Operating challenge
The plant used fixed or manually selected condenser-side frequencies, leaving pump and tower operation away from the lowest-energy point under changing weather and load.
Deployment registry
Real HVAC proofLive deployments
A small set of verified deployments shows how supervisory AI adapts to existing cooling plants, operating constraints, and contractual baselines without turning this page into a broad market taxonomy.
- Plant efficiency improvement
- 12.11%
- AI operating period
- 37 days
- Average plant efficiency
- 0.74 -> 0.65 kW/ton
Optimization approach
AI control adjusted condenser-water pump frequency and cooling-tower fan frequency inside defined bounds, with group-control and AI modes kept switchable.
Verified outcome
The measurement dataset compares the AI period against baseline operation and shows a 12.11% average plant efficiency improvement.
Verified fleet performance logs
Government office cooling plant
Government Office Cooling Plant
A compact office plant used AI supervisory control to optimize pump frequencies and chiller leaving-water temperature during alternating operation.
Read the full case- Energy savings result
- 16.81%
- Average daily savings
- 239.77 kWh
- Total measured savings
- 2,397.71 kWh
Hospital cooling plant
Hospital Cooling Plant
A mixed chiller plant used API-connected AI control to reduce comparable daily plant energy while keeping weather conditions close.
Read the full case- Energy reduction
- 32.18%
- Baseline daily energy
- 3,470.24 kWh
- AI daily energy
- 2,353.38 kWh
Solar manufacturing cooling plant
Solar Manufacturing Cooling Plant
A large industrial cooling plant with heat recovery was evaluated through comparable days matched by weather, production, and load.
Read the full case- Average savings
- 21.29%
- 20-day comparable result
- 24.51%
- Cooling towers
- 17
Regional energy center
Regional Energy Center
A shared chilled-water and ice-storage plant used supervisory optimization to balance load swings, tariff windows, and equipment staging.
- Verified savings
- 10.12%
- Monthly power-cost reduction
- CNY 150K
- Asset inventory
- 7 chillers / 12 towers
High-comfort venue hotel
High-Comfort Venue Hotel
A comfort-critical hotel plant used hard operating guardrails and local deployment to reduce tuning risk.
- Deployment tuning
- 28 days
- Recorded plant efficiency
- 0.55 kW/ton
- Safety boundary
- Hard guardrails
Data center cold plant
Data Center Cold Plant
A three-chiller data center plant used coordinated AI control across chillers, pumps, and towers.
- Verified savings
- 19.42%
- Chiller capacity
- 3 x 3340kW
- Controlled scope
- Chiller / pump / tower
Mode-split data center
Mode-Split Data Center
A data center plant separated free-cooling and mechanical-cooling modes so each operating state could be measured cleanly.
- Free-cooling savings
- 17.23%
- Chiller-mode savings
- 7.54%
- Annual PUE
- around 1.2
Phased data center campus
Phased Data Center Campus
A staged data center deployment used transfer reinforcement learning and advisory value push mode to reduce adoption risk.
- Verified savings
- 11.2%
- Learning strategy
- Transfer RL
- Interface mode
- Advisory push
Review your cooling plant against these patterns.
Compare your facility performance against this verified deployment set. Our engineering team can provide a tailored feasibility review.