01

What the evidence says

BOI’s 2025–2026 data-center work identifies energy efficiency, renewable integration, resilient infrastructure, advanced cooling, BESS and PPA planning as material investment topics.

A UPS, generator, PV inverter and BESS PCS have different transition, fault, autonomy and control duties; one label cannot prove that another duty is covered.

Solar generation and stored energy are valuable only inside an architecture that preserves the facility’s approved protection, transfer and recovery logic.

The operating reference changes by system stateControls, protection and responsibility must follow the transition.
  1. 01GRID PRESENT

    Follow the utility reference

  2. 02GRID LOST

    Establish a stable local reference

  3. 03RESTORATION

    Synchronize, transfer and recover safely

Practical decision guide

Design the power state machine before comparing equipment

The Philippine Board of Investments has recently connected data-center growth with long-term energy planning, renewable-energy procurement, advanced cooling and BESS. That policy and investment context does not approve a particular project or turn every battery proposal into resilient infrastructure. For an owner, the useful starting point is an operating boundary: which bus and which services must remain available, what interruption each load can tolerate, and which party has authority to change the power state. S3 treats those questions as engineering inputs rather than as conclusions supplied by a technology brochure.

A data center does not have one undifferentiated “critical load.” IT equipment, network equipment, cooling, pumps, controls, security, fire and life-safety systems, lighting and building services can have different continuity and restart requirements. A useful load map records steady kW, step and inrush behavior, allowable interruption, maximum restart time, dependency on another service and the consequence of failure. It also identifies what may be shed. Nameplate totals alone can hide a cooling load that arrives after the IT transfer or a pump start that exceeds the short-duration capability of the proposed PCS.

UPS, BESS, generator and solar PV are roles, not interchangeable labels. A UPS may provide tightly controlled ride-through to defined loads, while a BESS may serve a longer energy, demand or islanding duty only if its PCS, transfer arrangement, protection and controls support that duty. A generator adds fuel-based duration but has start time, minimum loading, maintenance and failure modes. Conventional grid-following PV normally disconnects when its voltage and frequency reference disappears. It can contribute during an outage only within an intentionally designed island in which another capable source forms the local grid and controls solar output.

Write the operating sequence as a state machine: grid healthy; grid disturbed; UPS ride-through; BESS-supported bus; generator starting; generator accepted; extended island; utility stable again; synchronized or open-transition restoration; and manual maintenance. For every state, assign the voltage and frequency reference, permitted breaker positions, reserve floor, charging limits, load-shed order, timeout and fallback. Then test failed transitions as well as successful ones. A dry contact that starts a generator does not prove that the generator and inverter can parallel safely, and a fast PCS response does not prove that the complete transfer path is fast enough for the protected load.

Size power and energy separately. The PCS kW rating must cover the supported steady load, planned steps, reactive demand, overload duration and environmental derating. Usable AC kWh must cover the required interval after SOC limits, conversion losses, auxiliaries, degradation and resilience reserve. A monthly utility bill, an annual PUE figure or a nominal DC battery capacity cannot establish either result. Model at least the credible outage and recovery scenarios, including delayed generator start, reduced solar production, a second start attempt and the energy needed to recharge without creating a new demand peak.

Cyber resilience is an operating requirement, not a dashboard feature. Document what the BMS, PCS, EMS, UPS and switchgear do when internet, cloud services, remote access, DNS, time synchronization or a supervisory link is lost. Essential protections should remain local, and loss of communications should lead to an agreed safe mode rather than an undefined stop or uncontrolled command. Separate monitoring from control privileges, use named accounts, record configuration and firmware, restrict remote paths, protect backups and keep a locally understandable recovery procedure. These measures do not prove compliance with every cyber framework; they create evidence for project-specific review.

The commercial model must protect the resilience duty. Peak management, solar self-consumption and energy shifting may create daily value, but dispatch should not consume the reserve promised for an outage or count the same benefit twice. Compare normal-operation value and resilience value in separate ledgers, then define the rule that arbitrates between them. A useful free initial review therefore asks for a one-line diagram, interval data, UPS and generator architecture, critical-load assumptions, outage history and the proposed control narrative. It identifies the highest-value uncertainty; detailed modelling and design remain a separately agreed engineering scope.

BOI — Data Center and Power Southeast Asia Summit 2026

Discuss the engineering scope →
02

What a business should do next

  1. Map IT, cooling and support loads by maximum interruption, restart sequence and consequence—not by monthly kWh alone.
  2. Write normal, utility-loss, generator-start, low-SOC, communications-loss and restoration states before selecting the PCS.
  3. Request a free initial S3 review of the single-line concept, load evidence and proposed operating duty before committing to equipment.