DEV Community: Jay Zhou

Solar Panel, Solar Inverter, Battery Storage: Why the System Decides the Result

Jay Zhou — Thu, 29 Oct 2026 05:52:17 +0000

A quotation can list every component at or above spec and still describe a solar system that underperforms by year three. The cause is rarely a poor solar panel or a weak solar inverter. It is the interface between them. In 2026 the industry has moved its competition from cell efficiency to system integration — TOPCon modules, liquid-cooled battery storage and grid-forming inverters are all answers at the system level, not the component level. For a buyer, that changes the checklist: five datasheets are less useful than five interfaces.

A completed 400 kW rooftop solar system in Butuan City, Philippines. The maintenance walkway was designed in, not added later.

A solar system is five interfaces, not five products

Most procurement documents are organised by product. A solar system is better described by what connects the products.

Array to inverter. String voltage, MPPT window, DC-to-AC ratio.
Inverter to grid. Voltage and frequency class, protection settings, export limits.
System to battery storage. Coupling architecture, charge and discharge windows.
Array to roof. Structural load, rear ventilation, access.
System to people. Cleaning, monitoring, spares — the operations and maintenance layer.

Every one of these is settled on paper, before anything ships. Four of them never appear in a price comparison.

Interface one: the solar panel and the solar inverter

A solar panel's voltage is not a constant. It falls as cell temperature rises, and it rises again on a cold, clear morning at first light. The solar inverter's MPPT window has to cover both ends of that range across the year, not at the datasheet's test condition. A string that sits comfortably inside the window at 25 °C can sit outside it at noon in the tropics, and the inverter will stop tracking where the array actually is.

The second half of this interface is the DC-to-AC ratio. Deliberately oversizing the array against the inverter is useful, because most of the day sits below peak irradiance. Oversize too far and the inverter clips at midday, so the extra panels earn nothing during the hours they were bought for. The ratio is a design decision with a cost on both sides — the arithmetic is in commercial solar inverter sizing.

Mismatch belongs here too. Two strings on the same MPPT with different orientation or different soiling will run at the weaker string's current. The loss is real, and it never appears on any single product's specification. It is also why solar panel specifications for hot, humid climates are read for temperature coefficient rather than headline efficiency.

Interface two: the solar inverter and the grid

This is the interface that most often decides whether a project works at all. The inverter has to match the site's voltage class and frequency — 50 or 60 Hz, confirmed rather than assumed — and its protection settings have to satisfy whoever operates the network.

On a weak grid, voltage drifts and the inverter trips to protect itself. It is doing its job; the design was wrong. That is why grid-tied, off-grid and hybrid are not lifestyle choices. A grid-tied system is the cheapest per kilowatt and the most exposed to a fragile supply. An off-grid system buys independence at the highest cost per kilowatt. A hybrid system carries battery storage and a grid connection at once, which is why it is the default where supply exists but is not reliable. The trade-offs are laid out in grid-tied vs off-grid vs hybrid solar systems.

The solar inverter and its protection panels: the interface between the array and the grid.

Interface three: the system and battery storage

Battery storage is the component most often bought for the wrong reason, and two mistakes account for most of it.

The first is buying a headline capacity figure. Usable energy depends on the depth of discharge the design allows and on the temperature the cabinet actually sits in, and neither is the number on the first page of the brochure. The second is buying storage for backup while designing it for peak shaving. Those are different duty cycles: one wants a long, shallow discharge a few times a year, the other wants a daily cycle. Sizing for the wrong one produces a system that is either oversized or short-lived.

This is where the 2026 shift is most visible. Liquid-cooled battery storage has been displacing air-cooled cabinets in high-ambient markets because tighter thermal control protects cycle life.

Battery storage sized against the load profile, not against a headline capacity figure.

Interface four: the array and the roof

A roof is a structure, not a blank surface. On a rooftop solar system the mounting has to match the sheet profile and the wind load for that site, and the array has to leave enough rear ventilation for the solar panel to shed heat. Pack the array tight against the roof and the cells run hotter, the voltage drops, and the loss surfaces in the interface above.

Access belongs to the same decision: on a large array the walkway decides whether the system can be cleaned and inspected at all.

Mounting rails and walkway sections in place before the first solar panel. Access is part of the mounting design, not an afterthought.

Interface five: operations and maintenance

Operations and maintenance on a solar system is usually sold as an after-sales service. It is better understood as a design output. If the monitoring points were specified, the strings are labelled, the inverter room is reachable and the walkway exists, routine service is a half-day job. If none of it was designed, the same work becomes a scaffolding contract.

In hot, humid and dusty markets, this is where the gap between two identical quotations actually lives.

What to change in the procurement document

Keep the datasheets, but add an interface page: for each of the five, write down who owns the decision, what the design condition is, and what the supplier confirms in writing. Then ask three questions — what happens to the string voltage at the hottest hour, what the inverter does when grid voltage drifts, and how a technician reaches row twelve with a hose.

A solar system is not the sum of its parts. It is the sum of what happens between them.

If you are specifying a commercial or industrial system for a hot, humid or weak-grid market, send the load profile and the destination — the reply is a design basis, not a catalogue. Request a system proposal

How a 400 kW Rooftop Solar System Actually Gets Built

Jay Zhou — Fri, 23 Oct 2026 10:23:25 +0000

A 400 kW rooftop array is not a difficult thing to buy, and it is not a difficult thing to install. What makes these projects go wrong is sequence: modules go up before the roof has been properly read, or a string gets re-arranged on the day because a row did not fit the way the drawing said it would.

This is the field sequence we follow on export projects, written up against a 400 kW factory rooftop build in Butuan City, Philippines. Nothing in it is exotic. That is the point — the failures are boring, and they are preventable.

Why a factory roof is a good candidate in the first place

A three-shift factory consumes most of its electricity during daylight. That is the entire argument for rooftop PV on an industrial load: generation and consumption overlap, so the array displaces grid energy directly instead of needing storage to time-shift it.

The grid stays connected as the balancing source. Solar covers what it covers; the utility covers the rest.

That arrangement is worth stating explicitly early, because it decides what the owner should expect. This is a bill-reduction system, not a backup system. If the requirement includes riding through outages, that is a different design conversation with different hardware in it — and discovering that difference after the array is on the roof is expensive.

Stage 1: the survey is an engineering task, not a sales visit

Before any hardware is discussed, the roof has to answer a fixed set of questions. I keep them as a list, because skipping one of them is usually where the trouble starts later.

Question	Why it is asked now
Usable area, after setbacks and obstructions	Sets the ceiling on array size before anyone quotes a number
Orientation and tilt of each roof plane	Sets the generation profile against the load profile
Shading from nearby structures, and at what time of day	A shaded string underperforms its neighbours and drags the MPPT input it shares
Condition of the metal sheets	Decides whether the roof needs work before it carries an array
Cable route from array to inverter room	Decides penetrations, trenching and how much of the DC run is exposed

The one that gets skipped most often is the last one. Cable routing gets treated as an installation detail, and then it turns out the only practical path crosses a roof penetration nobody wants to make.

Plan the walkway while you are still on the roof with a tape measure

A 400 kW array covers a large area. If access routes are not designed in from the start, cleaning and inspection become physically difficult within the first year — and modules nobody can safely reach are modules that quietly stop producing.

It is a layout decision, not an afterthought, and it is far cheaper to take at survey stage than to retrofit. The walkway has to be in the drawing before the rails are, because rails and walkway sections share the same fixing pattern.

Stage 2: substructure before modules

Rails, clamps and walkway sections go down first, and the mounting system is specified per project — against the sheet profile of that particular roof and against the local wind load. Neither of those is a default.

A clamp engineered for one standing-seam profile does not reliably transfer load on another. And wind load is a function of the site, the building height and the applicable local code, not of the array size.

Field note: if a mounting quote arrives before anyone has seen the roof profile, it is a placeholder. Ask what sheet profile and what wind speed the clamp spacing was calculated against.

Getting the substructure right is what makes the module installation fast. Every row is straight, every clamp point stays accessible, and the array remains serviceable for the life of the system.

Stage 3: modules row by row — and why the string plan is not negotiable

With the mounting complete, crews work row by row: modules are lifted, seated, clamped and connected into strings. On this build the roof crew and the electrical crew worked in parallel, one finishing the array while the other prepared the inverter and protection panels.

What must not happen is improvising the strings on the roof. The electrical reason is worth spelling out:

String length sets the voltage. The number of modules in series determines the string's open-circuit voltage, and the inverter's MPPT input has a window — a floor below which tracking is inefficient and a ceiling above which the inverter will not operate at all.
Cold raises open-circuit voltage. Module voltage carries a negative temperature coefficient, so the highest voltage a string ever sees is not on a hot afternoon but on a cold, clear morning. String length has to be chosen so the worst-case cold voltage stays inside the inverter's limit.
Strings sharing an MPPT input should match. If one string on an input is shaded or shorter than its neighbours, that input is pulled toward the weakest string rather than delivering the sum of the best ones.

All three are resolved on paper, before shipment. Re-arranging strings on the roof because a row did not fit means at least one of the three is now unchecked.

Stage 4: commissioning is a sequence, not a switch

The tests run in a specific order, and the order exists because each step protects the next one.

Insulation resistance — before anything is energised, confirm the DC installation is not faulted to earth.
Polarity — confirm each string is wired the way the drawing says, before it reaches the inverter.
String open-circuit voltage and current — compare against the expected value from the string plan. A string reading low is a wiring or shading problem, and it is much cheaper to find here than after handover.
Inverter start-up — only once the DC side is verified.
Monitoring configuration — so the owner has a baseline from day one, which is what makes later performance questions answerable at all.

Field note: the string-by-string voltage record taken at commissioning is the most useful document in the handover file. A year later, when somebody asks whether the array is underperforming, that record is the only thing that turns the question into an answer.

What goes in the handover file

For this project: the as-built layout, the bill of materials, and the documentation for each major component. Those three are what the owner's own maintenance people — or a third-party O&M contractor — will need, and all three are easier to produce during the build than to reconstruct afterwards.

For export projects the same discipline applies before shipment. Configuration records, inspection photographs and packing lists travel with the order, so that what arrives matches what was engineered.

The part that is not in scope

Worth stating plainly, because it is the most common mismatch between expectation and contract. On export projects the scope covers the system: design, hardware, factory testing, documentation and packing. It does not include installation crews abroad or a local service organisation. That work is done by the buyer's own contractor.

Two things follow from that.

First, the engineering package has to be complete enough for a contractor who has never seen the equipment to install it correctly — which is exactly why the string plan and the as-built drawing matter more here than they would on a domestic job.

Second, anything that depends on local conditions has to be confirmed locally, and confirmed before the inverter is specified rather than after it has shipped. Utility interconnection rules are the obvious one. How exported energy is treated — and therefore what the array is actually worth on the bill — is settled with the distribution utility, not nationally. For the Philippines, that process is described here.

Full field notes

The build this is written against, with the photographs, is documented here: How a 400 kW Rooftop Solar System Comes Together.

If you are working on a comparable roof, the sequence above is the part worth protecting. Hardware choices can be revisited; a compressed sequence usually cannot be.

Jay Zhou is a solar engineer working on commercial and industrial PV and storage systems for export markets. Photographs are from Force Solar installation records and describe the referenced project configuration; every system is sized against its own load data.

Why tropical C&I battery storage degrades faster than the datasheet suggests

Jay Zhou — Wed, 07 Oct 2026 09:51:47 +0000

A lithium iron phosphate cell rated for 6,000 cycles at 25 °C is describing a laboratory.

The moment that same cell sits in a rooftop enclosure in Lagos, Kano or Port of Spain, three
things change at once: ambient temperature rises, depth-of-discharge discipline changes
because the grid keeps failing, and the cell spends more time at high state of charge.

None of this is exotic. It is the normal operating condition of these markets. And the gap
between the datasheet number and the field number is where most storage projects quietly lose
their second year of economics.

Ambient temperature is the obvious one — and the most often skipped

Cycle-life figures are conventionally quoted at 25 °C. LiFePO4 tolerates heat far better than
NMC — that is its main argument — but "tolerates" is not "does not degrade". Capacity fade
accelerates with temperature, and the relationship is not linear.

The consequence for procurement is simple: ambient temperature at the installation site,
not the annual regional average, should be an input to the cycle-life claim. A system
specified against a 25 °C figure in a market with a 38 °C design summer temperature has been
specified against a number that will never be observed.

Ask for the degradation curve against ambient temperature, not a single cycle number. And ask
what the cabinet's thermal management is doing — the cell inside a cabinet in full sun can sit
10 °C above site ambient.

Procurement clause: cycle life shall be stated at the design ambient temperature of the
installation site, not at 25 °C, and the supplier shall state the measured cell temperature
inside the enclosure at rated load.

Grid availability changes the duty cycle, not just the backup time

The datasheet assumes a daily cycle at a defined depth of discharge. With unreliable supply, the
duty cycle is set by the grid rather than the designer.

This matters more than it first appears. When the grid fails repeatedly through the working
day, a backup system sized for a shallow daily cycle spends its life at deep discharge — the
condition that does the most damage per cycle. The same nominal 96 kWh cabinet can see a very
different annual throughput in Lagos than in a European installation with stable supply.

The common design error is sizing autonomy from the nominal outage rather than the distribution
of outages. The long tail is what consumes battery life: how often the outage lasts eight hours,
how often it lasts twenty.

Procurement clause: autonomy shall be sized from the measured outage duration
distribution for the site, not from a nominal or average figure, and the assumed annual cycle
count shall be stated explicitly.

Time at high state of charge is the quiet one

Less discussed, and relevant wherever solar oversupply is common. A system running from a large
PV array on a sunny grid day can sit near full charge for long periods. Calendar ageing at high
state of charge accumulates whether or not the battery is being cycled.

This is one argument for a mixed strategy — some storage for daily time-shifting, some reserved
for grid events. The reserve portion spends its life at moderate charge, and the frequently
cycled portion carries a shallower depth of discharge.

Putting it together

Take a 250 kW commercial load with a 400 kWh bank on a coastal West African site. Sized from a
datasheet cycle-life figure at one full cycle per day, the arithmetic looks comfortable. Adjust
for the three factors and it looks different:

Consideration	Effect on annual throughput
Design ambient well above the 25 °C reference	Reduce stated cycle life by the supplier's own temperature derating — in writing
Outage-driven deep discharge instead of a shallow daily cycle	Increases depth-of-discharge wear per cycle
Long periods at high state of charge	Calendar ageing in addition to cycle ageing
Combined	Annual throughput materially below the datasheet-derived figure

The conclusion is not that the system will underperform. It is that the procurement has to be
written so the supplier's stated numbers apply at real operating conditions. A supplier who
won't provide a temperature-derated cycle life hasn't understood the market they're quoting into.

Six clauses that close most of the gap

None of these cost anything to include in a tender document:

Cycle life stated at the installation site's design ambient temperature, not at 25 °C
Depth of discharge specified as a maximum, not a typical value
Autonomy sized from the measured outage duration distribution, not an average
Annual cycle count stated as an explicit design assumption in the contract
Enclosure thermal management described, including measured internal cell temperature
Warranty written against the specified duty cycle, with a defined remedy if actual operating conditions exceed the stated assumption

Why it matters commercially

Storage economics in these markets are usually modelled on datasheet cycle life. When field
performance falls short of the model, the project looks like a bad investment — when in fact the
model was built on a number that never described the site.

Requiring suppliers to state performance at real operating conditions isn't adversarial. It's
the only way both sides have the same expectation.

I'm a solar engineer at Force Solar, where we build commercial and industrial solar systems from
10 kW to 1 MW and LiFePO4 storage cabinets from 96 kWh to 5 MWh. Technical questions welcome:
https://forcesolarpower.com/contact/

Full version with the site-specific worked example: https://forcesolarpower.com/blog/battery-degradation-tropical-ci/