Solar-hybrid for off-grid telecom towers is a settled question across most of West Africa. The configuration is well understood, and most operators quote runtime fuel savings in the [RANGE] band with project economics defensible inside three to five years.
But usually is not always. There is a class of sites — larger than configurator software admits — where honest analysis still favours the older diesel-dominant architecture, or a far smaller hybrid than the standard sizing rule proposes. This piece describes that class clearly, because too many hybrid retrofits look poor after [TIMEFRAME] and the operator does not understand why.
Failure mode one — load profile versus the sizing rule
Standard sizing assumes a roughly symmetric load: similar draw at 11 a.m. and 11 p.m. For a busy urban macro site that is fair. For a rural site whose traffic is daytime-dominated and falls steeply at night, it is not.
A daytime-peaking site has its load conveniently aligned with generation; the battery does little work and the economics are excellent. A flat or night-peaking site — more common than assumed, since base load rarely falls much below [PERCENTAGE] of the daytime average once cooling, lighting and microwave radios are counted — has its load shifted away from generation. The battery does the work. The battery is the most expensive and shortest-lived component in the architecture, and the promised fuel savings quietly evaporate against accelerated replacement.
Failure mode two — ambient temperature
Lithium battery cycle life is published at 25 °C. The mean diurnal temperature inside a sealed shelter in [REGION] is closer to [TEMPERATURE], with in-shelter peaks of [TEMPERATURE] for several hours daily in the hot season. Cycle life at sustained 40 °C is on the order of [PERCENTAGE] of nameplate. If the financial model uses nameplate replacement intervals, the levelised cost is understated by a non-trivial margin. We have seen retrofits with a respectable modelled return that went underwater after the first unplanned battery replacement.
The remedy is not necessarily abandoning the hybrid. It is funding proper shelter ventilation, or an outdoor cabinet, or accepting real cycle life in the model and sizing accordingly. None of these are free, and none appear in the default configuration.
Failure mode three — site access and maintenance
A diesel generator at a remote site is a known quantity: fuel logistics solved, fault modes familiar, most failures recoverable on a first visit with a basic toolkit. A hybrid system at the same site has more components, more software, more state to interrogate, and more failure modes that present identically — a battery management fault, a string imbalance, a controller communication loss and a genuine cell failure can all read as “site down” from the network operations centre. Mean time to repair on a hybrid is typically [MULTIPLE] that of the equivalent diesel-only site, before spares availability for the specific inverter or firmware is considered. Where access is hours over poor road, this matters more than the spreadsheet shows.
Where the diesel still wins, plainly
- Sites with strongly flat or night-peaking load profiles where the battery does most of the work.
- Sites where shelter temperature cannot be brought within battery specification economically.
- Sites with severe access constraints where repair time is a real cost driver.
- Sites whose maintenance organisation cannot yet sustain the hybrid without multi-year vendor dependence.
- Sites whose remaining tower lease is shorter than the battery payback period.
What an honest pre-feasibility contains
- A seven-day logged load profile at the actual site, not an assumed average.
- A measured seven-day in-shelter temperature profile, not a regional climate average.
- Battery cycle life corrected to that temperature and depth of discharge.
- A diesel runtime forecast using corrected autonomy, not nameplate.
- An explicit line item for accelerated battery replacement.
- A diesel price sensitivity across project life — in both directions.
The principle. Hybrids work. The cases where they do not are not random — they cluster and are identifiable in advance. The cost of identifying them is a week of logging and a careful analyst. The cost of not identifying them is a stranded asset.