Turn chilled-water data into stronger compliance evidence, earlier warnings and more informed maintenance decisions.

Professional illustration of a Singapore industrial facility chilled-water plant connected to permanent meters, a BAS dashboard and an AI anomaly alert workflow.

For Singapore facility managers and industrial building owners, chilled-water performance is becoming more than an operational concern. It is also a compliance and evidence-management priority.

NEA’s Minimum Energy Efficiency Standards, or MEES, include requirements for covered water-cooled chilled-water systems. The requirements are relevant to certain industrial facilities, including systems with a capacity of at least 300 refrigeration tonnes (RT), subject to the applicable scope and conditions in the current NEA guidance. Separate MEES requirements also apply to certain new buildings from 1 April 2026.

For factories, warehouses, clean rooms and multi-user industrial buildings, the practical question is not simply whether a chiller is efficient. It is whether the facility can continuously measure the right operating information, identify abnormal performance, preserve reliable records and escalate issues to the people responsible for formal assessment and reporting.

AI can support this workflow. It cannot replace the Energy Manager, Professional Engineer, EEO Assessor or the formal submission and certification responsibilities required under the applicable regulatory process.

What MEES preparation means in practice

NEA’s MEES framework for covered water-cooled chilled-water systems requires suitable measurement and verification arrangements, including permanent measuring instruments connected to a Building Automation System (BAS) or standalone Energy Management System (EMS). The system must support trend logging, data aggregation and the preparation of assessment or monitoring reports.

This creates four practical responsibilities for operators:

  1. Define the system boundary. Confirm which chillers, pumps, cooling towers, heat exchangers and connected loads are included in the assessment.
  2. Measure consistently. Use suitable permanent meters and sensors rather than relying only on occasional manual readings.
  3. Preserve usable data. Ensure that readings are time-stamped, aggregated and retained in a form that can be reviewed.
  4. Act on exceptions. Investigate abnormal performance and involve the appropriate qualified professional when assessment, verification or reporting is required.

The exact instruments, calculations, reporting requirements and applicability should be checked against the latest NEA guidance and the facility’s circumstances.

Build the data foundation before adding AI

AI monitoring is only as dependable as the data underneath it. Before selecting an analytics platform, conduct a data and instrumentation review.

1. Map the chilled-water system

Create a simple equipment and metering map. Identify each chiller, chilled-water pump, condenser-water pump, cooling tower, primary and secondary loop, major air-side system and significant process load. For a multi-user industrial building, record which areas or tenants are served by each system where this information is available.

This map helps prevent a common problem: analysing one chiller while excluding pumps, cooling towers or connected loads that materially affect total system performance.

2. Check the permanent measurement points

Review whether the existing installation can capture the information needed for performance calculations and operational diagnosis. Depending on the system design, this may include chilled-water flow, entering and leaving water temperatures, electrical consumption, operating status, condenser-water temperatures, pump operation and cooling-tower conditions.

Do not assume that a value displayed on a BAS screen is automatically suitable for formal measurement and verification. Confirm sensor location, calibration arrangements, units, sampling intervals, time synchronisation and data quality with the responsible technical parties.

3. Connect BAS or EMS data into a controlled workflow

Data should move from meters and controls into a platform where authorised users can view trends, review exceptions and export records. The objective is not to create more dashboards. It is to create an auditable operating history that the facility team can understand and use.

Useful baseline records include daily and weekly cooling load, chiller energy consumption, operating hours, system availability and relevant weather or production context where appropriate. A change in production schedule, clean-room demand or warehouse operating hours can explain a change in energy use; without that context, an analytics system may generate misleading alerts.

Where AI adds practical value

AI and advanced analytics are most useful when they help people find important changes in a large volume of operating data. Several use cases are relevant to chilled-water systems.

Anomaly detection

An analytics model can learn normal relationships between cooling load, chilled-water temperatures, flow, power consumption and equipment status. It can then flag patterns such as rising power at a similar load, unusual temperature differences, extended operation at low load or a gradual decline in cooling-tower performance.

These are warning signals, not automatic diagnoses. A high-energy event may be caused by a fouled heat-transfer surface, a faulty sensor, a control sequence, a change in process demand or a temporary operating condition. The maintenance or engineering team must investigate the cause.

Trend compression and prioritisation

Facility teams may have weeks or months of BAS data but limited time to review it. AI-assisted tools can summarise recurring exceptions, rank assets by likely impact and highlight changes that deserve human review. This can reduce the dependence on manual spreadsheet inspection.

Maintenance decision support

When a warning is linked to a specific asset and operating pattern, the team can prepare a more focused inspection. For example, a persistent change in approach temperature or pump power may justify checking sensors, strainers, valves, heat-transfer surfaces or control logic, subject to engineering review.

Evidence preparation

Consistent trend logging and automated aggregation can make it easier to assemble operating records for review. However, automated data collection does not by itself confirm compliance. The records still need to be checked for completeness, validity, correct system boundaries and alignment with the applicable NEA requirements.

A practical implementation sequence

  1. Confirm applicability. Review the current NEA MEES guidance, system capacity, facility type and relevant deadlines with the responsible Energy Manager or qualified adviser.
  2. Establish ownership. Assign who maintains meters, who reviews alarms, who approves corrective actions and who coordinates formal reporting.
  3. Audit the data path. Trace readings from the field instrument to the BAS or EMS, database, dashboard and export file. Look for missing values, duplicated tags, incorrect units and time gaps.
  4. Set operating baselines. Compare performance by load, operating schedule and relevant process conditions rather than relying on one fixed benchmark for every situation.
  5. Configure alert rules. Start with a small number of meaningful exceptions, such as abnormal power intensity, sensor disagreement, missing data or persistent temperature deviation.
  6. Test the escalation process. Define response times and evidence requirements. An alert should result in a work order, engineering review, observation or documented reason for no action.
  7. Review monthly and before reporting. Check data completeness, closed actions, recurring faults and any changes to equipment or operating conditions.

What facility teams should avoid

Do not treat AI as a compliance certificate. A model can identify risk or inconsistency, but it does not replace the required professional assessment or formal NEA reporting process.

Do not install sensors without a data plan. Unused measurements create cost and complexity. Each point should have a defined purpose, owner and review method.

Do not ignore data quality. Missing timestamps, flat-lined sensors, implausible readings and unrecorded meter changes can weaken both operational decisions and evidence quality.

Do not optimise one asset in isolation. Chiller efficiency can be affected by pumps, cooling towers, controls, distribution losses and process demand. The review should consider the whole chilled-water system boundary.

Preparing for a more evidence-led operating model

Singapore’s Smart Facilities Management direction places greater emphasis on data-driven operations, productivity and sustainability. For industrial operators, MEES preparation is a practical starting point for this broader digital workflow.

A well-designed monitoring system can help the facility team answer three questions quickly: Is the system operating as expected? If not, what evidence supports the concern? Who needs to review and act on it?

ISS can support organisations exploring engineering, facility management and AI automation requirements for ACMV and chilled-water operations. Contact ISS to discuss how your facility can improve measurement, trend logging, anomaly detection and maintenance workflows while keeping qualified professionals responsible for compliance decisions and reporting.

Reference points