RMI BWFP Dispatch Study Results

Degradation-aware economic dispatch comparison for a simulated second-life battery at the UC Davis RMI Brewery, Winery, and Food Pilot Facility. 52,093 hourly data points, 2020–2026.

Site Focus

Primary: RMI Brewery, Winery, and Food Pilot Facility. Avg 55 kW demand, 120 kW peak. Behind-the-meter PV.

Secondary: Jackson Sustainable Winery Building. Load-only (PV telemetry unusable). Serves as a validation site for load profiles.

Tariff: SMUD CITS-1 proxy (21–299 kW). Published schedule — actual UC Davis contract is not publicly disclosed.

Strategy Comparison

Annual results against a $426,041 baseline bill (no battery).

Strategy	Annual Savings	Peak Reduction	Battery Life	EFC/yr	Best NPV	Best Size
Peak Shaving	$6,690 (1.6%)	13.8 kW (11.5%)	18 years	4.7	$85,871	50 kWh / 50 kW
TOU Dispatch	$5,724 (1.3%)	0 kW	2 years	614	$108,658	50 kWh / 25 kW
Forecast-Assisted	$3,490 (0.8%)	0 kW	14 years	20.8	$60,455	50 kWh / 25 kW

Key Findings

Demand charges drive peak-shaving value. $1,599 in demand charge savings vs $-24 in energy charges. The value is from reducing monthly peak, not shifting energy.

TOU arbitrage destroys second-life batteries. 614 EFC/year means 1.7 full cycles per day. The $109k NPV assumes replacement every 2 years — operationally untenable for most institutions.

Smaller batteries dominate NPV. 50 kWh outperforms 200–400 kWh across all strategies. The savings plateau while capex scales linearly.

Peak shaving is institutionally durable. One tunable parameter, 18-year battery life, deterministic demand-charge value, compatible with utility incentive programs.

Battery Degradation & Sizing

Linear additive model: cycle fade + calendar aging (2%/yr). Replacement at 60% remaining capacity.

Peak Shaving: Replacement at year 18. Total 94 EFC over project life. Battery survives nearly the full 20-year horizon.

TOU Dispatch: Replacement at year 2. 12,279 EFC total. Requires multiple replacements over project life.

Sensitivity: NPV stays positive from 3% to 12% discount rate for peak shaving at 200 kWh. Capex breakeven at ~$300/kWh.

Current Limitations

Battery is simulated. No battery telemetry exists in PI for either winery site. All SOC/charge/discharge profiles are modeled.

PV is reconstructed. Direct solar inverter tags are identified but not yet ingested. PV generation is inferred from bidirectional net meter deltas.

Tariff is a proxy. SMUD CITS-1 published rates, not the actual UC Davis custom contract.

JSWB is load-only. PV inverter telemetry is corrupted and the signal is dead since Sep 2025.

Next Phase: Forecasting & Operator Guidance

Load & PV Forecasting Short-horizon models to replace the persistence-forecast scaffold.

Forecast-Driven Dispatch Look-ahead optimization using real forecast accuracy.

Operator Recommendations Cost-outcome-linked advice with confidence bounds.

Backtesting Oracle vs forecast vs baseline comparison to quantify the value of forecast accuracy.