Hotel PV + Battery on a 355 m² Roof: What a Diesel Retrofit Can Save

A hotel in West Africa runs two buildings on two 200 kVA diesel gensets. It has a roof of about 355 m². The owner came with a fixed budget of USD 200,000 and one question: can the diesel bill come down without rebuilding the power system?

The answer is yes, but smaller than most people expect from a project described as solar plus storage. The simulated saving is about 21.9% of fuel, and the reason it is not 60% is worth more attention than the number itself. It comes down to the roof, and to what a dispatch controller does with a very small array.

This is a sample project and a preliminary proposal at retrofit concept stage. Every figure below is a simulation output, not a measured result, and nothing here is an offer. The load figures in particular come from design estimates and are waiting on metering.

The brief that existed before the design

The starting point was a two-building hotel with estimated consumption of 1,322.8 kWh per day — about 483 MWh a year — and a design peak demand of 199.4 kW. Supply came entirely from two diesel gensets rated 200 kVA / 160 kW each, one running and one standby.

An earlier study examined a fully off-grid system: enough PV and enough battery to run the hotel without the gensets at all. It did not close within the owner’s budget. That is not surprising at this scale. A hotel needing 1,323 kWh a day is a large daily energy demand, and autonomy for even two cloudy days calls for a battery measured in tonnes and an array several times the roof area available.

So the brief changed. It became: keep the existing gensets, add as much PV as the roof can carry, add enough storage to change how the gensets are loaded, and make the diesel that is still burned go further. That is a different project from an off-grid system, and it is sized differently.

The roof is the first constraint, and it does not negotiate

355 m² of usable roof, at a layout factor that accounts for walkways, parapets, obstructions and row spacing, yields about 0.21 kWp per square metre. That puts the ceiling at 74.46 kWp — 102 modules of 730 Wp each, arranged in six strings of seventeen.

Chart of installable PV capacity and PV coverage ratio against available roof area, with the hotel's 355 square metre roof marked at 74.4 kWp and 19.3 percent of annual consumption

The chart is the clearest way to explain why this project is not larger. Installed capacity rises linearly with available roof area: 200 m² gives about 42 kWp, 355 m² gives 74.4 kWp, and doubling the array to 146.8 kWp would need roughly 700 m².

The shaded band on the chart marks what that would actually require: a canopy structure over part of the site, or the roof of a neighbouring building. Both are real options, and both change the project from a rooftop retrofit into a construction project with a different budget, a different structural scope and a different timeline.

The same chart carries the second constraint. At 74.4 kWp on this load, PV covers 19.3% of annual consumption. Reaching 35% would need 700 m². Reaching 60% would need 1,200 m², which is more than three times the roof the hotel owns.

This is the honest framing of a rooftop project on a constrained site: the array is not sized to an energy target, it is sized to the roof. Everything else in the design follows from accepting that.

What the equipment actually is

The recommended configuration is modest in capacity and concentrated in control.

Single-line diagram of a 74.46 kWp hotel rooftop PV array, DC combiner, 125 kW hybrid storage inverter, 200 kWh LFP battery, EMS controller and two existing diesel gensets feeding a 400 V busbar
Layer Specification Note
PV array 102 × 730 Wp = 74.46 kWp 6 strings × 17 modules, 1,100 V DC
Combining 4 inputs / 1 output Fused, with lightning protection
Hybrid inverter 125 kW, 1,100 V DC / 400 V AC Capacity still to be verified against metered demand
Battery 200 kWh LFP, 0.5C Fire protection and thermal management included
Existing gensets 2 × 200 kVA / 160 kW One running, one standby; model to be confirmed
Distribution 400 V LV busbar Main building and annex; existing panels refurbished
Control EMS controller Dispatch, demand management, communications

Two points about the topology matter more than the equipment list.

The first is that the PV does not export to the public grid, and it does not back-feed the gensets. It parallels with them on the AC side and supplies the load directly. That removes a whole class of interconnection and metering questions that would otherwise consume months, and it is why a retrofit of this kind can proceed on a site with no export agreement at all.

The second is that a 125 kW hybrid inverter is smaller than the 199.4 kW design peak demand. For a meaningful part of the day, the inverter cannot carry the hotel alone. The genset has to be running, or the peak has to come down through demand management. That single number explains most of what follows.

How a typical day actually runs

The dispatch chart below is a simulated July day from the 8,760-hour model, and it shows the whole logic of the design in one picture.

Chart of hotel load served by direct PV, battery discharge and diesel genset over 24 hours, with battery state of charge falling overnight and rising during evening genset charging

Reading it from midnight:

00:00 to about 05:00 — genset off. Load sits at roughly 20 kW, the lowest it gets all day, and the battery carries it. This is the window the whole design is built around. A 20 kW load on a 160 kW genset is a 12% loading, the least efficient way to burn diesel, and it is also the window where the battery’s 200 kWh goes the furthest. The state of charge falls from about 65% to near zero over those five hours.

Around 05:00 — genset starts at a fixed 48 kW. The load begins climbing as the hotel wakes up, and the controller brings the genset on at a fixed output rather than letting it follow the load up and down. Any surplus above what the hotel needs goes into the battery. That is what makes a single fixed operating point possible.

Through the day — genset and PV together. Direct PV climbs to a modest peak and offsets load while the roof is in sun. The genset stays on, moving between roughly 55 kW and 80 kW as the hotel’s own demand rises. Because the array is only 74.46 kWp, the PV never approaches the inverter’s limit, and there is zero curtailment anywhere in the year.

18:00 to 21:00 — genset runs hard to refill the battery. Load peaks near 90 kW, the evening period that decides how much diesel the hotel will burn. The genset goes to about 138 kW output, which is above the hotel’s own demand, and the surplus charges the battery from about 20% back up to 93%. The controller is buying cheap stored energy for the night, using capacity the genset would otherwise waste by running light.

Late evening — genset steps down, battery takes over again. The cycle closes and repeats.

What the chart does not show is the constraint that makes all of it necessary: 200 kWh of battery covers about five hours at 20 kW of night load. Without the genset topping it up in the evening, the night window would shrink and the diesel saving would evaporate.

The genset is not a backup. It is the main source.

This is the point where the proposal has to be read carefully, and where a lot of retrofit designs are described in a way that will not survive commissioning.

In the 8,760-hour simulation, the gensets run about 5,655 hours a year. That is 15.5 hours a day, every day. And the PV penetration is 19.3%, meaning 80.7% of the hotel’s energy still comes from diesel.

Equipment that runs 15.5 hours a day is not a standby source. It is the primary source, with a solar and battery system bolted on to reduce what it burns.

That is not a criticism of the design, and it is not a hidden failure. It is the arithmetic of a 355 m² roof and a 199.4 kW peak. On this site there is no arrangement of 74.46 kWp of PV and 200 kWh of battery that removes the gensets, because the battery alone holds five hours of night load and the array produces a fifth of the year’s energy.

What the design does is change the genset’s job. Before the retrofit it ran at whatever the hotel demanded, much of the time well below its efficient range. After it, the genset runs at a controlled output, charges a battery deliberately, and shuts down entirely in the hours where running it was pure waste.

The proposal is honest about this in its own numbers: supply cost falls from about 0.5572 to 0.4494 USD/kWh, a 19.4% reduction, not a transformation. A hotel operator reading the headline “solar plus storage plus diesel hybrid” might expect more. The roof sets the ceiling, and the roof is 355 m².

Anyone planning a retrofit of this kind should require their supplier to state the simulated genset running hours and the PV penetration in the proposal, in plain figures. If a proposal describes the genset as backup and then models it running 15 hours a day, the document has a contradiction in it, and the place it will surface is the first fuel bill after commissioning.

Where the 21.9% actually comes from

The saving of about 38,133 litres a year is not the PV array’s doing. A 74.46 kWp array covering 19.3% of consumption would, on its own, save roughly that share of fuel — and it would do so at a poor efficiency, because the genset would still be running all night at low load.

The saving comes from three dispatch decisions that only exist because the battery does:

Shutting the genset down at night. The 00:00 to 05:00 window is the single largest contributor. A 20 kW load on a 160 kW genset burns disproportionately more fuel per kWh than the same load served from a battery that was charged at a better operating point.

Locking the genset to a fixed, efficient output. Instead of following the load across its whole range, the controller keeps the genset near the top of its efficiency curve and uses the battery to absorb the difference. This matters most in the evening, where the chart shows the genset deliberately over-producing to charge.

Shaving the daytime peak with direct PV. The array displaces diesel kWh during exactly the hours when the genset is at its most loaded and the hotel’s demand is highest.

Remove the battery and the first two stop existing. Remove the controller and all three stop existing — the hardware would still be there, but it would not be doing anything except passing energy through. An energy management system is what turns the equipment list into the saving.

3.16 years versus 6.88 years

Here is the number that should be in the acceptance criteria rather than only in the financial model.

Case Static payback
With EMS dispatch strategy implemented ≈3.16 years
Without it ≈6.88 years

Same array, same battery, same inverter, same gensets. A factor of 2.2 in payback, decided entirely by whether the dispatch logic is commissioned and then actually runs.

This is a common failure mode in hybrid retrofits. The equipment arrives, the inverters are set to a default operating mode, the battery cycles gently and the site sees a modest improvement that never reaches the modelled case. Nothing is broken. The system simply is not doing the thing that generated the business case.

The practical consequence is that dispatch performance has to be a contractual acceptance item, not a feature described in a brochure. What that means in concrete terms:

1. A recorded 24-hour dispatch profile at handover, showing genset output, PV output, battery charge and discharge, and load, at interval resolution.

2. A genset running-hours target derived from the 8,760-hour simulation, measured over the first full month.

3. A state-of-charge target at dawn, since the night window depends on the battery arriving at sunset with enough charge.

4. A written statement of the dispatch logic — the fixed genset output setpoint, the low-load shutdown threshold, and the conditions for restart.

5. A defined process for changing setpoints, since site conditions change and an untouchable controller becomes a black box.

A site that accepts a hybrid system without those five items is buying hardware and hoping. The retrofit survey checklist covers the physical side of the same conversation.

The shortest payback is not the recommended option

Three storage configurations were compared at the fixed PV ceiling: 100 kWh with a 100 kW inverter, 200 kWh with a 125 kW inverter, and 280 kWh with a 150 kW inverter.

Monthly chart of direct PV energy and diesel genset energy per day for a hotel hybrid system, with PV penetration between about 15.8 and 19.6 percent across the year
Option Static payback Position after 15 years
A — 100 kWh / 100 kW ≈2.77 years Lowest cumulative cash
B — 200 kWh / 125 kW ≈3.16 years Highest cumulative cash
C — 280 kWh / 150 kW ≈3.80 years Close to Option A

Option A returns capital fastest. Option B accumulates about USD 55,000 more over fifteen years. Option C costs the most and ends up in roughly the same place as Option A, having spent more to get there.

The recommendation is B, and the reason is not subtle once the chart is in front of you: payback measures how quickly the money comes back, not how much comes back. A smaller battery returns its cost sooner because it cost less, while delivering a smaller annual saving for the life of the equipment. For an owner whose constraint is cash flow in the first three years, A is a defensible choice. For an owner assessing a twenty-year asset, B is.

What neither option does is change the roof. The PV ceiling is identical in all three, so the entire comparison is about storage size, and storage size is the only variable the owner actually controls at this point.

The year, month by month

Solar projects in temperate climates show a pronounced seasonal swing. This one does not.

Line chart of cumulative cash flow for three hotel PV, battery and diesel options with static payback of 2.77, 3.16 and 3.80 years, the recommended option ending highest after 15 years

PV penetration runs between about 15.8% and 19.6% across the twelve months, with the weakest months being the wetter ones and the strongest falling at the end of the year. The variation is a few percentage points, not the factor of three or four that a northern European or North American site would show between December and June.

For a hotel in the tropics, that is a quiet advantage worth naming. The diesel saving is roughly constant every month, which makes the cash flow predictable and the budget line stable. It also means the battery cycles at a similar depth all year — about 406 equivalent full cycles annually — rather than being under-used for half the year and hammered for the other half.

What the numbers say about supply cost

Metric Before After
Cost of supply ≈0.5572 USD/kWh ≈0.4494 USD/kWh
Annual fuel use Baseline ≈21.9% lower, about 38,133 L saved
Annual net saving — ≈USD 52,043
Investment — ≈USD 164,241 against a USD 200,000 budget

The investment lands about USD 35,759 inside the owner’s ceiling. That headroom is not an accident — the three storage options were deliberately compared within a fixed PV and a fixed budget, so the choice was never between “build this” and “build something much larger”. It was between three ways of spending a similar amount.

The supply cost reduction of 19.4% is the figure that should be quoted to a hotel board rather than the fuel saving alone. Fuel savings in litres are sensitive to the price per litre, and prices in the region move. The cost per kWh delivered is a more stable way to describe what changed, and it is the number that survives a change in diesel price.

All of these figures carry the same caveat as everything else here: they are simulation outputs from a preliminary model, based on an estimated daily consumption of 1,322.8 kWh and an estimated peak demand of 199.4 kW, neither of which has been measured.

What must be measured before anything is ordered

Six items, in the order they can invalidate the design:

1. A 15-minute interval demand record, covering a full week including weekend and any event days. The 199.4 kW peak is an estimate, and the 125 kW inverter capacity depends on it.

2. Genset nameplates and repair history. The model assumes 2 × 200 kVA / 160 kW; the actual units and their condition determine both the dispatch strategy and whether they can accept parallel operation with an inverter.

3. A structural check of the 355 m² roof. Fully elevated mounting on a hotel roof carries wind loading, and the layout factor of 0.90 is a planning assumption until someone has walked the roof.

4. Confirmation of the 125 kW inverter size against measured demand, since the entire evening dispatch depends on the genset and inverter sharing the peak correctly.

5. The low-load shutdown threshold the gensets can tolerate. Some units handle frequent starts and stops well; others lose more in wear than they save in fuel.

6. A metered baseline of monthly fuel consumption, without which the 21.9% saving cannot be verified after commissioning.

A design built on estimated load is a design that will be adjusted on site. The purpose of listing these is not to delay the project but to make clear which figures are provisional, so that nobody is surprised when the first month of metered data moves them.

FAQ

Q: Is 21.9% a good diesel saving for a hotel retrofit?

A: It is a realistic one for this roof. The saving is capped by the array, and the array is capped by 355 m². A hotel with a larger roof, or with land for a ground-mounted array, could reach substantially higher penetration. On this site, a proposal claiming 50% or more from a 74.46 kWp array and a 200 kWh battery should be treated with suspicion unless it explains where the missing energy comes from.

Q: Why keep the gensets at all if there is a battery?

A: Because the battery holds about five hours of night load and the array produces a fifth of annual consumption. There is no arrangement of this equipment that removes a 199.4 kW peak demand from two diesel units. The retrofit changes how the gensets are loaded and how many hours they run, not whether they are needed.

Q: What if the hotel’s load grows?

A: The PV side is fixed by the roof, so additional load is served by the gensets and, to a degree, by the existing battery. Growth beyond roughly the current peak would need more storage or more generation, and any expansion of the array requires roof area that does not exist on this building. This is one of the reasons the roof constraint is stated so prominently early in the proposal.

Q: How do we know the EMS is actually working after handover?

A: By measuring the three things the business case depends on: genset running hours, state of charge at dawn, and the monthly fuel bill. All three are simple to record and all three should be in the acceptance criteria. If genset hours after three months match the pre-retrofit baseline, the dispatch logic is not doing its job regardless of what the controller screen shows.

Q: Does the PV ever back-feed the gensets or the grid?

A: No. The array parallels on the AC side and supplies the load directly, with no export to the public network and no back-feeding into the gensets. That simplifies the interconnection question considerably and is one of the reasons a project of this size can proceed without a grid export agreement.

About this proposal

This is a preliminary proposal preview prepared to show the design reasoning, not a delivered project. Load figures are estimates awaiting metering, the generation and dispatch results come from an 8,760-hour simulation rather than from measurement, and the equipment is described by reference specification. All financial figures are indicative estimates and do not constitute an offer. The genset running hours and PV penetration quoted here are model outputs; the design intent is described in the dispatch logic, and the two should be read together.

If you operate a hotel or resort on diesel and are considering a retrofit, the three things that decide the answer are a full year of interval consumption data, the nameplate and condition of your existing gensets, and a measured roof area with a structural opinion attached. The battery size and the dispatch strategy both follow from those three, and none of them can be inferred from the size of the property.

This proposal was first shared on Facebook and LinkedIn.

YURI POWER Engineering Team · November 10, 2026