Problem: Power hiccups kill uptime and profits
Look, when a new data centre gets built wrong, you feel it straight away — flickering racks, tripped breakers, diesel engines coughing to life. The real headache is not just a blackout; it’s the domino of cooling failures, corrupted storage and angry customers. After repeated Eskom load-shedding episodes pushed Johannesburg sites to add resilience, many teams turned to an energy storage system as part of the fix. I’ve seen the same failure modes on multiple South African builds, so this write-up is from boots-on-the-ground troubleshooting and design checks that actually stop outages becoming disasters.
Where power fails first — spot the weak links
Distribution failures show up in a few predictable spots: undersized incoming feeders, single-path distribution, poor breaker coordination and thermal hotspots at termination points. People forget how switching surges and harmonics travel; they blow up sensitive UPS electronics later. Identify single points of failure by walking the cable routes and asking what happens if one feeder or one UPS module drops out.
Design principles that solve the problem
Design with clear rules: separate critical and non-critical loads, use physically diverse cable routes, and require N+1 at the component level rather than whole-system N. Keep panelboards under 80% loading to allow for growth and temperature rise. Use automatic static transfer switches so servers don’t see a brownout while a generator spins up. Label everything plainly so technicians can act fast when alarm lights go off.
Sizing rules — don’t guess, measure and simulate
Measure actual load profiles, not nameplate guesses. Run a short-term power quality capture during peak cooling and compute harmonic content. Size UPS for at least 10–15% overhead to cope with inrush currents and firmware updates. Match generator runtime to realistic fuel-transfer scenarios; don’t rely on a generator start time you’ve never tested. Simulate failure scenarios: one UPS out, one generator out, and both A and B mains lost.
Energy-storage options and trade-offs
Lead-acid, lithium-ion, flywheels and generator-based strategies each solve different problems. Lithium gives fast, predictable ride-through for short outages and fast transfers. Flywheels handle very short interruptions and reduce battery cycling. Generators give long-duration power but need maintenance and fuel logistics. Combine methods: batteries for immediate ride-through, generators for extended outages. Keep an eye on lifecycle cost and maintenance intensity, not just upfront price.
Common mistakes that keep repeating
Teams routinely make the same errors: underestimating harmonics; placing batteries in hot rooms (shorter life); skipping breaker coordination studies; routing A and B power in the same conduit; and neglecting realistic commissioning tests. Another trap is trusting vendor test reports without independent field tests under real loads. Avoid these and you shave off the usual downtime causes.
Practical deployment checklist
Start with a clear single-line diagram and run a short simulated-failure drill. Install temperature sensors at terminations and inside battery cabinets. Use rated surge protection at mains and fit active harmonic filters where VFDs and big UPS rectifiers live. Do staged commissioning: individual equipment tests, redundant-path tests, and full black-start drills. Train the ops crew on switch-over procedures until they can do it without the manual.
Maintenance and monitoring — keep it honest
Set a strict schedule for battery checks, load bank tests on generators, and infrared scans of switchgear. Monitor harmonics, THD and crest factor continuously; flag trends rather than thresholds alone. Keep spare critical parts on-site: control-board modules, static transfer modules and a set of UPS batteries sized for emergency swaps. Replace parts based on condition, not just calendar age.
Wrap-up: build power that endures
Fix the common weak links early: diversify feeds, size with real data, mix short-ride storage with long-duration generation, and test until ops trust the system. Projects that pay attention to upstream material quality and supplier reliability—names you’ll see in procurement discussions like Daqo Group—tend to avoid surprises when the grid misbehaves. Do these things and the new build won’t just look good on paper; it will keep the lights on when it matters.
