Module degradation and inverter replacement: what ages in a solar park and what it means for returns

A solar park has hardly any moving parts, and yet it ages: modules lose a little output year after year, inverters eventually fail and are replaced. Both sit in the yield model of a direct investment as assumptions, usually in a single line each, and both act mainly in the late years, when the loan has been repaid and the yield belongs to the equity.

This article puts the measurement series on module degradation in context, explains what a performance warranty actually secures, why the inverter is the one planned wear part, who pays for its replacement, how that is treated for tax and which questions you should ask any provider about it. It describes orders of magnitude for the market, not a specific project.

What ages in a solar park, and what does not?

Almost nothing quickly, but every component differently. The modules are designed for 25 to 30 operating years and lose a small, largely linear share of their output over that time. The inverter, which converts the modules' direct current into grid-compatible alternating current, is a power-electronics device with fans, capacitors and switches; its lifetime is well below that of the park. Mounting structure, cabling and transformer station last the full term if properly built. A battery storage unit, if there is one, ages by rules of its own, described in The battery as a real asset: lifespan, degradation and warranties of a grid-scale storage system.

How fast do solar modules lose output?

More slowly than most models assume. Fraunhofer ISE measured a mean annual degradation of nameplate output of about 0.15 percent across 44 larger, quality-assured rooftop plants in Germany; the institute explicitly considers the common planning assumption of 0.5 percent yield loss per year rather conservative for plants with quality assurance. The broadest international review, the compendium by Jordan and colleagues at the US research institute NREL with more than 11,000 degradation rates from almost 200 studies, puts the median for crystalline silicon at 0.5 to 0.6 percent per year, with means of 0.8 to 0.9 percent because a few poor products pull the average up. The majority of modules age linearly; only the worst units show an accelerated decline towards the end.

Two things belong with these figures. First, initial degradation: depending on the cell material, modules lose one to two percent in the first days of operation through light-induced degradation (LID); according to Fraunhofer ISE measurements, the declared nameplate output usually already refers to the state after that. Second, the distinction between module and plant: soiling, snow, shading and mismatch between modules come on top at plant level, which is why the IEA speaks of a performance loss rate for the park rather than module degradation. A yield assessment should show both separately; how to read an assessment as a whole is shown in Reading a yield assessment: what P50 and P90 mean for a solar park direct investment.

What the rate means over the term is easy to recalculate. The following table is our own calculation with a linear rate and the first operating year as 100 percent; it shows how much output remains in the tenth, twentieth and thirtieth year and how much electricity is produced in total over 20 and 30 years, each relative to a plant that would not age at all.

The table shows why the rate in the model matters more than its sign: between 0.25 and 0.5 percent lie around 2.3 percent of cumulative yield over 20 years and 4.5 percentage points of output in the twentieth year. That sounds small, but it falls almost entirely into the years in which the yield after repayment belongs to the equity. A rate of 0.75 percent, which the NREL compendium reports as a mean across poor products, already costs a tenth of the yield over 30 years.

What is different with TOPCon and other new cell technologies?

The old worries are largely solved, but there are new open points. The 2025 report of IEA PVPS Task 13 on degradation and failure modes of new cell technologies records that light-induced degradation and its variant at elevated temperature (LeTID) have practically disappeared in current products through the switch from boron to gallium doping, thinner wafers and cleaner processes; cell cracks have also lost significance thanks to multi-wire interconnection. What remains open for TOPCon and heterojunction cells are UV-induced degradation (UVID) and potential-induced degradation (PID). According to the report, the most important test standard IEC 61215 does not cover UV degradation because its UV test is too short and does not take recovery effects into account. There are UV-stable modules on the market, but certification alone does not prove it.

For a direct investment this means: with modules using TOPCon or heterojunction cells, asking for an extended UV test of the specific module type is justified, and a degradation assumption below 0.5 percent should be backed by test reports, not just the datasheet. The field values the report cites for heterojunction modules (around 0.5 percent per year versus 0.7 to 0.9 percent for conventional cells in the same review) are encouraging, but come from few plants.

What does the performance warranty really secure?

A minimum value, not the yield. Module manufacturers give two warranties: a product warranty against material and workmanship defects, usually over 10 to 15 years, and a performance warranty which, according to the project documents reviewed by the IEA, is typically formulated linearly over 20 or 25 years, and for current products often over 30 years. According to Fraunhofer ISE, a guaranteed maximum output loss of 10 to 15 percent over 25 to 30 years is customary; the warranty line therefore falls by around 0.4 to 0.6 percent per year, somewhat above what good modules actually lose. A failure in the sense of the technical reports only exists once a module falls below that line or safety is affected.

Three limitations come with it. The warranty is a claim against a manufacturer, not money in the account: it presupposes that the manufacturer still exists in twenty years and that the output loss is proven, usually by a measurement under standard test conditions in a laboratory or with mobile flash testers, the cost of which the operator bears at first. The nameplate output itself has a tolerance band, usually zero to plus five watts, which plays into the proof. And only in individual cases is the warranty backed by external insurance; where that is the case, the policy belongs in the data room.

The practical yardstick for a yield model is therefore simple: the degradation assumption should not be more favourable than what the warranty allows, unless the assessment justifies it with measured values for this specific module type. An assumption of 0.25 percent against a warranty line of 0.55 percent calls for an explanation.

Why is the inverter the wear part?

Because it is power electronics. An inverter switches thousands of times per second, works through temperature swings between night and midday peak and contains components with limited life: electrolytic capacitors, fans, relays. IEA PVPS Task 13 puts the lifetime of a PV inverter at 10 to 15 years and describes that operations management usually plans a replacement shortly after the tenth operating year. Fraunhofer ISE and Bern University of Applied Sciences cite the same order of magnitude, and manufacturer warranties (usually 5 to 10 years, extendable at a surcharge) cover only about half the time for which the modules are warranted.

The failure pattern follows the so-called bathtub curve: early failures in the first years, then a long phase of rare random failures, wear towards the end. The IEA cites an evaluation of operating data from around 2,000 commercial plants in the Solar Bankability project: after six years of operation about 10 percent of inverters had already been replaced, the large majority of them in the first three years; the annual replacement rate fell from over 4 percent in the first year to under 1 percent in the fifth. The wear-out phase had not yet begun in these data. For the investor this means: statistically, the first years are the critical ones, and that is the phase in which the manufacturer warranty and the contractor's defects liability should still apply.

How many devices reach the second half is shown by a study at Bern University of Applied Sciences, which evaluated 2,121 inverters in 1,195 plants in Switzerland using survival curves: after 15 years, 34.3 percent of the devices had had a first yield-relevant failure, after 18 years 59.1 percent were still running without one. The strongest influencing factors were the installation site (outdoor devices failed earlier than devices inside a building), the manufacturer (the observed failure shares differed by about a factor of three) and the use of power optimisers. The study looks at residential and small commercial plants with string inverters; for a solar park with central or large string inverters outdoors, the direction carries over, not the number.

According to the operations and maintenance guide by NREL and Sandia, central and string inverters have different failure and replacement profiles: if a central inverter fails, a large part of the park stands still, but the replacement per kilowatt is cheaper; with string inverters only a small part fails at a time, the swap is simpler, but the sum over the term is higher. As a rule of thumb the guide puts the replacement cost over 20 years at about twice the factory price of the device, because removal, installation, adaptation and lost yield come on top.

When is it replaced, and who pays?

Planned around year 10 to 15, paid from three pots. The first pot is the manufacturer warranty: it covers the early failures, and the IEA recommends using the warranty period as the device lifetime in cost planning. Extended warranties take the risk out of the bathtub but cost a surcharge; according to the IEA the extension can be more expensive than an early replacement paid out of pocket when calculated over 20 or 25 years, and it is only worth as much as the manufacturer's continued existence.

The second pot is the maintenance contract. A full-service contract includes spare parts and device replacement for a fixed annual fee; a pure inspection contract bills repairs on a time-and-materials basis, and the replacement stays with the operator. The IEA points out expressly that the contract must set out the procedure for warranty replacement, the allocation of costs and the reaction and repair times; even a creeping fault such as a defective fan with reduced output costs more yield over months than a clear total failure. Which contract type sits in the project is among the first questions, because it decides whether the replacement is an operating cost or a reserve.

The third pot is the reserve in the model. The O&M guide by NREL and Sandia treats inverter replacement as an expected normal loss that no insurance carries, and calculates a reserve account for it that grows with the probability of failure per year. In the yield model of a direct investment this appears either as an annual provision or as a one-off item in a specific operating year. What should not happen: the item is missing because the contractor has shifted it into a warranty that expires after ten years. The replacement itself is worth a look at the design, because with a new inverter the ratio of module to inverter capacity can be chosen afresh; the background is explained in South, east-west or vertical: which solar park layout pays off for investors.

How is the inverter replacement treated for tax?

As a rule as immediately deductible maintenance expense. For tax purposes a photovoltaic plant is a single asset that is depreciated over 20 years under the depreciation table of the Federal Ministry of Finance (item 3.1.6); the Bavarian State Tax Office also assigns components installed on the DC side to this unit. The inverter is therefore not a separate asset with its own depreciation but part of the plant. Where an existing part is renewed, this is regularly maintenance expense under R 21.1 (1) of the income tax guidelines (EStR), and that applies even if the new part is more modern or of higher quality than the old one. Production costs that would have to be capitalised and depreciated only arise under Sec. 255 (2) sentence 1 of the Commercial Code (HGB) in the case of an extension or a substantial improvement beyond the original condition.

For the direct investment this means: the replacement reduces the profit of the year in which it occurs in full as an operating expense. No new depreciation line arises, and there is no new investment deduction (IAB), because no new asset is acquired; how the depreciation of the plant itself runs is explained in Sonder-AfA §7g (5) vs. declining-balance AfA §7 (2): which combination, when?. Borderline cases are a replacement that actually enlarges the plant (say, markedly more AC capacity at the grid connection) or a repowering in which modules and inverters are renewed together; then the line to production costs may be crossed.

Where does this sit in the yield model, and what does it mean for returns?

In three lines that are rarely read together. The first is the degradation rate, which either comes from the yield assessment or is an assumption of the provider; according to the IEA review of project models it is usually 0.5 to 0.6 percent per year, and between 0.3 and 0.8 percent in the maintenance contracts reviewed. The second is the reserve for the inverter replacement, as an annual provision or a one-off item. The third is technical availability, the share of time in which the park actually feeds in; it captures the downtime around an inverter fault and is usually set higher again after the replacement.

How strongly the return reacts to these lines is shown by our table above: 0.5 instead of 0.25 percent costs around 2.3 percent of cumulative yield over 20 years, almost entirely in the years after repayment. The inverter replacement acts differently, as a one-off outflow in a year in which the profit is correspondingly lower. A model that shows both and names the assumptions is therefore not more pessimistic than one that leaves them out; it is only more honest about the second half of the term. How we read the cost items of an offer as a whole is described in Transparent costs: which fees a direct investment involves, and which ones are hidden.

Questions you should ask any provider about this

1. Which degradation rate is in the yield model, does it come from the assessment or from the provider, and how does it relate to the module manufacturer's warranty line?
2. Which module type and cell technology are installed, and are there extended test reports on UV-induced and potential-induced degradation for this specific type?
3. Who is the guarantor of the performance warranty, how is proof regulated, and is the warranty backed by insurance?
4. Which inverter type is used (central or string devices), how long is the manufacturer warranty, and has it been extended?
5. In which operating year is the inverter replacement planned in the model, and with what amount: as an annual reserve or as a one-off item?
6. Does the maintenance contract include spare parts and device replacement, or are repairs billed on a time-and-materials basis? Which reaction and repair times are agreed?
7. How is spare-parts stocking arranged, and what happens if the inverter manufacturer no longer supplies the type?
8. What availability does the model assume, and how is a creeping loss of output detected in the monitoring?
9. Is the inverter replacement treated in the model as an operating expense of the replacement year, and has that been agreed with the tax adviser?

How we check this in our project review

For every solar park we present to investors, we put the model's degradation assumption next to the module manufacturer's warranty line and next to the measured values from the technical reports, check whether the inverter replacement is in the model as its own item with year and amount, and read in the maintenance contract who actually bears it. We make no return promises; we show which figure rests on which assumption and how the return shifts if the assumption does not hold. That is exactly what we do in a non-binding initial consultation, gladly on the basis of a specific prospectus.