC). Even the most extreme real-world scenarios rarely exceed $12K/ton fully installed. All values can be overridden per-unit in M&V project.json or per-proposal in the estimator custom $/ton field.

📊 Interactive Cost Model Charts (for sales walkthrough)

Part II — Dashboard Savings (Sensor-Based)

These formulas are used by the live dashboard once sensors are installed. They use real amperage and EER data instead of model-based estimates. Source: dashboard/calculations.py.

D1. Energy Savings

Measures the reduction in electrical consumption from improved EER after coating. Uses the escalating degradation gap method — untreated equipment degrades faster than treated, widening the savings gap each year.

-- Per-unit power draw (kW) — 3-phase --
multiplier = voltage × √3 × power_factor / 1,000
kW_untreated = baseline_amperage × multiplier

-- Per-unit power draw (kW) — 1-phase --
multiplier = voltage × power_factor / 1,000

-- Treated draw assumes EER returns toward SEER-10 baseline --
kW_treated = kW_untreated × 10 / baseline_EER
Only applies when baseline_EER > 10 (SEER-10 federal minimum)

-- Annual energy cost (Year 0 baseline) --
annual_energy_untreated = kW_untreated × EFLH × electricity_rate
annual_energy_treated = kW_treated × EFLH × electricity_rate

-- Year N savings with degradation escalation --
energy_savings(N) = annual_energy_untreated × (1 + deg_without)^N
− annual_energy_treated × (1 + deg_with)^N

Variable	Source	Default
baseline_amperage	Unit baseline (sensor or manual)	—
voltage	Unit equipment spec	208V
phase	Unit equipment spec	3-phase
power_factor	Unit equipment spec	0.85
baseline_EER	Unit baseline (sensor or manual)	—
electricity_rate	Site setting	$0.20/kWh
EFLH	Unit override > site default	2,000 hrs
deg_without	Unit: coil degradation without coating	2%/yr
deg_with	Unit: coil degradation with coating	1%/yr

D2. R&M (Repair & Maintenance) Savings

Baseline annual R&M cost is a percentage of total replacement cost. Coating reduces R&M proportionally to the EER efficiency improvement. Both untreated and treated R&M costs escalate annually, but at different rates.

-- Replacement cost --
replacement_cost = tonnage × replacement_cost_per_ton

-- Baseline annual R&M --
baseline_rm = rm_pct × replacement_cost

-- EER improvement factor --
eer_improvement = (baseline_EER − 10) / baseline_EER

-- Treated R&M (reduced by EER improvement) --
treated_rm = baseline_rm × (1 − eer_improvement)

-- Year N savings with escalation --
rm_savings(N) = baseline_rm × (1 + rm_esc_untreated)^N
− treated_rm × (1 + rm_esc_treated)^N

Variable	Source	Default
tonnage	Unit equipment spec	—
replacement_cost_per_ton	Unit setting	$3,000 (>15T) / $2,200 (≤15T)
rm_pct	Site: baseline R&M %	4%
baseline_EER	Unit baseline	—
rm_esc_untreated	Site: R&M escalation (untreated)	2%/yr
rm_esc_treated	Site: R&M escalation (treated)	1%/yr

D3. Capital Expenditure (CapEx) Savings — NPV / EUL Amortization

M&V CapEx savings use NPV / EUL Linear Amortization. For each unit we compute the present value of deferring its replacement from RUL to EUL, then recognize that NPV linearly over the unit's extended useful life. This is purposely conservative for backward-looking M&V — we never claim savings that elapsed time hasn't yet earned.

-- Per unit --
RUL = max(3, install_year + useful_life − current_year)
EUL = RUL + extension_yrs
replace_cost = tonnage × cost_per_ton

-- Present value of replacement events --
pv_without = replace_cost × ((1 + infl) / (1 + r))^RUL
pv_with = replace_cost × ((1 + infl) / (1 + r))^EUL
npv_unit = pv_without − pv_with

-- Recognition: M&V always uses EUL (cap_at_horizon = false) --
annual_rate_unit = npv_unit / EUL years 1..EUL, then 0

-- Fleet aggregation --
capex_yearly(y) = Σ_units annual_rate_unit if y ≤ EUL_unit
npv_deferral_total = Σ_units npv_unit

Variable	Source	Default
replace_cost	tonnage × replacement_cost_per_ton	—
r (discount_rate)	project.json	7%
infl (rm_pct as proxy)	project.json	3%
useful_life_yrs	project.json	18 yrs
extension_yrs	project.json	10 yrs
RUL	install_year + useful_life − current_year (floor 3)	—
EUL	RUL + extension_yrs	—

Why amortize over EUL (not horizon)? M&V reports actuals. Recognizing the entire NPV in year 1 would overstate realized savings; spreading it over EUL matches the period during which the deferred replacement remains in service.

Forward proposals (Estimator/Proposer) use the same formula but cap recognition at min(EUL, project_horizon) so the full deal NPV is shown within the contract term.

Sinking fund mode remains available as a toggle for backward-compatible display.

D4. Cumulative Savings (Time-Based)

The dashboard counters show cumulative savings accruing in real time since each site's coating date.

-- Elapsed time --
elapsed = now − coating_date
elapsed_yrs = elapsed / seconds_per_year

-- Cumulative savings = sum of each full year + partial current year --
cumulative(T) = Σ_{y=0..floor(T)-1} annual_savings(y) + annual_savings(floor(T)) × frac(T)

-- Current accrual rate (for counter animation) --
rate_per_sec = annual_savings(current_year) / seconds_per_year

The SSE stream pushes updated cumulative values every 1 second. The browser uses requestAnimationFrame to smoothly interpolate between ticks.

D5. ATCO (Annual Total Cost of Ownership)

ATCO_without = annual_energy_untreated + baseline_rm + reserve_without
ATCO_with = annual_energy_treated + treated_rm + reserve_with

total_savings = ATCO_without − ATCO_with
savings_pct = total_savings / ATCO_without × 100

Part III — Sensor & Live Reading Calculations

S1. Amperage → kW / kWh Conversion

Monnit current sensors report raw amperage. We convert to kW using per-unit voltage, phase configuration, and power factor, then to kWh by multiplying by the time interval between readings.

-- Instantaneous power (3-phase) --
kW = amperage × voltage × √3 × power_factor / 1,000

-- Instantaneous power (1-phase) --
kW = amperage × voltage × power_factor / 1,000

-- Energy consumed between two readings --
dt_hrs = (t₂ − t₁) / 3,600
kWh = kW × dt_hrs

-- Annual projection --
annual_kWh = kW × EFLH
annual_cost = annual_kWh × electricity_rate

Variable	Source	Notes
amperage	Monnit current sensor reading	20A or 150A sensor variants
voltage	Unit equipment spec	Default 208V
phase	Unit equipment spec	3-phase uses √3 multiplier
power_factor	Unit equipment spec	Default 0.85
EFLH	Estimated from sensors or unit/site override	Equivalent full-load hours/yr
electricity_rate	Site setting	$/kWh

S2. Enthalpy Delta (Psychrometric)

Measures the heat energy removed by the evaporator coil using moist-air enthalpy from dry-bulb temperature and relative humidity at inlet and outlet.

-- Step 1: Convert dry-bulb °F → °C --
T_c = (T_f − 32) / 1.8

-- Step 2: Saturation vapor pressure (Buck equation, kPa) --
Pws = 0.61121 × exp( (18.678 − T_c / 234.5) × (T_c / (257.14 + T_c)) )

-- Step 3: Actual vapor pressure --
Pw = (RH / 100) × Pws

-- Step 4: Humidity ratio (lb water / lb dry air) --
Pw_psi = Pw × 0.14504 kPa → psi
W = 0.62198 × Pw_psi / (14.696 − Pw_psi)

-- Step 5: Moist air enthalpy (BTU/lb) --
h = 0.240 × T_f + W × (1061 + 0.444 × T_f)

-- Enthalpy delta (coil heat extraction) --
Δh = h_inlet − h_outlet

Inlet and outlet readings are matched by nearest timestamp within a 5-minute window. The chart shows Δh over time — a declining trend may indicate coil fouling.

S3. Discharge → Outlet Delta (Condenser Health)

Temperature spread between compressor discharge and condenser outlet. Not true subcooling (requires pressure), but a practical proxy.

ΔT_condenser = T_discharge − T_condenser_outlet

Low (< 10°F): possible low charge or poor airflow. OK (10–20°F): healthy heat rejection. High (> 20°F): possible liquid line restriction or overcharge.

S4. EER — Energy Efficiency Ratio

Real-time EER from live sensor data. Combines enthalpy delta (cooling output) with amperage (power input).

CFM = tonnage × cfm_per_ton (default 400 CFM/ton)

BTU/hr = Δh × CFM × 4.5 (4.5 = 60 min × 0.075 lb/ft³)

Watts = Amps × Voltage

EER = BTU/hr / Watts

Reference lines at EER 10 (minimum), 12 (baseline), 14 (good). Points with EER < 6 or > 25 flagged as suspect.

S5. EFLH — Runtime Method (Method A)

is_ON = amperage ≥ amp_on_threshold (default 6.0 A)

runtime_hours = Σ interval_hours where is_ON

PLF = avg_amps_when_on / max_amps_in_window

EFLH_runtime = runtime_hours × PLF

S6. EFLH — Energy Method (Method B)

BTU_interval = Δh × CFM × 4.5 × interval_hours
full_load_BTU_hr = tonnage × 12,000

EFLH_energy = Σ BTU_interval / full_load_BTU_hr

Requires evap inlet/outlet temp/RH sensors. Returns null if not mapped — graceful degradation to Method A only.

S7. EFLH Annualization (CDH Scaling)

Scales measured EFLH to full-year estimate using Cooling Degree Hours (CDH, base 65°F) from Open-Meteo weather data.

CDH = Σ max(0, temp_f − 65) per hour from weather data
CDH_scale = CDH_annual / CDH_current_period

EFLH_annual = min(EFLH_runtime × CDH_scale, EFLH_energy × CDH_scale)

Confidence scoring: < 2 weeks → Low • 2–6 weeks → Moderate • > 6 weeks → Good • Full season → High • CDH < 5% annual → Low (winter penalty)

S8. Latent vs Sensible Cooling Decomposition

Q_total = h_inlet − h_outlet
Q_sensible = 0.240 × (T_inlet − T_outlet)
Q_latent = Q_total − Q_sensible
Latent % = Q_latent / Q_total × 100

Typical commercial HVAC: 20–40% latent fraction. Higher outdoor humidity → higher latent fraction.

S9. EER Scatter — Regression & R²

Plots instantaneous EER vs outdoor temperature. Linear regression quantifies efficiency degradation at higher ambient temps.

slope = (n·Σxy − Σx·Σy) / (n·Σx² − (Σx)²)
intercept = (Σy − slope·Σx) / n

SS_res = Σ (y_i − (slope·x_i + intercept))²
SS_tot = Σ (y_i − ȳy)²
R² = 1 − SS_res / SS_tot

Higher R² means CDH-based annualization is more trustworthy. Points filtered: 6 ≤ EER ≤ 25. Fixed x-axis: 20–100°F.

S10. Hourly Load Factor vs Outdoor Temp

avg_amps_hour = mean(amps in hour) when amps ≥ threshold
max_amps_unit = max(amps) across all data
load_factor = avg_amps_hour / max_amps_unit

Per-unit regression with R² validates CDH annualization. Fixed axes: x = 20–100°F, y = 0–1.0.

S11. Per-Unit Summary Table Metrics

avg_EER = mean(EER_i) filtered 6 ≤ EER ≤ 25

avg_kW = avg_amps_on × voltage × phase_mult × power_factor / 1,000

subcool_ΔT = mean(T_discharge − T_condenser_outlet) matched ± 5 min

Metric	Color Coding	Thresholds
Avg EER	Green / Yellow / Red	≥ 14 / ≥ 10 / < 10
Subcool ΔT	Green / Yellow / Red	10–20°F / < 10°F / > 20°F

Part IV — IPMVP Option B Baseline Readiness

These metrics assess whether a unit’s baseline data collection is sufficient for IPMVP Option B (retrofit isolation with metered data) M&V analysis. Computed in dashboard/baseline.py and displayed in the Site Workshop.

V1. Overview & Scope

IPMVP Option B isolates the retrofit measure (coil coating) by metering the individual HVAC unit before and after treatment. The full IPMVP spec has ~15 criteria covering whole-building Option C and multi-system retrofits. For individual unit isolation with a passive coating ECM, 7 metrics are the ones that actually gate baseline readiness.

What doesn’t apply (and why):

Criterion	Why Excluded
CV-RMSE / NMBE	Model-fit outputs, not baseline prerequisites. Shown as informational only.
FSU (Fractional Savings Uncertainty)	Requires post-treatment data. Irrelevant during baseline collection.
12-month rule	Formal spec says 12 months for weather-sensitive measures. Direct metering + OAT range coverage across the cooling season is a defensible short-term Option B. OAT range matters, not calendar time.
Non-routine adjustments	Apply at reporting/analysis time, not during collection.
Meter certifications, t-stats	Apply at reporting/analysis time.

V2. OAT Bin Coverage

The entire M&V analysis is kW-vs-OAT. Gaps in temperature bins = gaps in the savings story. We bin outdoor air temperature into 5°F buckets across the cooling range.

-- 5°F bins from 60°F to 100°F --
bins = [60–65), [65–70), [70–75), [75–80), [80–85), [85–90), [90–95), [95–100)

filled_bins = count(bins with ≥ 1 compressor-on reading)

-- Pass criterion --
Target: filled_bins ≥ 5

8 possible bins. Requiring 5 ensures coverage across most of the cooling range without penalizing sites that never see extreme temperatures. Only compressor-on readings (kW ≥ threshold) are counted.

V3. Min Readings per Bin

Statistical power per temperature bin. Sparse bins produce noisy averages that weaken the regression.

thinnest_bin = min(count per filled bin)

Target: thinnest_bin ≥ 10 readings

Only filled bins are considered (empty bins are an OAT coverage issue, not a per-bin density issue). 10 readings per bin is the minimum for a defensible bin average.

V4. Total Readings

Overall dataset size for regression confidence. More points = tighter confidence intervals on the baseline model.

total_on = count(readings where kW ≥ min_kw AND OAT is not null)

-- kW derivation --
kW = amperage × voltage × phase_mult × PF / 1,000

Target: total_on ≥ 500

Variable	Source	Default
voltage	units.voltage	208
phase_mult	√3 if 3-phase, 1 if single	√3
PF	units.power_factor	0.85
min_kw	units.amp_on_threshold × V × √3 × PF / 1000	~1.8 kW

V5. Data Completeness

Missing intervals create a biased sample. Completeness is estimated from the median reporting interval vs actual reading count.

median_gap = median(consecutive timestamp differences)
expected = total_span / median_gap
completeness = min(actual_count / expected, 1.0)

Target: completeness ≥ 90%

Using median gap (not mean) avoids being skewed by a single long outage. Completeness is capped at 100%.

V6. Monitoring Window

Calendar days from first to last compressor-on reading. Short windows risk capturing only one weather pattern.

mon_days = (last_timestamp − first_timestamp) / 86,400

Target: mon_days ≥ 14

14 days captures at least two weekday/weekend cycles and likely some weather variation. The formal IPMVP 12-month rule is relaxed because direct metering + OAT bin coverage provides equivalent statistical power for a passive ECM.

V7. kW-vs-OAT R²

Does electrical power actually correlate with outdoor air temperature? Low R² indicates an oversized unit, variable internal loads, or staging behavior that masks the OAT relationship.

-- scipy.stats.linregress(OAT, kW) --
slope, intercept, r_value, p, stderr = linregress(oat[], kw[])
R² = r_value²

Target: R² ≥ 0.75

R² ≥ 0.75 means OAT explains at least 75% of kW variance. Values below this still allow baseline collection to continue but flag that the OAT-based savings model may have limited explanatory power for this unit. Minimum 10 data points required for regression.

V8. Sensor Channel Completeness

All configured sensor channels (source_mappings) must have at least one reading. kW + OAT are the minimum for Option B; supply/return temp+RH enable EER calculation.

total_channels = count(source_mappings for unit)
reporting = count(channels with ≥ 1 raw_reading)

Target: reporting = total_channels (100%)

A channel that was mapped but never reported data indicates a configuration error, dead sensor, or connectivity problem that must be resolved before baseline can be considered complete.

V9. Overall Readiness Score

Composite score combining all 7 metrics. Simple average of individual percentages (each capped at 100%).

metric_pct_i = min(value_i / target_i × 100, 100)
overall_pct = mean(metric_pct_1..7)

Status	Color	Condition
Ready	Green	overall ≥ 90% AND no metric below 60%
Partial	Amber	overall ≥ 50% (but not meeting Ready criteria)
Collecting	Red	overall < 50%

The “no metric below 60%” gate prevents a unit from being marked Ready when it has excellent coverage in most areas but a critical gap in one (e.g. only 2 OAT bins filled, or no sensor channels reporting).

V10. Informational Metrics (Non-Gating)

These are computed and displayed for reference but do not gate baseline readiness. They become gating criteria at the analysis/reporting stage.

-- CV-RMSE (Coefficient of Variation of RMSE) --
predicted_i = slope × OAT_i + intercept
RMSE = √(mean((predicted − actual)²))
CV-RMSE = (RMSE / mean(kW)) × 100%

Metric	Typical Range	When It Matters
CV-RMSE	< 25% good, < 15% excellent	Model fit quality — assessed at analysis time
NMBE	± 5%	Model bias — assessed at analysis time
FSU	Varies	Requires post-treatment data

ASHRAE Guideline 14 thresholds for hourly data: CV-RMSE ≤ 30%, NMBE ≤ 10%. For our sub-hourly metered data with direct measurement (Option B), these are typically well within bounds. They’re shown in the Site Workshop IPMVP card for early visibility.

V11. Energy Savings Approach Selection

After computing the efficiency improvement (Methods A–D in the M&V skill), translate to annual kWh and $/yr savings. Always compute multiple approaches and present as a comparison table. Each M&V library entry shows all applicable approaches with the primary highlighted.

Code	Method	Formula	Best When
A	kW-only	kW_reduction% × avg_baseline_kW × EFLH	kW/CT-only data, no coil-face RH sensors
B	EER-based (humidity-normalized)	EER_improvement% × avg_baseline_kW × EFLH	Full sensor suite with humidity normalization. Default when data supports it.
C	Decomposed (kW + sensible)	(kW_share + sensible_share) × EER% × kW × EFLH	When latent component is controversial or audience demands conservatism
E	Regression × TMY	∑(kW_pre(OAT) − kW_post(OAT)) over TMY hours	Gold standard IPMVP. Accounts for actual temp distribution.
F	Vendor’s method, corrected inputs	savings_frac × avg_baseline_kW × EFLH	Apples-to-apples with vendor; or third-party validated (e.g. NYSERDA)

Primary Approach Selection

Data Available	Primary	Rationale
Full sensors + humidity-normalized EER	B	Full thermodynamic picture including latent capacity
Third-party validated (NYSERDA, utility regression)	F	Don’t override independently validated savings
OAT-dependent improvement (oversized unit)	E	Improvement concentrated at high OAT; uniform % overstates
kW-only / CT data	A	Only honest option without psychrometric data
Mixed fleet, no common EER	A per-unit	Can’t apply a single EER% to heterogeneous equipment

EER Improvement Decomposition

Required on every entry with evap inlet/outlet T+RH data. Decompose total EER improvement at matched OAT × humidity conditions into three components:

Component	Mechanism	Weather Sensitivity
kW reduction	Cleaner condenser → lower head pressure → less compressor work	Low — direct mechanical improvement
Sensible capacity	Colder supply air from improved heat transfer	Low — stable across conditions
Latent capacity	Lower evap temp → more moisture condensation	High — varies with ambient humidity

-- Decomposition at matched (OAT, h_return) bins --
sensible_tons = 1.08 × CFM × (T_return − T_supply) / 12,000
latent_tons = total_tons − sensible_tons
kW_component = tons_pre × 12 / kW_post − tons_pre × 12 / kW_pre
latent_component = total_ΔEER − kW_component − sensible_component

Every M&V library entry includes: (1) a justification paragraph explaining why the primary approach was chosen, (2) the decomposition table + stacked bar chart, and (3) the full approach comparison table. The primary approach is agreed upon with the analyst before finalizing the entry. See SKILL.md §9 for the complete methodology and audience-based recommendations.

V12. Outdoor Air Fraction Estimation

A first-principles diagnostic that estimates what fraction of the air entering the evaporator coil is outdoor air vs. recirculated return air. This is a heuristic sanity check, not an IPMVP method or a correction factor for savings calculations. It exists to explain unusual EER-vs-OAT patterns (e.g. positive slope) so engineers don’t mistakenly flag them as data quality issues.

Mixed Air Temperature Equation

For a unit with a fixed outdoor air fraction, the temperature of the mixed air entering the evap coil is:

T_mixed = OA% × T_outdoor + (1 − OA%) × T_return

If T_return is roughly constant (thermostatically controlled zone), then a linear regression of evap inlet temperature against outdoor air temperature yields OA% as the slope. A slope of 0.72 means ~72% outdoor air, 28% recirculated.

Method

-- Data: evap_inlet_rh sensor temp paired with weather OAT at same hour --
-- Requires ≥30 matched readings, sampled up to 2,000 --

slope, intercept, r, _, _ = linregress(weather_OAT, inlet_temp)
R² = r²
estimated_OA% = clamp(slope × 100, 0, 100) -- only if R² ≥ 0.4

What R² Means Here

R²	Interpretation
≥ 0.7	Consistent OA fraction — fixed dampers or code-mandated minimum OA. Reliable estimate.
0.4 – 0.7	Moderate correlation — could be variable dampers (economizer), occupancy-driven return air swings, or inconsistent fan scheduling. Estimate is approximate.
< 0.4	OAT does not drive inlet temp — mostly recirculated air, or too much noise. No OA% reported.

Why High-OA Units Can Show Positive EER-vs-OAT Slope

For a standard recirculated-air RTU, the evap load is roughly constant (thermostat-controlled zone). Higher OAT means higher condensing temperature → compressor works harder → EER drops.

For a high-OA unit, the mechanism is part-load efficiency: at low OAT the evap load is very small (outdoor air barely needs cooling), so the compressor runs at 20–30% capacity where DX part-load curves are steep and inefficient. As OAT rises, load increases toward the 60–80% capacity sweet spot where most compressors peak in COP. This can produce a positive EER-vs-OAT slope up to the part-load peak.

Above the part-load peak (often around 80–85°F OAT depending on the unit’s specific performance curves), condensing pressure effects dominate and EER should begin to decline — the familiar negative slope returns. The full curve may therefore be inverted-U shaped: rising EER through part-load, peaking, then falling at full load. More data at higher OATs is needed to confirm where the inflection occurs for any given unit.

Variable airflow rate may also contribute: if the unit modulates fan speed with load, the changing CFM affects the enthalpy calculation and can steepen the positive slope in the part-load region.

Common High-OA Applications

Application	Typical OA%	Reference
MRI / imaging suites	70–100%	ASHRAE 170 Table 7-1
Procedure rooms, ORs	100% (code)	ASHRAE 170
Labs / clean rooms	100%	Pressurization requirement
Data center economizers	Variable (0–100%)	ASHRAE TC 9.9
Printing / manufacturing	30–80%	VOC / particulate exhaust makeup

Does OA% affect savings calculations? No. Our bin-matched kW and EER comparisons (pre vs post at same OAT) are valid regardless of OA fraction because the OA% is a constant physical characteristic of the unit — it doesn’t change pre-to-post treatment. TMY3-weighted annual projections are also unaffected. The OA% is purely a pattern-explanation diagnostic that prevents engineers from misinterpreting EER chart shapes. It is not an IPMVP method, not a correction factor, and not an input to any savings formula.

Limitation — economizer units: If a unit has a variable-OA damper (economizer), the OA fraction changes with OAT. Below the switchover (~55°F) it may be 100% OA; above, it drops to minimum code ventilation. The linear regression gives a misleading “average” in this case. A nonlinear inlet-vs-OAT relationship is the telltale sign. This does not affect savings calculations but means the reported OA% should be interpreted as approximate.

Calculation Methodology

Contents

Part I — Financial Model (Estimator & Proposer)

Part II — Dashboard Savings (Sensor-Based)

Part III — Sensor & Live Reading Calculations

Part IV — IPMVP Option B Baseline Readiness