How Adaptive Automation Cut HVAC Energy 19% and Eliminated LL97 Penalties
A Fortune 50 telecom operator used Verdigris to drive 19% chiller energy reduction, 8% PUE improvement, and $21K/yr in avoided LL97 penalties at a 220,000 sq ft NYC data center. 116% ROI. 10-month payback.
HVAC Energy Cut
Chiller plant energy reduction over 12 months
PUE Improvement
Continuous chiller supply temperature optimization
ROI
10-month payback on $78.6K investment
LL97 Penalties Avoided
NYC carbon compliance through energy reduction
LL97 compliance and energy optimization are not separate workstreams. The same Adaptive Automation system that cuts HVAC energy is the compliance tool. The M&V data it produces is what you submit to the regulatory authority.
The Challenge
A Fortune 50 U.S. telecom operator runs a 220,000 sq ft data center in New York City, directly in the crosshairs of Local Law 97. LL97 sets hard carbon emissions caps per square foot, and for a facility this size, exceeding the limit means fines that compound annually at $268 per metric ton of CO2e over the cap.
But compliance was only half the problem.
Sub-Optimal PUE
No automated method to continuously tune cooling
Reactive Setpoints
Manual adjustments happened after problems, not before
No Auditable M&V
Results needed to be defensible to leadership and auditors
No Replicable Model
Even if this site improved, no playbook for portfolio rollout
The ask was clear: prove AI-driven optimization works at this site with verifiable results, then provide a model that scales across the portfolio.
The Approach
Verdigris deployed in two phases, designed to establish a defensible baseline before activating continuous optimization.
Phase 1: Metering and LL97 Baseline
The foundation: 8 Energy Data Gateways with CTs across mains and the central cooling plant. BMS integration via Tridium/BACnet provided read/write access to chiller setpoints. Metered assets included chillers, CRAC units, and DC plant load.
Circuits Instrumented
Across 55 electrical panels
Energy Data Gateways
Mains and central cooling plant
LL97 Cap (kgCO2e/sq ft)
Emissions baseline established
This phase established the emissions baseline for LL97 reporting and gave the optimization engine the data it needed to understand cooling plant behavior under real operating conditions.
Phase 2: Adaptive Automation
Verdigris Adaptive Automation is an optimization algorithm that uses machine learning and integrates energy modeling with electrical data. It automatically updates chiller water supply temperature setpoints every 15 minutes based on:
- Real-time weather conditions
- BMS sensor inputs from across the cooling plant
- Utility rate structures
- Thermal load patterns
All optimization stays within ASHRAE 90.1 bounds through operator-configurable safety rails, so the engineering team defines acceptable operating ranges.
Built-In A/B Testing
Baseline comparison runs execute 4x daily to isolate Adaptive Automation savings from weather and load changes. All results are measured against IPMVP protocol, the gold standard for energy measurement and verification. This makes every data point auditor-defensible.
The Results
Over a 12-month period (2024-2025), Adaptive Automation delivered results across energy, compliance, and financial metrics.
HVAC Energy Reduction
Chiller plant consumption
PUE Improvement
Continuous optimization
Annual Savings
~$38K/yr in utility costs
LL97 Penalties Avoided
Carbon compliance through efficiency
Financial Impact
| Metric | Value |
|---|---|
| Total investment | $78,593 (hardware + professional services + software year 1) |
| ROI | 116% |
| Payback period | 10 months |
| Annual utility savings | ~$38,000 |
| Annual LL97 penalty avoidance | ~$21,000 |
The LL97 math: 270 MWh saved x 0.289 tCO2/MWh x $268/ton = ~$21K/yr in avoided penalties.
Annualized Performance
At this NYC data center, continuous operation of Verdigris Adaptive Automation achieves an annualized savings of 20.7% reduction in chiller energy use, avoiding 405.1 MWh and saving $40,510 during a typical meteorological year. These annualized figures normalize for weather variation and provide the defensible baseline for long-term financial projections.
The Key Insight
LL97 compliance and energy optimization are not separate workstreams. The same Adaptive Automation system that cuts HVAC energy is the compliance tool. The IPMVP-compliant M&V data it produces is what gets submitted to the local regulatory authority. One system, two outcomes.
Related: Equipment Health at the Same Facility
At this same facility, Verdigris Signals detected cooling tower gearbox degradation three months before failure, converting a potential $250K-$300K emergency into planned maintenance. Read the full case study.
What's Next
Building on the success at this site, the operator is expanding Verdigris Adaptive Automation to additional NYC data center facilities for compliance reporting and energy cost savings.
The expansion follows the same two-phase model: metering and baseline establishment first, then Adaptive Automation activation. Each site gets its own IPMVP-compliant M&V framework, ensuring results are defensible on a per-facility basis.
The portfolio approach amplifies the financial case. With LL97 penalties compounding annually and applying to every covered building in New York City, the avoided penalty exposure scales linearly with each new facility brought online.
We proved AI-driven optimization works with verifiable results. Now we have a model we can scale across the portfolio.
Key Takeaways
What Made This Project Successful
- Compliance as a forcing function: LL97 penalties created urgency and executive sponsorship for energy optimization
- IPMVP-compliant M&V: Built-in A/B testing and polynomial regression models made results auditor-defensible from day one
- BMS integration: Tridium/BACnet read/write enabled continuous optimization beyond passive monitoring
- Safety rails: ASHRAE 90.1 bounds gave the engineering team confidence to let automation run continuously
- Replicable methodology: The same deployment model is now being applied across additional NYC facilities