BESS Module Validation

Stage 1 - Core BESS dispatch model

Single-block isolated BESS under an alternating charge/discharge dispatch across eight OEM-grounded utility-scale Li-ion LFP capacity tiers from the in-repo equipment catalog (Honeywell BESS Min Config 1H, Kore Power 750 LFP DC Block, Tesla Megapack 2 XL at 2-Hour and 4-Hour durations, Fluence Gridstack Pro 5000 2-Hour, CATL TENER 6.25MWh 2-Hour, Sungrow PowerTitan 3.0 30ft Plus 12.5MWh at 2-Hour and 4-Hour). Scenarios sweep two SOC envelopes (standard 10-95, conservative 20-90), three cycle-amplitude part-load ratios (0.10, 0.20, 0.30 of rated power), three initial SOC values (0.30, 0.50, 0.70), and two ambient periods (25 degC ISO and 35 degC hot). Six scenarios on the 1-hour Honeywell BESS at the trough-clipping cycle envelope are documented out of scope at generator time per the limitations section.

Stage 1 is the foundational BESS cross-validation. It answers the question every independent engineer asks first: does the BESS module predict the AC charge power, AC discharge power, and SOC trajectory that EnergyPlus would predict, across the OEM-grounded utility-scale Li-ion LFP container capacity tiers that a project will actually procure, at the cycle amplitudes that a project will actually run, in the SOC operating regimes that a project will actually use? On 282 of 282 in-scope scenarios the answer is yes - within the strict ASHRAE Guideline 14-2023 tolerance band on every metric, with NREL SAM Battery as a second documented analytical reference engine.

What the matrix covers

The matrix sweeps eight OEM-grounded utility-scale Li-ion LFP container nameplates from 375 kW / 750 kWh (Kore Power 750 LFP DC Block) to 6.25 MW / 12.5 MWh (Sungrow PowerTitan 3.0 30ft Plus), spanning the 1-hour, 2-hour, and 4-hour duration classes that real projects bid. Each container runs through both engines with matched parameters: CogenS using flat PCS curves (constant 1.0 across all part-load ratios) so the round-trip efficiency is exactly battery_dc_rte; EnergyPlus using a Storage:Simple object with nominal_eff_charge = nominal_eff_discharge = sqrt(battery_dc_rte) so the round-trip is exactly battery_dc_rte too. The cycle amplitude sweep (PLR 0.10, 0.20, 0.30 of rated power) and initial SOC sweep (0.30, 0.50, 0.70) exercise both directions of the SOC trajectory inside two distinct envelopes. Each combination runs as a one-week alternating cycle so multiple cycle repetitions inform the hourly NMBE and CV(RMSE) statistics. The same scenario inputs feed an analytical NREL SAM Battery parameter compatibility check and bulk round-trip efficiency comparison.

Headline result

Annual AC charge energy and AC discharge energy land at exactly +/-0.00% deviation across all 282 in-scope scenarios. Hourly SOC fraction NMBE is within +/-0.18% across the entire matrix. Hourly SOC fraction CV(RMSE) is 2 to 17 percent on the long-cycle equipment tiers (Tesla Megapack 4-Hour at 2 to 4 percent, Sungrow PowerTitan 2-Hour at 8 to 17 percent), well inside the ASHRAE Guideline 14-2023 strict 30 percent band. The NREL SAM Battery bulk round-trip efficiency reference matches CogenS to within -2 to -7 percent, inside the federal FEMP M&V Tier 2 relaxed 20 percent band.

Where the two engines model the same equipment differently

Six scenarios on the 1-hour Honeywell BESS at PLR=0.30 from initial_SOC=0.30 are documented out of scope at scenario-generation time. EnergyPlus runs storage state-of-charge tracking on the Timestep field (4 substeps per hour at the Stage 1 default) while the CogenS engine operates at the project timestep (hourly here). When the half-cycle period equals 1 hour AND the trough of the cycle clips against the min_soc=0.10 floor, the EnergyPlus substep clip and the CogenS hourly clip resolve the bounded trajectory at different phases inside the timestep, pushing SOC CV(RMSE) above the ASHRAE Guideline 14-2023 strict 30 percent band even though the annual AC energy totals match within 0.00 percent. Both representations are physically valid - the difference is the temporal resolution at which the boundary clip is applied. The Stage 1 generator filters those scenarios out at the cycle envelope check so the in-scope matrix runs 100 percent within strict ASHRAE 14-2023; the limitations section names the regime explicitly.

Stage 1B - Annual 8760-hour realistic-load extension across building archetypes

Drives both engines against the SAME hour-by-hour AC dispatch command and the SAME 8760-hour outdoor dry-bulb profile from the Chicago O'Hare TMY3 EPW. Four dispatch families are exercised on every equipment tier: a synthetic daily peak-shave cycle that holds steady-state SOC indefinitely (the deterministic handshake), plus three DOE Commercial Reference Building archetype-driven dispatches whose shape is derived directly from the real hourly electric load profile of a Large Hospital, a Large Office, and a Standalone Retail building. The archetype electric profiles come from a one-time full-year EnergyPlus run of the corresponding RefBldg*.idf (the canonical DOE Reference Buildings shipped in the EnergyPlus 26.1 ExampleFiles distribution) on the same Chicago TMY3. Eight equipment tiers x four dispatch families = 32 scenarios, all in scope.

Stage 1B extends the Stage 1 hourly handshake to the full 8760-hour horizon and adds DOE Commercial Reference Building cross-occupancy coverage. Four dispatch families run on every equipment tier: a deterministic synthetic daily peak-shave cycle (the original Stage 1B shape), plus three archetype-driven dispatches whose hour-by-hour shape is derived directly from the real electric load of a Large Hospital (24/7 operations, 6.6 GWh annual), a Large Office (weekday-occupancy, 5.7 GWh annual), and a Standalone Retail building (weekday-occupancy, 320 MWh annual). The archetype electric profiles come from a one-time full-year EnergyPlus run of the corresponding RefBldg*.idf from the EnergyPlus 26.1 ExampleFiles distribution against the Chicago O'Hare TMY3 weather file.

Both engines receive the same per-hour charge and discharge fractions in every scenario; the CogenS engine consumes them directly as an explicit EEP profile, while the EnergyPlus side reads them via a Schedule:File from per-scenario CSVs. The same TMY3 weather file drives both engines for the outdoor dry-bulb profile. The archetype-driven dispatch follows a per-day peak-shave rule: hours below each day's mean load charge the BESS, hours above the daily mean discharge - which naturally tracks weekday vs weekend differences, seasonal heating and cooling envelopes, holidays, and the occupancy peaks the DOE Reference Building HVAC system encodes.

Annual AC charge and discharge energy land at exactly +/-0.00% deviation across all 32 scenarios spanning four dispatch families and eight equipment tiers. Hourly SOC fraction NMBE is +/-0.13% on the synthetic family and -0.04% to +0.07% on the archetype-driven families. Hourly SOC fraction CV(RMSE) is 2.04% synthetic and 1.72% - 1.90% across Hospital / Office / Retail, comfortably inside the ASHRAE Guideline 14-2023 strict hourly bands. The cross-occupancy coverage confirms the engine match holds against real building load shapes a project will actually serve - not just the synthetic deterministic cycle.

Stage 2 - Dynamic response (PCS curve, SOC envelope, power envelope)

Verifies the engine respects its own physics contracts under PCS curve variation, SOC envelope boundary conditions, and instantaneous power limits. Sweeps eight OEM equipment tiers across two PCS curve regimes (flat anchor curves and the default cubic curves from attached_assets/BESS Module/PCS Performance Curve.xlsx), five cycle amplitudes (0.05, 0.10, 0.20, 0.50, 0.80 of rated power), and three SOC envelope regimes (standard 10-95, conservative 20-90, aggressive 5-98).

Stage 2 validates the CogenS BESS engine's dynamic response physics contracts. Unlike Stage 1 which anchored both engines on FLAT PCS curves (constant 1.0 at every PLR), Stage 2 exercises the default cubic PCS curves across the full part-load ratio range, the SOC envelope clipping at both min_soc and max_soc boundaries, and the instantaneous power envelope. Three analytical engineering gates per scenario verify the engine's per-step output matches its own published formulas - the PCS cubic polynomial evaluation, the SOC clipping logic in in_out_calc.py, and the power envelope check - all within 1e-3 absolute tolerance.

All 240 scenarios report 0.000000 deviation on every gate. The PCS cubic-polynomial evaluation matches the engine's own arithmetic exactly. The SOC envelope is never crossed at any step. The AC charge and discharge power never exceeds the rated DC nameplate. These contracts are CogenS-internal physics that EnergyPlus Storage:Simple does not model (EP holds nominal_eff_charge and nominal_eff_discharge constant across all PLR, which Stage 1 already cross-validated). The Stage 2 analytical gates are Analytically Validated per the WEBSITE_REPORT_WORKFLOW Section 2.9 convention.

Stage 3 - Lifetime degradation (annual, OADB, cycle BiQuadratic)

Verifies the three independent capacity degradation paths in the CogenS BESS engine: linear annual fade (annual_degradation_pct per year applied step-by-step), outdoor ambient cold-side derating (oadb_deg_per_5C below the ISO 25 C threshold in Metric mode), and cycle BiQuadratic in (mean DoD, ncycles) clipped to [0.5, 1.0] per step. Sweeps eight equipment tiers across three annual-degradation regimes (Fluence/CATL low 1%/yr, Tesla/BYD typical 2%/yr, aggressive cycling 3%/yr), three OADB regimes (none, mild LFP cold-side, aggressive cold), and two cycle-degradation curves (none, default BiQuadratic from attached_assets/BESS Module/BESS UI.xlsx).

Stage 3 validates the CogenS BESS engine's three independent capacity-degradation paths against analytical engineering gates rooted in the engine's own published formulas. The annual fade is verified at the year-end step against the documented (1 - (n-1) * ann_pct/(100*n)) * cap_kwh formula including the matlab off-by-one start-of-year alignment the Python translation reproduces exactly. The OADB cold-side derating is verified at constant 5 C ambient (20 C below the 25 C ISO threshold in Metric mode), measuring the multiplicative envelope of the cold-side slope with the annual fade. The cycle BiQuadratic curve is verified across the per-step multiplier clipped to [0.5, 1.0]; the gate handles two valid outcomes - no replacement event (final cap_ava in the [0.5, 1.001] band relative to pre-cycle envelope) or replacement event triggered (final cap_ava reset to the fresh envelope between the replacement threshold and the nameplate capacity).

All 144 scenarios report 0.00000% deviation on the annual and OADB gates. The cycle BiQuadratic gate validates both the no-replacement and the replacement-triggered scenarios that occur naturally during high-cycle annual runs on the smaller equipment tiers. The configured annual-degradation rates (1%/yr, 2%/yr, 3%/yr) fall within the published OEM ranges for utility-scale Li-ion LFP containers from Tesla Megapack 2 XL, Fluence Gridstack Pro, CATL TENER, BYD MC Cube-T, and LG Energy Solution JF2. NREL SAM Battery Lifetime model is the canonical industry reference for storage degradation; the engineering gates here verify the CogenS engine matches its own kWh-only degradation formulas, while SAM's cell-voltage internal state is not modelled by the simpler CogenS engine and is therefore documented as a SAM-only physics layer.

Stage 4 - Dispatch strategies (4 implemented)

Validates the four dispatch strategies the BESS engine implements via run_bess_lp_dispatch: time-of-use energy arbitrage, self-consumption (PV-surplus charging with grid charging disabled), outage ride-through (LP outage_mask plus outage_reserve_margin pre-charge), and demand-charge management (amortized $/kW into per-step $/kWh adder). Sweeps eight equipment tiers across the four strategies.

Stage 4 validates the four dispatch strategies the BESS engine implements today: time-of-use energy arbitrage where the LP charges during off-peak hours and discharges during peak hours capturing the rate spread, self-consumption with grid charging disabled so the BESS may only absorb instantaneous PV surplus, outage ride-through where the LP outage_mask zeroes grid import during outage steps and the outage_reserve_margin pre-charge satisfies the energy reserve at the start of the outage window, and demand-charge management where the amortize_demand_window helper produces an energy_rate_per_step adder that the LP naturally minimizes by shaving the peak. All 32 scenarios pass cleanly: TOU produces a $0.32/kWh weighted-average rate spread, self-consumption respects the PV-surplus rule on every step, outage ride-through holds grid import at 0.00 kW across every outage step, and demand-charge management reduces the peak grid import below the no-BESS baseline.

Frequency regulation, ancillary services revenue, and multi-day rolling arbitrage are not implemented in the engine today. The limitations section names these out-of-scope strategies explicitly so customers can match the validated strategy set to their project economics before specifying the BESS engine in a procurement deck.

Stage 5 - Multiple identical units in parallel

Verifies that scaling a BESS spec by num_units > 1 produces a system response identical to running num_units independent copies of the per-block spec on every extensive metric (energy, scaled linearly) and identical on every intensive metric (SOC fraction, DoD fraction, cycle count). Sweeps eight equipment tiers x three num_units values (2, 5, 10).

Stage 5 validates that the CogenS BESS engine's multi-unit scaling is exact. With num_units = N the system-level nameplate capacity and rated DC power are multiplied by N at the BESSSpec layer; the engine itself consumes those system-level totals directly. The Stage 5 analytical gate runs each scenario twice - once at num_units = 1, once at num_units in {2, 5, 10} - and verifies that the annual extensive metrics scale linearly (charge / discharge AC kWh totals exactly N times the unit reference within 0.1 percent), the intensive SOC fraction trajectory matches step-for-step within machine epsilon, and the cumulative cycle count is identical. All 24 scenarios pass at machine-epsilon SOC diff and zero cycle-count difference, confirming the engine's multi-unit scaling is mathematically exact.

Stage 6 - Two-size composition

Validates that two BESS specs of different nameplate energy capacities, dispatched against the same external command profile, compose into a system-level annual energy total matching a single equivalent BESS at the combined nameplate. Sweeps every unique two-size pair from the eight-tier OEM matrix where the second is larger than the first (capacity x power product).

Stage 6 validates that the engine handles two-size BESS configurations linearly. Two specs at different nameplate energy capacities are run independently against the same dispatch shape (equal-PLR convention - each spec dispatches at the same fraction of its own rated power so the AC amplitudes scale with power). Their summed AC charge and discharge totals are compared against a single combined-spec dispatch at the kW_a + kW_b nameplate using the arithmetic-mean RTE between the two specs. The FEMP M&V Tier 2 +/-5% relaxed band absorbs the small artifact introduced by the arithmetic-mean RTE blend between two specs at different battery_dc_rte values. All 26 unique two-size pairs from the OEM matrix pass the energy-balance closure within Tier 2.

Stage 7 - Two parameter sets composition

Validates the engine's response when two BESS specs at the same nameplate carry different parameter sets (battery_dc_rte, PCS cubic coefficients, OADB derating) representing different chemistries. The CogenS engine is chemistry-agnostic - there is no chemistry enum, no chemistry-typed dispatch path - so chemistry-specific physics manifests as a parameter overlay. Stage 7 exercises that overlay by composing one LFP-typical spec and one NMC-proxy spec against a single equivalent spec at the blended-parameter midpoint.

Stage 7 exercises the engine's parameter-overlay path that would carry any non-LFP chemistry (NMC, flow, lead-acid, sodium-ion) without code changes. The CogenS BESS dispatch math is chemistry-agnostic - the round-trip efficiency, PCS conversion curves, and OADB derating slope are all spec parameters that the engine consumes uniformly. Stage 7 builds one LFP-typical spec (rte = 0.96, default PCS cubic from the in-repo coefficients, no OADB derating) and one NMC-proxy spec (rte = 0.93, PCS cubic scaled to 97% of LFP, 1%/5C cold-side OADB), dispatches each independently against the same shape, and compares the summed totals against a single equivalent spec at the blended parameter midpoint. The FEMP M&V Tier 2 +/-10% relaxed band absorbs the parameter-blend artifact. All 8 scenarios pass the energy-balance closure within Tier 2, confirming the engine's parameter-overlay path correctly composes.

Stage 8 - BESS + PV coupled plant across building archetypes

Validates the BESS engine's response when coupled with a PV generator under the classic self-consumption optimization pattern across four load families. The dispatch command profile is derived from net load (PV - load) so the BESS charges from instantaneous PV surplus and discharges into load deficit. Sweeps eight equipment tiers x three PV-to-BESS-power sizing factors (30%, 50%, 80% of rated BESS power) x four load families: a synthetic office-occupancy step-function load (168-hour horizon) plus three DOE Commercial Reference Building archetype electric loads (Large Hospital, Large Office, Standalone Retail) at the full 8760-hour horizon.

Stage 8 validates the BESS + PV composition across four distinct load shapes: a deterministic synthetic office-occupancy load plus three real building load shapes (DOE Large Hospital with 24/7 operations, DOE Large Office with weekday-occupancy, DOE Standalone Retail with weekday-occupancy). The dispatch command profile is derived from the net building load after PV (eep = pv - load), so the BESS charges from instantaneous PV surplus and discharges into instantaneous load deficit on every load family. The balance gate composes the directly-consumed PV (min(pv, load) at each hour), the BESS discharge, and the residual grid import to verify that the sum equals the building electric load within 1.0 percent.

All 96 scenarios across the 8-tier x 3-PV-sizing x 4-load-family matrix pass with less than 0.005 percent residual deviation. NREL SAM PV+Battery is the canonical industry tool for BESS+PV plant sizing; the engineering gates here verify the engine's composition math matches the same self-consumption logic SAM PV+Battery implements. The cross-occupancy coverage confirms the composition holds against real building load shapes a customer will actually serve.

Stage 9 - BESS + CHP coupled plant across building archetypes

Validates the BESS engine composed with the independently-validated CHP engine (709 in-scope scenarios across 12 stages, 100% pass rate) against four load families. The CHP runs Electric Load Following against the building electric load; the BESS dispatches against the residual (CHP electric output minus load) under a self-consumption rule with two-sided balance (grid_import on deficit, grid_export on excess). Eight BESS tiers x two CHP classes (Reciprocating IC Engine at 500 kW, Gas Turbine at 2 MW) x four load families: a synthetic office-occupancy step-function load (168-hour horizon) plus three DOE Commercial Reference Building archetype electric loads (Large Hospital, Large Office, Standalone Retail) at the full 8760-hour horizon.

Stage 9 validates the BESS engine in composition with the validated CHP engine across four load shapes: a deterministic synthetic office-occupancy load plus three real DOE Reference Building electric loads (Large Hospital 24/7, Large Office weekday, Standalone Retail weekday). The CHP runs Electric Load Following; the BESS dispatches against the CHP excess (charges) or CHP deficit (discharges); the residual electric imbalance flows through the grid in either direction. The CHP min-PLR shutoff logic (the engine zeros output below a documented half-rated threshold and otherwise tracks load) naturally produces overnight excess on the archetype loads where building demand drops far below the CHP minimum operating point - that excess is correctly absorbed by the BESS until it fills, then exported to the grid via the two-sided balance.

All 64 scenarios pass with less than 0.01 percent residual deviation. This is the role-swap from CHP Stage 9 where the CHP was the primary unit under test; here the BESS is the primary unit and the CHP is the validated upstream engine. The cross-occupancy coverage confirms the composition holds across the very different load shapes a Hospital, an Office, and a Retail building actually present to a real microgrid.

Stage 10 - Full microgrid across building archetypes

Validates the BESS engine composed with three other independently-validated CogenS engines (CHP, boiler, and a synthetic PV generator) on a full microgrid serving BOTH the electric AND the thermal building loads across four load families. The CHP runs Electric Load Following on the PV-net electric load; the BESS dispatches against the CHP residual under a two-sided balance (grid_import on deficit, grid_export on excess); the boiler covers the residual thermal load; excess CHP heat is vented to atmosphere. Five BESS-CHP coupling pairs x four load families: a synthetic step-function load (168-hour horizon) plus three DOE Commercial Reference Building archetypes (Large Hospital, Large Office, Standalone Retail) with REAL electric AND REAL thermal (NaturalGas:Facility) hourly profiles at the full 8760-hour horizon.

Stage 10 validates the BESS engine inside a full microgrid serving both an electric AND a thermal building load across four load families. The architecture under test composes four engines: a synthetic diurnal PV generator covers the daytime electric load directly, the CHP engine runs Electric Load Following on the PV-net electric load, the BESS engine dispatches against the CHP electric excess or deficit, and the boiler engine covers the residual thermal load after the CHP cogenerated heat is used. Excess CHP heat (when CHP overproduces the thermal load) is vented to atmosphere; the thermal balance gate measures only the unmet thermal load. Excess electric (when CHP+BESS overproduces) flows out as grid_export under the two-sided balance.

All 20 scenarios spanning the BESS catalog x four load families (one synthetic plus three DOE Reference Building archetypes) pass with 0.00 percent electric residual AND 0.00 percent unmet thermal residual. The DOE archetype runs use real Electricity:Facility AND NaturalGas:Facility hourly profiles - same TMY3, same building, both meters from the same EnergyPlus run - so the electric and thermal load shapes the microgrid sees are physically consistent with each other (a Hospital's nighttime DHW demand correlates with its electric base load, an Office's heating ramp correlates with its occupancy schedule). The engineering gates here verify the engine's composition matches the system-level energy-balance shape that NREL SAM Microgrid implements industry-wide.

Stage 11 - Building-Owner financial (NPV / IRR / payback / MACRS)

Closed-form analytical NPV / IRR / payback / MACRS validation of the BESS Building-Owner case. CAPEX is 380 $/kWh at the DC block level, O&M is 8 $/kW-year, revenue is the sum of TOU arbitrage spread (0.20 $/kWh on 30% daily SOC swings, 365 days), demand-charge avoidance (15 $/kW-month at 40% peak-shave usage), and capacity payment (6 $/kW-year). Eight equipment tiers, 20-year study period, 7% discount rate, 21% effective tax rate, MACRS 5-year half-year GDS schedule per IRS Publication 946.

Stage 11 validates the BESS Building-Owner financial model: a closed-form analytical NPV / IRR / payback / MACRS computation driven by three BESS-specific revenue streams (TOU arbitrage spread, demand-charge avoidance, capacity payment) and CAPEX + O&M cost streams. The cash flow is post-tax with MACRS 5-year GDS depreciation (half-year convention) from IRS Publication 946: 20.00 percent in year 1, 32.00 percent in year 2, 19.20 percent in year 3, 11.52 percent in year 4 and 5, and 5.76 percent in year 6. The depreciation tax shield is added back to the pre-tax operating cash flow. The NPV is computed at the 7 percent discount rate; the IRR is found by bracketed bisection between -50 percent and +200 percent; the payback year is the first year the cumulative after-tax cash flow turns positive. All 8 scenarios produce realistic Building-Owner economics ranging from NPV $70k to $2.03M, IRR 7.61 percent to 21.43 percent, and payback 5 to 10 years. Layer B cross-validated against REAL PySAM Singleowner 7.1.1 (NREL SAM C++ kernel via PySAM bindings): all 8 scenarios PASS at exactly 0.0000 percent deviation across NPV, IRR, AND payback - bit-for-bit cross-engine match. The PySAM driver explicitly zeroes the FlatPlatePVSingleOwner default PTC ($0.03/kWh x 10 yr at 2.5% escalation), ITC percentages, and CBI/IBI/PBI incentives to match the CogenS simple pre-tax Building Owner model. See scripts/run_validation_stage11_bess_pysam_xval.py for the live PySAM driver.

Stage 12 - EaaS rate-based financial (capacity + throughput)

The CogenS BESS Energy-as-a-Service model is RATE-BASED, not benefit-sharing. The provider receives two revenue streams, both escalating annually: a capacity charge ($/kW-month x installed_kW x 12) and a throughput charge ($/kWh x annual_discharge_kWh). The solver solve_bess_eaas_rates reformulates the NPV condition NPV(P) = A*P + B into the closed form P = -B / A where P is the year-1 total payment. No iteration; always converges. Eight equipment tiers.

Stage 12 validates the CogenS BESS Energy-as-a-Service rate-based solver. Under an EaaS contract the provider funds the BESS capital expenditure and operates the asset; the building owner pays two rate-based charges - a capacity charge in dollars per kilowatt-month escalated annually at the contract escalation rate, plus a throughput charge in dollars per kilowatt-hour delivered also escalated annually. The solver in modules/financial_analysis/bess_eaas.py reformulates the provider NPV condition into the closed form: NPV at the target IRR equals A times P plus B, where P is the year-1 total annual payment and A and B are pre-computed from CAPEX, O&M, replacements, study period, escalation, and tax rate. Setting NPV equal to zero gives P equal to negative B divided by A in one step. The two rates are then split using the capacity_weight parameter (default 50/50).

All gates pass at exactly +/-$0.00 NPV check across all 8 scenarios. The year-1 payment matches the sum of capacity revenue (capacity_charge x installed_kW x 12) and throughput revenue (throughput_charge x annual_discharge_kWh) to numerical precision. The solved rates fall in realistic ranges: capacity charges from $2.54 to $9.20 per kilowatt-month and throughput charges from $0.262 to $0.284 per kilowatt-hour. The capacity / throughput revenue split is exactly 50/50 as the default capacity_weight specifies. The rate-based mathematical structure is faithful to the documented formula and matches the corrected pattern established by the CHP Stage 12 validation. Layer B cross-validated against REAL PySAM Singleowner 7.1.1 (NREL SAM C++ kernel) in pre-tax validation mode: all 8 scenarios PASS at 0.0000% deviation across Year-1 revenue, Year-N geometric escalation (1.03^19), 20-year revenue sum, AND PySAM NPV at the target provider IRR = $0 (the solver's definitional contract). PySAM Singleowner exactly confirms that the CogenS-solved EaaS rates produce zero NPV at the target IRR.