Power, water, and everything the building assumes

Nigeria has roughly 13,000 megawatts of installed generating capacity and in practice delivers somewhere between 4,000 and 5,000 to a population above 200 million. The estimated requirement is around 30,000. The national grid collapsed more than a dozen times in 2024, and several hundred times over the preceding two decades.

The consequence is a fact that should stop any business plan in its tracks. The privately owned generator fleet in Nigeria accounts for something close to 14 gigawatts of the electricity actually used, which is roughly three times what the grid itself delivers. Nigerian manufacturers spent over a trillion naira on alternative energy in a single year, a rise of more than forty per cent on the year before.

Read that as an operating condition rather than as a news story. A hospital built in that market is not connecting to a utility with occasional faults. It is building its own primary supply and treating the grid as a useful supplement, and the design, the capital plan, the plant rooms and the staffing all follow from that inversion.

Reliability is not a property you specify, it is a quantity you purchase, and each layer multiplies the failure probability of the one beneath it. Grid alone, grid plus a generator, grid plus two, and uninterruptible supply on the loads that cannot survive the seconds it takes a generator to pick up.

The numbers below use stated assumptions rather than measured ones, which is the point: a real design starts by measuring the grid at the actual parcel, because the difference between forty per cent unavailability and twenty changes the whole plant specification.

Availability arithmetic tells you how many layers to buy. The configuration decides what each layer carries, and in a hospital that decision is made once, in the electrical design, and then lived with for sixty years.

The organising idea is that a hospital does not have one electrical system. It has three, separated deliberately. A no break supply for loads that cannot survive even the seconds a generator takes to pick up. A short break supply for the much larger group that can tolerate a few seconds but not an hour. And a non-essential supply that is designed to go off, in a planned order, so that the decision about what to shed is taken at design stage by people with time to think rather than at three in the morning by whoever is on.

That last point is the one most often missed. Load shedding is going to happen. The only question is whether the sequence was designed or improvised, and improvised shedding in a hospital means somebody makes a clinical priority decision using a distribution board.

The generators only guarantee anything for as long as there is diesel, which makes fuel autonomy the single most consequential number in the configuration and the one that turns an electrical decision into an architectural one.

Carrying a hospital’s essential load burns fuel continuously and at a rate that surprises people who have only specified standby plant for an office. A week of independence is tens of thousands of litres, which is bunded storage, fire separation, tanker access and a planning conversation, all of which occupy position on the site and therefore belong in the massing rather than in a later package.

How many days is the judgement, and it is a supply chain judgement dressed as an engineering one. Three days is a normal standby figure in a market where the grid comes back and fuel arrives on schedule. Neither of those assumptions holds in the market under consideration, so the honest answer is longer, the cost is real, and it should be argued for explicitly rather than discovered when the tankers stop.

It is also the same argument as the stock page, in a different fluid. Lead time variability drives the buffer, and fuel is the buffer whose exhaustion stops everything else simultaneously.

Solar is the first thing anybody proposes for a hospital in West Africa and it deserves a serious answer rather than either enthusiasm or dismissal. Lagos has good irradiance, panel prices have collapsed, and the fuel it displaces is among the most expensive electricity in the world. The economics are genuinely attractive.

The constraint is not sunshine. It is area, and the arithmetic is unforgiving. A hospital carrying a continuous essential load consumes a great deal of energy over a day, and photovoltaic generation needs roughly seven square metres of installed array for each kilowatt peak.

Run those together on a realistic roof, with plant, air handling and access already occupying much of it, and the share of daily energy solar can supply is real but modest. Chasing a large share means covering land, and on a constrained urban parcel that land is in direct competition with the building, the car park, the ambulance turning circle and any expansion the group wants to keep possible.

None of which is an argument against it. It is an argument for sizing it honestly and describing it as what it is: a reduction in what you burn rather than a reduction in what you must be able to carry.

It is also an argument for questioning the assumption underneath the constraint, which is that the array has to occupy land or roof that something else wanted. The next chapter drops that assumption, and the number moves a long way.

The area constraint above assumes the array has to sit on ground or roof that something else wanted. Drop that assumption and the arithmetic changes completely, because a hospital site is full of surface that is already committed to a purpose incompatible with nothing.

Start with the car park, which is usually the largest single area on a hospital site and is currently a heat island with cars on it. Canopies over it produce power, shade the vehicles, shelter the people walking to the door, and collect rainwater that otherwise has to be attenuated somewhere. The parking capacity is unchanged. Then the roof, on frames raised above the air handling plant rather than competing with it. Then covered walkways and service yards, which need a canopy in a rainy season anyway. Then the west and south facades, which are going to face the sun whether or not anything is mounted on them, and which produce less per square metre and cost nothing in land.

And then the piece of land the group is holding for expansion, which will earn nothing for a decade and is the single best argument for relocatable frames rather than permanent foundations. The array pays for itself while the plot waits and moves when the plot is built on.

Add them and the total is the difference between solar as a gesture and solar as a supply. None of it took clinical floor area. None of it took a parking space.

The second half of the idea matters as much as the first. Build this as one large engineered system and it becomes a single capital decision taken at the worst possible moment, which is before the hospital has opened and before anybody knows what the load profile actually looks like.

Build it as repeated modules instead, each a standard assembly of array, inverter, battery and controls, and several things change at once. Capital becomes incremental and follows the ramp, so the plant grows as the occupancy does rather than sitting idle waiting for it. The load profile gets measured on the first module and informs the tenth. Each module is commissioned and proven on its own, so a fault is partial rather than total, which is the same redundancy argument as the second generator in a different form. And procurement gets easier, because a repeated standard item is a negotiation you have once.

It also fits what this group is actually doing. Meridian builds hospitals repeatedly and intends the operating standard to travel between them. A containerised, repeatable energy module is that standard expressed in plant: the same unit, the same commissioning tests, the same spares and the same maintenance regime in Lagos and at whatever site comes after it. That is worth more to a business selling demonstrated performance than a bespoke system that has to be explained from first principles to every buyer.

The word microgrid is often used to mean solar with a battery attached. What it actually means is a network that can run connected to the grid or disconnected from it, under its own control, and that distinction is the reason it belongs on this page rather than in a sustainability appendix.

A hospital that can island does not experience a grid collapse as an event. It experiences it as a change of source, decided by a controller in milliseconds rather than by a transfer switch in seconds, and with solar and battery already carrying part of the load the generators may not need to start at all. Fewer starts is fewer failures, because the moment a standby generator is most likely to fail is the moment it is asked to start.

Segmentation is the other half. Several smaller islandable cells rather than one system means a fault, a fire or a maintenance shutdown takes a part of the hospital rather than the whole of it, and the clinical priority of each cell is decided in design rather than improvised at three in the morning, which is the load shedding argument from earlier made physical.

So the honest description of the target design is a layered, segmented microgrid: grid where present, solar across every surface that was doing nothing, battery sized to ride through and to shave the peaks, gas or trigeneration as the economic prime mover if a firm supply exists, and diesel held last because it is still the only fuel you can keep a week of on your own land. The alternatives do not replace the storable buffer. They reduce how often it has to be spent, which is exactly what this study has been asking of every buffer it has examined.

There are more choices here than diesel, and in this market the interesting ones are gas and heat. Nigeria has abundant gas and expensive diesel, and a hospital has a continuous demand for hot water, sterilising and cooling, which makes trigeneration unusually well suited: one fuel producing power, heat and, through absorption chillers, cold.

The right way to read the table is the last column. Every option gives something and every option leaves a specific thing unsolved, and the design is the combination whose unsolved parts do not overlap.

On cost per unit diesel loses to gas. On running cost and emissions it loses to solar. On response time it loses to a battery. It stays in the design anyway, and the reason is worth stating precisely because it is the thesis of this entire study in one property.

Diesel is the only one of them you can store. You can hold a week of it in a tank on your own land, under your own control, and take it out when everything else has failed. You cannot store a week of sunshine. You cannot store a week of gas pressure, because gas arrives by pipeline or by truck and both are somebody else’s reliability. A battery holds hours, not days, and that is a physical limit rather than a budget one.

So the layers arrange themselves by what each is good at rather than by what each costs. Grid first because it is cheapest when present. Solar to cut what you burn while the sun is up. Gas or trigeneration as the economic prime mover where a firm supply exists. Battery for the seconds and the peaks, sized larger than a pure uninterruptible supply because it is then doing two jobs instead of one. And diesel last, unglamorous and expensive, because it is the only buffer you can physically hold.

Which is the same sentence this study has now written four times in four currencies. Beds, people, stock and now power. The buffer is the thing that looks like waste and is not, and the one that survives the cost round is the one somebody wrote down the reason for.

All of this lands on land, which is why it belongs in the siting decision rather than in a later engineering package.

A generator hall, fuel storage with its separation distances, a gas compound if gas is chosen, a battery room with its own fire strategy, water treatment and storage, an oxygen plant, and a solar array large enough to matter. Add those together and the infrastructure footprint on a constrained urban site is a material fraction of the parcel, competing directly with clinical floor area and with whatever expansion capacity the group wants to preserve.

On a tight city site that competition may be the argument for a different parcel, and that is a conclusion worth reaching deliberately rather than discovering in year three when there is nowhere to put the second generator. It is also exactly the kind of interaction that a drawing cannot test and a run can: change the energy strategy, and the massing, the capital cost and the expansion options all move together.

Every availability figure on this page assumes the standby plant starts and carries the load when it is called. That assumption is the weakest part of the whole calculation, and it is an operational discipline rather than a capital item.

What makes it true is a test regime nobody enjoys: a short no load run weekly, an on load run monthly with the building actually transferred, and at least annually a full black start in which the incoming supply is opened and the hospital runs on its own plant while somebody watches what fails. The last of those finds the things the first two never will, which is usually a load that was quietly connected to the wrong board during a refurbishment three years earlier.

It also needs the people. A site generating its own power, treating its own water and producing its own oxygen is running industrial plant continuously and needs the engineering establishment to match, on shift. A second generator that nobody maintains is not redundancy, it is a second thing to fail, and it was counted in the availability figure as though it were not.

Power failure does not stop a hospital cleanly. It degrades it selectively, and each load fails on its own timescale into a different operational consequence. That is the part a capital business case almost never contains, because it is written in engineering terms and the cost lands in clinical and operational ones.

Read the column on the right rather than the one on the left. The engineering question is how long the load can survive. The operational question is what the day looks like afterwards.

The most consistently underestimated number in this whole area is the ratio between the length of an interruption and the amount of hospital time it costs.

Forty minutes without power does not cost forty minutes. It costs the theatre session that could not be safely restarted, the analyser run that has to be recalibrated before any result can be issued, the cold chain stock discarded on a rule rather than on evidence, the scans not done and therefore the decisions not made, and the backlog of movements that the porters absorb over the following day. A short interruption at nine in the morning is felt at four in the afternoon by people who have no idea why the day went badly.

This matters commercially as well as clinically, because it is the mechanism by which an infrastructure decision becomes an operational cost recorded under a different heading. The cancelled list appears as theatre utilisation. The discarded stock appears as waste. The extended stay appears as length of stay. Not one of them names the power event, which is the misattributed cost from the entity model appearing for the fourth time on this site.

Power is the one everybody thinks of and it is not the one that closes a hospital fastest. That is water. Theatres, sterile services and dialysis stop immediately without it, dialysis needs water treated to a standard rather than merely present, and a site in a constrained market needs storage, treatment and a tested quality regime rather than a connection.

Oxygen is the load with no tolerance at all and the one a generator does not solve. A site that cannot rely on delivered cylinders builds its own plant and holds reserve, and the pandemic demonstrated what the alternative looks like in several countries at once.

Then the ones that get left out of the brief entirely. Fuel, because the generators that guarantee the power are themselves a logistics operation with a lead time, and fuel is the single stockout that stops everything simultaneously. Connectivity, because the record, the imaging and increasingly the remote reporting all depend on it. Clinical waste, which needs a compliant route out of the building that exists in fact rather than in contract. Sewerage and drainage, in a city with a rainy season. And roads, because ambulance journey time, staff arrival and delivery reliability are all functions of traffic, and in Lagos traffic is a clinical variable rather than an inconvenience.

Three consequences, and they arrive in that order because each one constrains the next.

The design first. Generator halls, fuel storage, water treatment and storage, an oxygen plant, larger electrical and UPS rooms and the distribution to reach them all occupy real floor area and real position. That is massing, it is fixed at the earliest stage of the project, and it cannot be retrofitted into a plan drawn for a site with a dependable utility. A design imported from a market with good infrastructure is not merely optimistic in this one. It is the wrong shape.

Then the staffing. A hospital generating its own power, treating its own water and producing its own oxygen is operating industrial plant continuously, and it needs biomedical and estates engineering capability to match, on shift, permanently. That is an establishment line that clinical business cases routinely omit and that the availability figures above depend on entirely, because a second generator maintained by nobody is not a second generator.

And then the cost of a case. Self generation costs a multiple of grid supply, and it runs whether the theatre is busy or empty, so it behaves as fixed cost and falls on every patient. It belongs in the costing method explicitly rather than disappearing into an overhead apportionment, because a group that cannot say what infrastructure adds to a case cannot price one, defend one at diligence, or compare a Lagos site with a London one honestly.