Equipment replacement

Solar assets do not run 30 years on their original components. Inverters get replaced or refurbished, tracker drives wear out, monitoring electronics become obsolete, and a rooftop array may need to come off and go back up if the roof itself requires work. Folding these costs into a blended percentage of CAPEX loses their timing and their economics, in particular the fact that power electronics get cheaper in real terms while field labor gets more expensive.

The model prices every replacement as a discrete event.

Event cost

Each event's ticket in base-year money is:

$$C = Q \cdot c_0 \cdot (1+\lambda)(1+\tau)(1+\kappa) \cdot P$$

Term	Meaning	Typical range
Q	quantity (kWp, kVA, physical units)	from design
c_0	unit cost in base-year money	reference registers below
λ	install / logistics / recommissioning uplift	12–35% by site access
τ	import duty + VAT	0 where a renewable-energy import exemption applies (as in the current portfolio); its loss is a downside scenario
κ	contingency	10–15%
P	probability, for uncertain events	e.g. mid-life transformer at 25%

Dual-currency escalation with a deflation floor

The ticket splits into a USD-denominated hardware share s and a local-currency share 1-s covering installation labor, logistics, and permits (roughly 70/30 for utility-scale, 55/45 for rooftop). Each share escalates with its own index and a component-specific real delta Δ:

$$C(y) = C \cdot s \cdot \frac{U(y)}{U(y_b)} \cdot \rho_{usd}(t) \cdot FX_y \;+\; C \cdot (1-s) \cdot \frac{CPI(y)}{CPI(y_b)} \cdot \rho_{loc}(t)$$

where t = y - y_b and ρ(t) = (1+Δ)^t, floored at 0.6 cumulative for declines. Compounding a 3% real decline for 25 years would price hardware at 47% of today's cost, which supply-chain and commodity floors make implausible, so cumulative real decline is capped at 40%. Typical deltas: inverters and monitoring electronics decline 2 to 3% per year in real terms, copper-heavy gear like transformers runs at CPI or slightly above, and local labor runs 0.5 to 1.5% over CPI.

The USD share converts at the exchange rate of the event's year. Revenue is indexed in local currency while replacement hardware is dollar-denominated, and pricing that mismatch year by year keeps it visible instead of burying it in a blended escalator.

Downtime

Every replacement has downtime, and it lands in the same year as the cash outflow. Each event declares a generation haircut h (0.3 to 0.8% of annual yield for block-by-block utility swaps, 1 to 2% for a full rooftop shutdown), applied at expected value P · h to that year's generation and, through it, to energy revenue, certificates, and the fee.

Reference archetypes

The model ships component registers per archetype, instantiated against project capacity and editable per project:

Archetype	Key events	Undiscounted total (≈ % of initial CAPEX)
Utility ≥1 MWp, central inverters + trackers (30 y)	Inverter refurb yr 10, full replacement yr 19; tracker drives yr 15/27; SCADA, metering and security refresh yr 10/20; transformer yr 25 at P=25%	~19%
Utility, string inverters + trackers (30 y)	String replacement yr 13 (full) and yr 25 (partial, 60%); same balance of system	~23%
Utility, fixed-tilt (30 y)	As central, without tracker rows	~15%
Rooftop self-consumption 50 kWp–1 MWp (20 y)	Weighted monitoring refresh yr 10 (P=50%: half the range rides the inverter's integrated monitoring); bundled inverter replacement yr 11 (protections, connectors and meter included) with 0.8% downtime. No roof R&R: the asset is handed over to the client at end of contract	~12%
Rooftop self-consumption 30–50 kWp (20 y)	Simplified register, two events: bundled inverter replacement yr 11 (integrated monitoring included, no standalone datalogger) and a weighted electrical corrective yr 16 at P=50%	~12%
Rooftop self-consumption 50 kWp–1 MWp (30 y)	Two inverter cycles (yr 11 and 22) and two weighted monitoring refreshes (yr 10/20), plus an electrical corrective yr 24 at P=50%. No roof R&R	~30%

On 20-year horizons these totals annualize to roughly 0.5 to 0.8% of CAPEX per year; the 30-year rooftop register runs closer to 1% because two full replacement cycles compound three decades of price escalation. Industry experience puts a floor around 0.3% (below it, labor escalation or contingency is usually missing); totals well above these marks usually mean something is being double-counted against O&M.

Timing convention

Events are scheduled by operating year. An event falling in the valuation year is prorated by the remaining year fraction, an expected-value convention that assumes uniform timing within the year. A scheduled event can deviate from that assumption in either direction, so the engine emits a warning whenever a valuation lands in an event year and that NAV gets reviewed.