Abstract
The growth of urban solar installations will increase the demand for decommissioning and recycling. This review examines treatment technologies for end-of-life photovoltaic (PV) modules, recycling costs, and the responsibilities of market participants. It draws on research publications, IRENA and IEA PVPS reports, national statistics, legislation, and company disclosures available through 9 November 2026. Industrial recycling still relies mainly on mechanical treatment. Recovering silver and silicon suitable for use in manufacturing requires delamination and purification. Results from laser and abrasive delamination and aqueous perovskite recycling have yet to be demonstrated in continuous industrial operation. Comparable global recycling statistics are unavailable. In France, 83.9% of the module mass treated in 2025 went to material recycling. The main economic barriers are low plant utilization, dismantling and transport costs, low glass prices, and weak incentives to recover valuable components. A comparison of seven cities identifies three deployment models. The proposed full solar life cycle city model combines an installation registry, pre-decommissioning assessment, batch consolidation, and supplier requirements. It also specifies metrics for tracking where recovered materials go and identifies tasks that artificial intelligence can support.
Keywords photovoltaic modules; solar panel recycling; delamination; silver recovery; circular economy; urban solar deployment; smart city; extended producer responsibility; artificial intelligence; life cycle
Full Article
22 pages · 9 tables · 57 references
Read the text version
Page 1
Review Article
Photovoltaic Module Recycling in Urban Solar Energy Systems:
Technologies, Economics, Regulation, and Artificial Intelligence
Shukhrat Yuryevich Shipulin
LLC SARDOR-TECH, Tashkent, Uzbekistan
E-mail: shukhrat.shipulin@SARDOR-TECH.uz; ORCID: 0000-0001-7125-0654
Abstract
The growth of urban solar installations will increase the demand for decommissioning
and recycling. This review examines treatment technologies for end-of-life photovoltaic (PV)
modules, recycling costs, and the responsibilities of market participants. It draws on research
publications, IRENA and IEA PVPS reports, national statistics, legislation, and company
disclosures available through 9 November 2026. Industrial recycling still relies mainly on
mechanical treatment. Recovering silver and silicon suitable for use in manufacturing requires
delamination and purification. Results from laser and abrasive delamination and aqueous
perovskite recycling have yet to be demonstrated in continuous industrial operation.
Comparable global recycling statistics are unavailable. In France, 83.9% of the module mass
treated in 2025 went to material recycling. The main economic barriers are low plant utilization,
dismantling and transport costs, low glass prices, and weak incentives to recover valuable
components. A comparison of seven cities identifies three deployment models. The proposed
full solar life cycle city model combines an installation registry, pre-decommissioning
assessment, batch consolidation, and supplier requirements. It also specifies metrics for
tracking where recovered materials go and identifies tasks that artificial intelligence can
support.
Keywords: photovoltaic modules; solar panel recycling; delamination; silver recovery; circular
economy; urban solar deployment; smart city; extended producer responsibility; artificial
intelligence; life cycle
1. Introduction
Solar panels incur recycling costs long after their purchase and installation. According to
the U.S. Environmental Protection Agency (EPA), panels typically last more than 25 years.
Damage or system upgrades can lead to earlier removal [1]. Testing can identify removed
modules that are suitable for repair or reuse before they enter the waste stream [2], [3].
A ground-mounted solar plant usually has one operator and concentrates its modules at a
single site. In cities, modules are spread across buildings and may have many owners.
Recovering even a small batch can require several site visits and separate arrangements for
rooftop access. Studies in Amsterdam link the potential for urban solar generation to the built
1Page 2
environment, property ownership, and the condition of the distribution grid [4], [5]. If contracts
leave dismantling and recycling costs unassigned, building owners, property managers, or
municipal waste services may ultimately bear them.
Glass and aluminum account for a large share of the mass recovered during recycling.
These materials are easier to recover as secondary feedstocks than silver or silicon that meets a
downstream buyer's purity requirements [6]. This review distinguishes material extraction from
the production of marketable feedstocks and tracks whether those feedstocks enter PV
manufacturing or other industries. A closed PV material loop requires recovered materials to
re-enter module production [3].
Early recyclers used equipment designed for glass, metals, and electronic waste. Most
processes documented in an IEA PVPS study from 2016–2017 were not designed specifically
for PV modules [6]. The EU Waste Electrical and Electronic Equipment (WEEE) Directive
2012/19/EU made producers responsible for organizing and financing the management of waste
from their products [3], [7]. In 2016, IRENA and IEA PVPS projected 60–78 million tonnes of
cumulative panel waste by 2050, depending on the scenario [8]. The 2020s saw the emergence
of dedicated processes for delamination, silver recovery, and silicon purification. All seven
cases in the 2024 IEA PVPS report involved dedicated PV recycling processes [6].
Previous reviews cover recycling technologies, legislation, and waste projections in detail
[9]–[11]. Less attention has been paid to cities. This article examines how cities can manage
modules from equipment selection and procurement through dismantling and the sale of
recovered materials. It identifies technical, economic, and regulatory barriers and proposes a
municipal management model.
The review has six objectives:
– Compare commercial processes and emerging technologies in light of the available
evidence.
– Define recycling metrics and identify when they can be compared.
– Examine solar deployment models in leading cities.
– Identify economic and regulatory barriers and describe manufacturers' strategies.
– Assess potential applications of artificial intelligence (AI).
– Propose a municipal model for the full solar life cycle and define metrics for evaluating
it.
2. Methods
2.1. Review design and scope
This study is a narrative review. It did not use a PRISMA protocol, an exhaustive database
search, a formal risk-of-bias assessment, or a meta-analysis. The review prioritized publicly
available IRENA and IEA PVPS reports and primary peer-reviewed studies, followed by
2Page 3
review articles, legislation, and publications from governments and companies. Trade
publications helped establish facility start-up dates and when results were released. The review
covers information available through 9 November 2026.
The scope is limited to PV modules and excludes inverters, batteries, and mounting
structures. The main focus is on crystalline silicon (c-Si). Thin-film modules (CdTe and CIGS)
and perovskite devices are examined separately because their composition and treatment
requirements differ [2], [10]. The cities were selected to illustrate different deployment models.
2.2. Evidence categories
The review uses four evidence categories (Table 1). Laboratory delamination tests
characterize an individual operation. Establishing the profitability of an entire line requires
additional evidence. Material yield and product purity also measure distinct aspects of process
performance. IEA PVPS reports discuss limits on the comparability of industrial data [6], [57].
Table 1. Evidence categories used in this review
Category Definition
Industrial experience The process operates at a commercial facility. Operating scale and
reporting detail vary.
Experimental result The authors measured the result on specified samples in a laboratory
or pilot facility.
Model estimate The value was calculated under stated assumptions.
Proposed approach The approach has been proposed or anticipated. Its performance
across the full industrial process remains unverified.
2.3. Data verification and limitations
The review examined how each reported indicator was defined. In 2025, the French
organization SOREN collected 13,760 tonnes of end -of-life PV modules and treated 12,736
tonnes. Of the mass treated, 10,684 tonnes went to material recycling, equivalent to 83 .9%.
Using the total mass collected as the denominator gives 77.6%. This difference in reporting
basis can distort comparisons across facilities and national recycling systems.
The review also distinguished between design capacity and actual throughput. Design
capacity indicates how much waste a facility can accept under its design assumptions. Actual
throughput records how much it processes over a given period. The two figures a re reported
separately. Manufacturer-reported figures were interpreted with caution, particularly where
independent audit results were unavailable.
Bibliographic details and digital object identifiers (DOIs) were checked against publisher
pages and the publications themselves. The study relied on available sources. No laboratory
experiments, facility visits, or independent measurements were undertaken .
The municipal management model proposed in Section 4 draws on experience from
several countries. Its practical applicability could be assessed through a pilot study in a city
3Page 4
with a registry of PV installations and data on equipment condition, dismantling costs, and
module transport costs.
3. Results
3.1. Projected waste volumes and available statistics
Under IRENA's 2026 scenario for limiting warming to 1.5 °C, cumulative end-of-life
panel mass could exceed 200 million tonnes by 2050, with annual waste exceeding 25 million
tonnes [2]. The estimate depends on deployment rates and module lifetimes. Comparing the
figure directly with the 2016 projection would confound projected waste growth with
differences in scenario assumptions.
Current collection volumes remain far below these projections. IEA PVPS reported
48,395 tonnes across 18 European countries in 2022. Data for the 2025 report were collected
mainly in 2024 [3]. Investors must time capacity expansions carefully to avoid maintaining
underused lines while waiting for waste volumes to rise.
France publishes detailed national data (Table 2). SOREN collected 13,760 tonnes of
modules in 2025 and 9,477 tonnes in 2024. Of the 12,736 tonnes treated, 10,684 tonnes went
to material recycling, 1,298 tonnes to energy recovery, and 754 tonnes to disposal [12], [13],
[55], [56]. These streams accounted for 83.9%, 10.2%, and 5.9% of the treated mass,
respectively. Material and energy recovery together accounted for (10,684 + 1,298) / 12,736 ×
100% ≈ 94.1%. The material recycling share alone was 83.9%. The reported tonnages give a
year-on-year increase in collection of (13,760 − 9,477) / 9,477 × 100% ≈ 45.2%, whereas
SOREN's 2025 report and the trade press report 40% [13], [55]. Both calculations use the
published rounded masses.
Table 2. Verified recycling indicators and their interpretation
Entity and period Indicator Interpretation
France, SOREN, 13,760 t collected, 12,736 t treated, and Three distinct indicators. The percentage
2025 10,684 t sent to material recycling (83.9% uses treated mass as its denominator.
of treated mass)
First Solar, 2025 59,522 t of PV materials collected, with The input stream includes manufacturing
an approximately 95% recycling rate scrap. The company-reported figure has an
uncertainty of approximately ±2%.
First Solar, 2026 Network design capacity of 136,000 Neither figure measures actual annual
report t/year. Cumulative total of approximately throughput.
500,000 t.
SOLARCYCLE, 480,406 panels recycled Actual throughput as reported by the
2024 company
SOLARCYCLE, Thousands of panels per week. Target of Current throughput, target throughput, and
Cedartown, January 1 million panels/year by the end of 2026. design capacity are distinct measures.
2026 Design capacity of up to 5 GW of panels
per year.
4Page 5
Entity and period Indicator Interpretation
18 European 48,395 t of module waste collected Historical collection figure. It does not
countries, 2022 measure recycling in Europe in 2025–
2026.
For 2025, First Solar reported the following shares of collected mass: glass, 92%; metals
other than semiconductor materials, 1.9%; semiconductor materials, 1.3%; and other materials,
0.3% [14]. The previously published figure of 41,921 t/year comes from t he 2024 IEA PVPS
dataset [6]. SOLARCYCLE reports two different indicators: recovery of 96% of material value
and diversion of 100% of waste from landfill [15], [16]. Neither measures the share of mass
sent to material recycling.
Available data provide no reliable denominator for calculating a global recycling rate for
2025–2026. National reports often give collected waste mass, while company reports give
incoming equipment mass or the mass of recovered materials. Reporting practic es for
manufacturing scrap, reuse, and energy recovery also vary. The same waste may appear in two
reports. For example, First Solar is a partner of France's SOREN [13]. Adding their figures
could result in double counting.
3.2. Module composition and established recycling processes
A silicon module contains glass, solar cells, a polymer encapsulant (typically EVA),
electrical interconnects, and a junction box. Many designs also incorporate an aluminum frame
and a polymer backsheet. Glass–glass modules use a second glass sheet in plac e of the
backsheet [2], [17]. Recycling must separate strongly bonded layers while limiting cross -
contamination. CdTe modules require controlled recovery of cadmium and tellurium. For CIGS
modules, the target elements are copper, indium, gallium, and selen ium. Treatment of lead-
containing perovskites must recover functional materials and prevent lead losses [6], [10], [18].
Module type should therefore be identified before processing.
Table 3. Main PV module recycling methods (based on [6], [9], [19])
Method Operating principle Advantages Limitations
Mechanical size reduction Crushing and Industrial maturity, high Material mixing, dust,
and sorting separation by particle throughput, and low energy and lower value for some
size and density demand outputs
Thermomechanical Separation of layers Can preserve intact glass Performance depends on
delamination with a heated tool panel design and
condition
Thermal treatment Polymer Liberates glass and cells Energy demand, off-gas
decomposition or treatment, and loss of
removal polymers
Chemical delamination Treatment with Selective breaking of bonds Reagent consumption,
solvents and other treatment time, and
reagents solution regeneration
5Page 6
Method Operating principle Advantages Limitations
Hydrometallurgy Leaching and metal Selective recovery of Wastewater, chemical
recovery valuable components hazards, and additional
purification
Pulsed-light and laser Localized treatment at May reduce material mixing Scale-up and economic
treatment layer interfaces viability remain to be
demonstrated
A mechanical line removes the frame, cables, and junction box before shredding the rest
of the panel and sorting the material. The method is established commercially and requires
relatively little energy. It does not typically produce separate, high -purity silicon and silver
fractions [6], [57]. Mixed mineral fractions and contaminated glass can find uses outside PV
manufacturing.
In equipment made by Japan's NPC, a hot knife at approximately 300 °C softens the EVA
and separates the glass from the laminate. For one model, the company reports a processing
time of about 60 s per panel when the glass is intact [20]. The figure covers d elamination alone.
It excludes metal extraction and material purification.
Thermal delamination removes polymers by heating. T. Dobra and colleagues obtained
solid products of comparable quality after 65 min at 500 °C and 33 min at 600 °C [21]. Raising
the temperature by 100 °C reduced the treatment time by about 49%. Estimating energy savings
would also require data on heating power and heat losses. Panels with fluorinated backsheets
require prior separation and emissions control. Polymer combustion with energy recovery is
counted separately from material recycling [3], [17].
Chemical treatment can remove polymers or treat a separated metal -bearing fraction.
Leaching brings silver into solution. The metal is then recovered by precipitation or
electrochemical methods. The fraction dissolved during leaching and the yield of the f inal
product are separate measures [17], [22]. Researchers are studying solution regeneration and
the treatment of enriched fractions to reduce reagent consumption. Assessing environmental
performance requires material and energy balances for the process [ 22].
3.3. Research advances in 2025 –2026
Laser delamination of aged silicon modules. In a manuscript accepted for publication
in November 2026, B. N. Alsulami, J. Kettle, and colleagues applied a laser at the EVA –silicon
interface [19]. The process produced three fractions: a largely intact EVA film with residual
metal, busbars, and a silicon-rich powder. Aluminum was removed, and silver was extracted in
a medium containing iron(III) chloride and choline chloride. Selectivity exceeded 99% for the
REC240PE and 95% for the CS6P-245P. These values describe extraction selectivity. Whole -
module silver recovery and product purity are separate measures. Localized heating may use
less energy than bulk heating, but industrial energy data are not yet available.
6Page 7
Layer-by-layer abrasive separation. In the EVERPV project, F. Coustier and
colleagues at CEA used abrasive belts to remove the rear layers while keeping the glass intact
[17]. They collected the backsheet separately from the encapsulant, cells, and interconnects.
Tests covered more than 400 end-of-life modules from four manufacturers. At a belt speed of
2.3 m/min, the pilot line processed about one panel per minute, excluding subsequent powder
treatment. The method was described for intact glass–backsheet modules less than 1,300 mm
wide. Laboratory tests yielded silicon at 99.9% purity and a silver-bearing product containing
about 90% silver. Recovery yields were around 98% for both materials. The industrial silver
recovery estimate of 99.9% remains a projection. Polymer fractions also require further
preparation before shipment to a recycler.
Aqueous perovskite recycling. X. Xiao and colleagues dissolved the perovskite layer in
a heated aqueous medium with additives. Crystals precipitated as the solution cooled [18]. The
process recovered 99.0 ± 0.4% of the perovskite material by mass. Devices made from
recovered material achieved a mean power conversion efficiency of 21.9 ± 1.1%, compared
with 21.8 ± 0.8% after five degradation and recycling cycles. The means differ by 0.1
percentage points. Further statistical information would be needed to assess the significance of
that difference. The tests used research devices, including devices subjected to accelerated
degradation. Commercial readiness for full-size perovskite and tandem modules remains
unproven.
Use of recovered silicon in batteries. Y. Zhao and colleagues (3 November 2026)
blended recovered silicon with graphite to make lithium-ion battery anodes. The mixture was
heat-treated at 800 °C in nitrogen [23]. After 100 cycles at 0.5 A/g, the composite with a Si:C
mass ratio of 2:8 retained 47.5% of its capacity, compared with 12.8% for the untreated waste
silicon. The difference was 34.7 percentage points. Even the best sample lost 52.5% of its
capacity. The composite outperformed the control, but the test did not establish durability for
commercial use.
3.4. Industry participants and manufacturer strategies
Table 4 summarizes the operations in which industry participants specialize. Incomplete
reporting precludes comparison by actual recycling throughput. Equipment suppliers, collection
organizations, and recycling plants serve different functions.
Table 4. Selected participants in the recycling industry (based on [3], [6], [15], [24]–[28])
Participant Country Role Interpretive limitation
First Solar United States, CdTe module manufacturer Figures apply to the company's
international with in-house recycling technology.
network
SOLARCYCLE United States c-Si recycling and material Opening a plant does not establish
recovery that it is operating at design capacity.
7Page 8
Participant Country Role Interpretive limitation
ROSI France, expanding in Recovery of valuable silicon The announced figures of 3,000
Europe and metal fractions t/year (France, 2022) and 10,000
t/year (planned plant in Spain) are
capacities.
Reiling Germany Industrial mechanical The process does not necessarily
recycling yield high-grade Ag and Si.
NPC Japan Hot-knife equipment Equipment supplier for the separation
step
LONGi China, international PV CYCLE participation and Membership does not disclose actual
markets partnerships with regional recycling throughput.
operators
Trinasolar China Demonstration module using Demonstration unit only
recovered materials
PV CYCLE, Europe Collection, compliance, and Scheme administrators; no plant
SOREN coordination operation
Manufacturers use four approaches to managing end -of-life modules. The first is in-house
recycling. First Solar operates facilities in the United States, Malaysia, Vietnam, Germany, and
India. A third party separates and refines cadmium and tellurium [24]. For sales before 2013,
the company used a prefunded program. Current contracts use pay -as-you-go pricing. Owners
are responsible for removal, packaging, and delivery, and prices vary by region. The program
does not include free rooftop pickup.
The second approach uses collective collection schemes and partnerships. LONGi reports
participating in PV CYCLE and working with Soren, Open, and 3Drivers. In Brazil, it works
with PV CYCLE and SunR [27], [29]. In 2025, LONGi declared equipment with a combined
rated capacity of 4,589.27 MW under the European WEEE system [30]. The figure records
producer obligations and does not measure waste treated.
The third approach uses recovered materials in new modules. In October 2024, Trinasolar
announced a TOPCon module made using recovered silicon, glass, aluminum, and silver. The
company reported an efficiency of 20.7% and an output above 645 W [28]. It described the
prototype as the first “fully recycled” c-Si module. The announcement gives insufficient detail
on how much of the component mass came from secondary sources, whether the module could
be mass-produced, or what it would cost. It remains a demonstration using recovered materials.
The fourth approach is to design modules for recycling. Manufacturers are reducing the
use of materials that are difficult to separate, using fluorine-free backsheets, increasing recycled
content, and extending product life. LONGi reported that rear glass used in its modules
contained 10–30% recycled glass in 2023 [27]. Manufacturing waste also needs to be tracked
separately. For example, the more than 2,200 t of packaging material recycled by Trinasolar
did not come from the modules themselves [31].
8Page 9
3.5. Leading cities and models of urban solar deployment
A city with the highest installed capacity may rank lower in capacity per capita or in the
share of buildings equipped with PV. Table 5 compares deployment models across cities using
metrics relevant to each.
Table 5. Cities illustrating different models of urban solar deployment
City Deployment model or Indicator or benchmark Implications for recycling
strength
Honolulu, Capacity per capita Highest among surveyed Per-capita metrics do not capture
United States cities in the U.S. Pacific equipment age or condition.
region, 2022 [32]
Las Vegas, Capacity per capita Highest among surveyed Continuous supply requires grid
United States cities in the U.S. Mountain integration. Installed capacity
region, 2022 [32] alone does not measure it.
Tokyo, Japan New construction Requirements for large Procurement can require
requirements housing suppliers since disclosure of material
April 2025. Target of about composition and specify take-
1 GW on residential back terms.
buildings by 2030 [33], [34]
Singapore Centralized deployment Panels on more than 4,300 A single procuring agency
on public housing HDB blocks (January 2025) simplifies condition assessment
[35] and take-back contracts.
Rotterdam, Size of urban PV fleet More than 1 million panels The installed fleet needs
Netherlands (December 2024) [36] portfolio-level management.
Amsterdam, Rooftop PV in a historic 315 MW installed, with a Deployment is shaped by shared
Netherlands city target of 550 MW by 2030 ownership and heritage
(TU Delft study) [4] protection.
Dubai, United Utility-scale solar park Mohammed bin Rashid Al Dismantling can be planned by
Arab Emirates Maktoum Solar Park phase. Part of the capacity is
capacity of 3,860 MW in concentrating solar power.
2026, with a target above
8,000 MW by 2030 [37]
Tokyo has applied, since April 2025, requirements to large housing suppliers that deliver
at least 20,000 m² of floor area annually. Solar exposure and roof geometry affect panel
placement. The requirements do not apply to all existing homes [33]. The Tokyo Met ropolitan
Government projects about 1 GW of new residential capacity by 2030, including around 700
MW attributable to the new requirements. It expects annual emissions to fall by approximately
430,000 t CO₂ [34]. These figures are projections.
Amsterdam could generate more than 6.9 PJ of electricity per year (about 1.92 TWh),
according to TU Delft. This is roughly half of city electricity use under the study's comparison
baseline. Installations are less common on historic buildings and multifamily buildi ngs with
shared ownership [4]. In 2026, Amsterdam extended support for panels that meet composition
and carbon footprint requirements. It also developed a module reuse program for households
with limited means [38]. These measures support reuse. Closing th e material loop also requires
a route into manufacturing.
9Page 10
Rotterdam reported in December 2024 more than one million installed panels and the
largest solar capacity among the four largest Dutch cities [36]. Managing an installed base of
that size requires data on module age and condition as well as scheduled roof repairs.
Singapore installs panels on public housing through the Housing & Development Board
(HDB). As of 31 March 2024, 325 MWp was in operation. Eight tenders awarded about 455
MWp, and the 2030 target is 540 MWp [39], [40]. The electricity powers lifts, lighting, pumps,
and other common services. Surplus electricity is exported to the grid. Individual apartments
still rely on the grid [41]. Solar-ready designs reserve space for panels, supports, and electrical
infrastructure from the outset [40].
Dubai illustrates the utility-scale solar park model. The 950 MW fourth phase of the park
includes 700 MW of concentrating solar power and 250 MW of PV [37]. PV accounts for 250
/ 950 × 100% ≈ 26.3% of capacity in this phase. Waste projections for PV modules sh ould
therefore use the 250 MW PV capacity.
For future dismantling, the cases illustrate three organizational models. In distributed
rooftop systems, the priorities are identifying owners, arranging building access, and
consolidating small batches. Centrally managed installation programs can standar dize service
and take-back terms. Large parks require phase -by-phase replacement schedules and separate
collection for different module technologies. The classification was developed for this review
and is not an industry standard.
3.6. Economic barriers
IEA PVPS reports that recycling costs are high relative to revenues from recovered
materials [42]. Seven factors affect the balance.
Low and variable utilization. Over a given operating range, a plant's annual fixed costs
can be treated as constant. As actual throughput falls, fixed cost per tonne increases:
cfixed = Cfixed / Mactual (1)
Here, cfixed is the fixed cost per tonne (currency units/t), Cfixed is annual fixed cost
(currency units), and Mactual is annual treated mass (t). With annual fixed costs held constant, a
plant operating at 50% of design capacity has twice the fixed cost per tonne of a fully utilized
plant. The illustration excludes changes in variable costs. Design capacity alone says little about
profitability [42].
Logistics. Urban cost estimates should separate dismantling, consolidation of small
batches, and transport. Recycling fees may exclude rooftop work and lifting equipment [42]. A
larger plant increases throughput and requires a larger collection area. The optimal service
radius depends on where waste arises and what it costs to transport.
Low value of bulk materials. In a silicon module, inexpensive glass accounts for a large
share of the mass. Valuable metals are present in small quantities and take several steps to
recover [3]. A high mass yield can therefore generate modest revenue.
10Page 11
Advanced treatment costs. Chemical purification improves product quality and adds cost
[43]. The step pays for itself when incremental revenue exceeds incremental cost. Otherwise, it
requires separate funding.
Markets and impurities. Even a high silicon content may fall short of the specifications
for new solar cells [42]. Before purchasing equipment, recyclers need to establish prospective
customers' requirements for composition and purity.
Weak incentives to recover valuable components. Heavy fractions can contribute most of
the mass needed to meet WEEE targets. IEA PVPS notes that such targets provide limited
incentive to invest in costly recovery steps for minor fractions [43].
Changes in module composition. Using less silver in manufacturing reduces the potential
recycling revenue per module [2], [43]. Long -term financial assessments should account for
future feedstock composition as well as metal prices.
A recycler's financial result over a given period can be represented by the following
illustrative equation:
Π = Pgate + Rmaterial − Clogistics − Cprocessing − Cresidue − Cfixed (2)
Here, Pgate is gate-fee revenue, Rmaterial is revenue from the sale of recovered materials,
and Clogistics, Cprocessing, Cresidue, and Cfixed are logistics, processing, residue management, and
fixed costs over the same period. All terms are expressed in the same currency. At break-even
(Π = 0), the minimum gate-fee revenue equals total costs less material sales revenue. When
material sales fall short, the owner, a producer responsibility scheme, or another contractual
arrangement covers the difference. Process improvements may still leave a need for gate fees.
3.7. Regulation
Extended producer responsibility (EPR) determines who organizes and pays for the
management of product waste. Producers may meet their obligations independently or through
a collective scheme [44]. For solar panels, the financing arrangement must have a long time
horizon, as decades can elapse between sale and decommissioning.
In the EU, PV panels have been within the scope of WEEE since 13 August 2012.
Directive (EU) 2024/884 clarified producer obligations after the Court of Justice of the
European Union ruled against retroactive application. Financial responsibility applies to
products placed on the market from that date [44]. The amendments call for assessing whether
panels should have their own category, whether collection targets should reflect service life,
and how to cover costs if a producer becomes insolvent. These matters remain subject to further
review [44].
In the United States, federal requirements differ from state rules. The EPA describes the
inclusion of solar panels under the universal waste regime as a proposed change. California and
Hawaii already classify qualifying panels as universal waste [45]. This approach simplifies
certain hazardous-waste procedures and does not require producers to fund recycling. Panels
11Page 12
are safe in normal use. Whether a discarded panel is classified as hazardous waste depends on
its composition and the leaching of metals such as lead and cadmium [1].
Washington State's program, adopted in 2017, requires manufacturers to fund take -back
and recycling without charging panel owners [46]. Under the 2025 amendments, plans are due
by 31 January 2030, and the ban on selling panels without an approved plan take s effect on 31
January 2031 [47]. The timeline illustrates the interval that can separate enactment from
implementation.
Table 6. Scope and limits of regulation
Regulation can Regulation alone cannot guarantee
Assign responsibility and identify a funding Plant profitability
source
Set collection and treatment requirements High-purity silicon or glass
Mandate reporting Reliable reporting without oversight
Restrict improper handling Buyers for recovered materials
Create demand for recycling services Commercial readiness of new technologies
3.8. Artificial intelligence in module life cycle management
AI in PV recycling is most useful for decisions involving heterogeneous inputs, such as
selecting modules for reuse, identifying materials, and forecasting waste volumes [11], [48],
[49]. Delamination and metal recovery are still physical and chemical proc esses. The evidence
for AI varies by application (Table 7).
Table 7. AI applications across the module life cycle and strength of evidence
Stage Task Role of AI Evidence
Intake Select modules for reuse, Condition classification Company-reported
repair, or recycling
Planning Forecast where and when Deployment forecasts Peer-reviewed studies
waste will arise combined with lifetime
models
Treatment Detect incomplete Analysis of process Company-reported
separation deviations
Sorting Distinguish materials and Recognition and Company-reported
contaminants classification of material
streams
Robotic disassembly Adapt to heterogeneous Machine vision Potential application
panels
Process development Select treatment Models fitted to Research application
conditions experimental data
SOLARCYCLE reported that its AI model helped select more than 1,000 previously
operated modules for a 500 kW power plant in Texas [48]. The company also reports using AI
to monitor separation and identify materials. It has not published classification erro r rates, the
12Page 13
subsequent service life of the modules, or results from a control process without AI. The
available information does not support a quantitative estimate of the algorithm's effect.
A. Franzoni and colleagues used installation data for 2008 –2024 and socioeconomic
indicators from 20 Italian regions [49]. They compared linear regression, ARIMA, a
backpropagation neural network (BPNN), and a long short -term memory network (LSTM). The
BPNN performed best. It forecast installed capacity, which the authors then converted to
estimated waste volumes using a Weibull distribution. Four coauthors work for the energy
company A2A S.p.A., an affiliation relevant to assessing the study's independence .
4. Discussion
4.1. Why high recycling rates do not establish a closed material loop
Four metrics are needed to compare technologies: the fraction of mass sent to material
recycling, the recovery yield for each target element, product purity, and the fraction of product
that actually re-enters the intended manufacturing process. For silver, recovery yield is the mass
of silver in the final product divided by the mass of silver in the original module. Purity is the
silver mass divided by the total mass of the product. Both are percentages with different
denominators. Even silicon at 99.9% purity may fail to meet solar cell specifications because
of the impurities present. Silver recovered from powder also tells us little about losses upstream
[6], [17]. The 2024 IEA PVPS report compared some technologies only at the stage of
separating cells from glass and polymers [6]. Those data cannot establish which process
performs best across the full recycling chain.
4.2. Consequences of limited progress in technology and organization
Basic PV module recycling is already available [6], [50]. The risks examined here arise
from continued reliance on current practices, which leave some valuable materials unrecovered
and limit the quality and traceability of recovered fractions.
– Plants are likely to continue recovering mainly aluminum frames and glass. High mass-
based recovery can coexist with losses of silver and silicon [6]. Even advanced recycling
cannot quickly replace virgin feedstocks. In a growing market, new modules require more
material than retired modules supply [51].
– When recycling costs exceed recovered material value, panels may remain in storage for
extended periods. Without records of condition and intended use, it is difficult to
distinguish modules held for reuse from waste awaiting disposal. The risk depends on
storage rules and contract terms [2], [3].
– If processes are developed only after modules reach end of life, plants may be unprepared
for large volumes of retired glass–glass, multilayer, perovskite, or tandem modules [10],
[18].
13Page 14
– Improper treatment can cause losses of recoverable materials, generate fine mixed
fractions, release hazardous constituents, and contaminate process solutions [1], [6].
– Public confidence in municipal environmental programs may decline if they lack
provisions for end-of-life modules.
– Distribution grids must accommodate bidirectional flows from rooftop installations,
storage, and electric vehicles. Grid investment is lagging behind generation growth [52],
[53]. Without upgrades, some generation may be curtailed or go unused.
The risks point to gradual losses of materials and money, idle recycling capacity, and
limits on usable solar generation.
4.3. Implications for urban development
The studies reviewed do not establish an effect of solar energy on urbanization rates or
the share of the population living in cities. The main change is in how buildings interact with
the energy system. They generate electricity as well as consuming it [5 2]. This affects roof use,
grid operation, how benefits are shared among residents, and how equipment is maintained.
Rooftops serve several functions. Design must account for heritage constraints,
maintenance access, building services, vegetation, and stormwater retention [4], [40].
Preliminary results from the Mannoury project in Amsterdam showed that panels over an
irrigated blue-green roof generated about 4.5% more electricity than those over a bitumen roof
[54]. The result applies to this site. The vegetation and roof assembly provide cooling and water
retention.
In multistory buildings, roof area per unit of gross floor area falls as the number of stories
increases. With identical floor plates, the ratio is approximately 0.1 for a 10 -story building and
0.05 for a 20-story building. This geometric illustration excl udes mechanical floors and
installation constraints. In Singapore, rooftop panels primarily serve common building loads
[41]. The example provides no basis for recommending a shift to low -rise construction.
Generation and demand do not always coincide. At the district scale, this calls for load
management, storage, and shared resources [4], [52]. Even an annual electricity surplus cannot
guarantee self-sufficiency in every hour.
Residents' access to solar generation depends on ownership, roof access, and the ability
to pay for installation. Multifamily buildings and public housing therefore require specific
arrangements [4], [38].
Wider deployment creates demand for testing, repair, dismantling, sorting, and recycling
[50]. Sustaining these services depends on plant utilization, workforce skills, and markets for
recovered materials.
14Page 15
4.4. The full solar life cycle city model
The proposed full solar life cycle city model is a municipal framework for recording and
servicing solar installations and deciding in advance where the equipment will go after
dismantling. It draws on published approaches [2], [3], [6] and requires testin g in a specific
city.
Table 8. Stages of the proposed module life cycle management model
Stage Actions Verifiable outcome
1. Design and Select materials with future separation in Defined disassembly route without
procurement mind, document their composition, and reduced durability
include take-back terms in the supply
contract.
2. Commissioning Enter the installation in the municipal Traceability of each installation
registry, assign responsibility, and confirm
the supplier's participation in an EPR
scheme.
3. Operation Record defects, repairs, and replacements. Records that support decisions
Coordinate with roof maintenance. about continued use
4. Pre-decommissioning Check performance, electrical safety, and Allocation to reuse, repair, or
assessment repairability. recycling
5. Dismantling and Use qualified contractors and local Module identification and
consolidation consolidation points. condition preserved
6. Intake and sorting Identify technology, design, damage, and Separate streams for c-Si, CdTe,
hazardous components. and other technologies
7. Recycling Carry out preliminary disassembly, Products with verified
delamination, recovery, and purification at a specifications
regional facility.
8. Material recovery and Send recovered materials to users according Verified end uses and a
reporting to their specifications, manage residues, and documented route for the entire
reconcile the mass balance. batch mass
Five measures support implementation of the stages in Table 8.
– Equipment registry. The registry records the installation address, owner, module models
and quantities, commissioning date, composition, repair history, and take -back terms. It
supports forecasting and planning. A fixed retirement date 25 years after insta llation
would ignore variation in actual service life. In a simple model for a cohort of mass M
installed in the same year, the expected mass retired between ages a and b is M × [F(b) −
F(a)], where F is the cumulative probability of retirement. A numerical forecast requires
distribution parameters for different module types and ages. This review did not estimate
those parameters.
– Pre-decommissioning assessment. Higher efficiency in a new panel is insufficient to
establish the life cycle benefits of replacing a functioning installation. Removed modules
should be reused only after they pass safety and performance checks [2].
15Page 16
– Regional recycling. Cities can host collection and assessment points. Large specialized
facilities can serve a wider region, with location and capacity chosen in light of expected
waste volumes and transport requirements [3].
– Procurement requirements. Supply contracts should specify the technical documents the
supplier must provide and the method of identifying individual modules. They should
itemize the services covered, including removal, packaging, transport, and treatment.
Contracts should also define what happens if a building changes hands or the supplier
ceases trading. Reports should distinguish accepted module mass, material recycling,
energy recovery, and residues. These terms supplement EPR obligations [44].
– Digital tools and AI. Algorithms can read labels and documents. Where composition data
are missing, the registry should record the composition as “unknown.” A generative
model should not infer the missing composition. AI can support condition assessment,
but safety clearance requires a separate decision. Uncertain cases go to manual review.
Waste forecasts should report scenario ranges and may use hybrid models [49].
Conventional optimization methods are sufficient for routing. Process control should
retain independent safety interlocks.
Installed PV capacity is one measure of program performance. Table 9 also covers
operational, social, and material outcomes.
Table 9. Proposed performance indicators for a full solar life cycle city
Area Indicators
Energy Actual generation, curtailment, and self-consumption
Reliability Failures, repair time, and verified service life
Equitable access Participation of public housing and lower-income households
Reuse Number of modules that passed testing and were reinstalled
Collection Share of end-of-life equipment entering a controlled system
Recycling Recovery of individual materials, their purity, and their end uses
Residual waste Mass and verified management route
Economics Total life cycle cost and cost of meeting obligations
Evaluating AI requires comparing the same process with and without the algorithm.
Relevant measures include treatment cost per tonne, including digital system operation, as well
as target material yield and purity and consumption of energy, water, and reag ents. For
condition assessment, additional measures include the number of modules safely reinstalled,
the false clearance rate, and the share of cases requiring manual review. The false clearance rate
can be defined as the number of unsafe modules incorrec tly cleared for operation divided by
the total number of unsafe modules in the validation sample. Calculating it requires an
independent reference assessment, the sample size, and the error count. Comparisons should
use similar batches to avoid attributing differences in feedstock quality to AI.
16Page 17
4.5. Study limitations and future work
This narrative review did not attempt an exhaustive systematic search. Some industrial
data come from company disclosures without independent audits. City indicators differ in dates,
geographic boundaries, and methods of calculation. The municipal model is conceptual. Further
technology assessment requires prolonged operating trials with mixed batches, material
balances, life cycle assessment, and cost analysis [6]. The registry should be piloted in a
municipality, and digital algorithms should be evaluated against a baseline process without AI.
5. Conclusions
1. Mechanical recycling lines can recover the bulk of the material in silicon modules.
Retaining silver and returning silicon to manufacturing require dedicated delamination
and purification steps.
2. Laboratory and pilot results for laser and abrasive delamination and aqueous perovskite
recycling are promising. Stable operation across a full industrial process has yet to be
demonstrated.
3. The available statistics do not support a global material recycling rate for 2025–2026.
Collection, treated mass, material recovery, energy recovery, design capacity, and
recovered material value should be reported separately.
4. Financial assessments should account for actual plant utilization, dismantling and
transport costs, glass prices, and the cost of producing marketable silver and silicon.
Improving one operation leaves constraints elsewhere in the process.
5. Regulation assigns responsibility and determines who pays. Washington State's revised
compliance deadlines illustrate the distinction between enactment and implementation.
6. Distributed rooftop systems need consolidation of small batches, centralized programs
benefit from common contract terms, and utility-scale parks require phase-by-phase
replacement schedules. Deployment alone provides no assurance of energy self-
sufficiency or that recovered materials will return to manufacturing.
7. The most substantiated applications of AI are condition assessment, waste flow
forecasting, and separation monitoring. Economic evaluation requires comparable line
data with and without the algorithms.
8. The proposed municipal model includes module registration from procurement onward,
pre-decommissioning assessment, regional collection, contractual requirements, and
reporting. A pilot is needed to establish its practical value. Evaluation should cover
installed capacity, actual service life, waste management costs, and the end uses of
recovered materials.
17Page 18
Declarations
Author contributions (CRediT). S. Y. Shipulin: conceptualization, methodology, source
analysis, writing and editing. The author approved the final manuscript and takes responsibility
for its content.
Competing interests. The study by Coustier and colleagues was funded by the EU through the
EVERPV project (grant No. 101122208) [17]. The 2024 IEA PVPS report received support
from the EPRI Technology Innovation Program, and recyclers supplied some of the data [6].
Four coauthors of the study by Franzoni and colleagues are employees of A2A S.p.A. [49].
Data availability. This review generated no experimental data. The information used is
available in the sources listed in the references. The review covers information available
through 9 November 2026.
Ethics approval. The study involved no human participants, personal data, or animal
experiments and did not require ethics committee approval.
References
[1] U.S. Environmental Protection Agency, "End-of-life solar panels: Regulations and
management." [Online]. Available: https://www.epa.gov/hw/end-life-solar-panels-regulations-
and-management
[2] IRENA, End-of-life management for a circular economy: Solar PV panels. Abu Dhabi, UAE:
International Renewable Energy Agency, 2026. ISBN 978-92-9260-760-9. [Online]. Available:
https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2026/Jul/IRENA_POL_End-
of-life_solar_PV_2026.pdf
[3] K. Komoto, C. Agraffeil, C. Alonso-Garcia, D. Costa, et al., "Status of PV module recycling in
IEA PVPS Task 12 countries," IEA PVPS, Rep. T12-31:2025, Sep. 2025, doi:
10.69766/XLFG7020.
[4] TU Delft Urban Energy, "Amsterdam could meet nearly half its electricity needs by better
utilizing its rooftops." [Online]. Available: https://www.tudelft.nl/urbanenergy/news-
events/amsterdam-could-meet-nearly-half-its-electricity-needs-by-better-utilizing-its-rooftops
[5] AMS Institute, "Solar panel potential of Amsterdam rooftops considering local grid
infrastructure." [Online]. Available: https://www.ams-institute.org/urban-challenges/urban-
energy/solar-panel-potential-amsterdam-roof-tops-considering-local-grid-infrastructure/
18Page 19
[6] K. Wambach, C. Libby, and S. Shaw, "Advances in photovoltaic module recycling: Literature
review and update to empirical life cycle inventory data and patent review," IEA PVPS Task 12,
Rep. T12-28:2024, 2024. ISBN 978-3-907281-56-7. [Online]. Available: https://iea-
pvps.org/wp-content/uploads/2024/06/IEA-PVPS-T12-28-2024-Report-PV-Recycling-
LCI_EPRI.pdf
[7] Directive 2012/19/EU of the European Parliament and of the Council of 4 July 2012 on waste
electrical and electronic equipment (WEEE), Official Journal of the European Union, L 197,
pp. 38–71, 2012.
[8] S. Weckend, A. Wade, and G. Heath, End-of-Life Management: Solar Photovoltaic Panels.
IRENA and IEA PVPS, 2016, doi: 10.2172/1561525.
[9] M. H. Saleem, Z. Huang, Y. Duan, X. Du, X. Wang, and J. Li, "Recycling routes for end-of-life
crystalline silicon photovoltaic modules: mechanical, thermal, chemical, and environmentally
friendly strategies," J. Mater. Cycles Waste Manag., 2026 (published 5 Nov. 2026), doi:
10.1007/s10163-026-02776-0.
[10] K. Tran, N. V. D. Long, C. Nguyen, V. Hessel, K. D. P. Nigam, and P. Kwong, "Recycling of
photovoltaic modules: Technologies, comparative insights, challenges, and future outlooks,"
Waste Manag., vol. 220, Art. no. 115587, 2026, doi: 10.1016/j.wasman.2026.115587.
[11] S. Xia, Y. Yang, and J. P. H. Poon, "How to tackle the looming challenge of solar PV panel
recycling," Proc. Natl. Acad. Sci. U.S.A., vol. 122, no. 27, Art. no. e2417921122, 2025, doi:
10.1073/pnas.2417921122.
[12] K. Terzi and F. Tuillé, "National survey report of PV power applications in France 2025," IEA
PVPS Task 1, Aug. 2026, doi: 10.69766/FEAF6408.
[13] G. Deboutte, "France sees sharp rise in solar module recycling," pv magazine, 12 Aug. 2026.
[Online]. Available: https://www.pv-magazine.com/2026/08/12/france-sees-sharp-rise-in-solar-
module-recycling/
[14] First Solar, Inc., "Corporate Responsibility Report 2026," 2026, pp. 17–23. [Online]. Available:
https://www.firstsolar.com/-/media/First-Solar/Sustainability-Documents/2026_Corporate-
Responsibility_Report.pdf
[15] SOLARCYCLE, "SOLARCYCLE begins operations at Cedartown, Georgia recycling facility,"
press release, 29 Jan. 2026. [Online]. Available: https://www.solarcycle.us/press-
releases/solarcycle-begins-operations-at-cedartown-georgia-recycling-facility
[16] Solar Power World, "SOLARCYCLE recycles nearly half a million solar panels in 2024," Apr.
2025. [Online]. Available: https://www.solarpowerworldonline.com/2025/04/solarcycle-
recycles-nearly-half-a-million-solar-panels-in-2024/
[17] F. Coustier, X. Mackré-Delannoy, R. Riva, J. Aimé, R. Duwald, and Y. Karar, "Mechanical
glass-backsheet photovoltaic modules delamination: Toward materials recycling," Adv. Energy
Sustain. Res., vol. 7, no. 2, Art. no. e202500276, 2026 (published online 21 Nov. 2025), doi:
10.1002/aesr.202500276.
[18] X. Xiao, N. Xu, X. Tian, T. Zhang, B. Wang, X. Wang, Y. Xian, C. Lu, X. Ou, Y. Yan, L. Sun,
F. You, and F. Gao, "Aqueous-based recycling of perovskite photovoltaics," Nature, vol. 638,
no. 8051, pp. 670–675, 2025, doi: 10.1038/s41586-024-08408-7.
19Page 20
[19] B. N. Alsulami, J. Cameron, A. Obst, P. Prentice, P. Roghaieh, A. Feeney, and J. Kettle, "Laser-
induced delamination of aged silicon photovoltaic modules to achieve >95% selective silver
recovery," Commun. Mater., 2026 (published 3 Nov. 2026, accepted manuscript), doi:
10.1038/s43246-026-01424-2.
[20] NPC Incorporated, "Solar panel recycling machine." [Online]. Available:
https://www.npcgroup.net/eng/solarproduct/dismantling
[21] T. Dobra, D. Vollprecht, and R. Pomberger, "Thermal delamination of end-of-life crystalline
silicon photovoltaic modules," Waste Manag. Res., vol. 40, no. 1, pp. 96–103, 2022, doi:
10.1177/0734242X211038184.
[22] H. Saffarian, K. Galvin, and M. Firouzi, "A critical review of leaching pathways for silver
recovery from EoL photovoltaic modules and prospects for flotation," Miner. Process. Extr.
Metall. Rev., vol. 47, no. 6, pp. 755–769, 2026 (published online 5 Oct. 2025), doi:
10.1080/08827508.2025.2569684.
[23] Y. Zhao, R. Wang, C. Zhang, T. Zhou, and Y. Zhao, "High-value recycling of end-of-life
photovoltaic Si: study on the fabrication and lithium storage performance of waste-
silion@graphite composite anodes," J. Mater. Sci.: Mater. Electron., vol. 37, Art. no. 2366,
2026 (published 3 Nov. 2026), doi: 10.1007/s10854-026-18806-4.
[24] First Solar, Inc., "Recycling." [Online]. Available:
https://www.firstsolar.com/en/Solutions/Recycling
[25] ROSI, "ROSI raises €7.4 million in equity," 2022. [Online]. Available: https://www.rosi-
solar.com/news/rosi-raises-e7-4-million-in-equity
[26] ROSI, "ROSI RHE – a long-term industrial project in Baden-Württemberg," 2026. [Online].
Available: https://www.rosi-solar.com/news/rosi-rhe---a-long-term-industrial-project-in-baden-
wurttemberg
[27] LONGi Green Energy Technology Co., Ltd., "ESG: Environment – circular economy."
[Online]. Available: https://www.longi.com/en/sustainability/esg-environment-circular-
economy/
[28] Trinasolar, "World's first fully recycled c-Si module," Oct. 2024. [Online]. Available:
https://www.trinasolar.com/us/resources/newsroom/Worlds-First-Fully-Recycled-c-Si-Module/
[29] LONGi Green Energy Technology Co., Ltd., "First PV CYCLE certificate in Brazil." [Online].
Available: https://www.longi.com/en/news/first-certificate-pvcycle-brazil/
[30] LONGi Europe, "Carbon footprint and low-carbon manufacturing," Oct. 2026. [Online].
Available: https://eu.longi.com/esg/carbon-footprint-and-low-carbon-manufacturing
[31] Trinasolar, "Trinasolar releases 2024 Sustainability Report," 2025. [Online]. Available:
https://www.trinasolar.com/us/resources/newsroom/Trinasolar-Releases-2024-Sustainability-
Report
[32] Environment America Research & Policy Center, "Shining Cities 2022." [Online]. Available:
https://environmentamerica.org/texas/center/resources/shining-cities-1/
[33] Tokyo Metropolitan Government, "Tokyo Solar Power." [Online]. Available:
https://www.english.metro.tokyo.lg.jp/w/021-101-001016
20Page 21
[34] Tokyo Metropolitan Government, "Tokyo's mandate for the installation of solar power
generation equipment on new residential buildings." [Online]. Available:
https://www.english.metro.tokyo.lg.jp/documents/d/english/rooftops-pdf-pdf
[35] Housing & Development Board, "Status of initiatives under HDB Green Towns Programme,"
Annex, Feb. 2025. [Online]. Available: https://www.hdb.gov.sg/-/media/hdb-
pulse/news/2025/hdb-to-extend-cool-coatings-initiative-to-all-existing-hdb-estates/03022025-
Annex.pdf
[36] Rotterdam Style, "Rotterdam leads in solar panels among Dutch cities," Dec. 2024. [Online].
Available: https://rotterdamstyle.com/government/rotterdam-leads-in-solar-panels-among-
dutch-cities
[37] Dubai Electricity and Water Authority, "Mohammed bin Rashid Al Maktoum Solar Park."
[Online]. Available: https://www.dewa.gov.ae/en/about-us/strategic-initiatives/mbr-solar-park
[38] pv magazine, "Amsterdam tightens conditions around sustainable solar panel subsidy," 14 Jul.
2026. [Online]. Available: https://www.pv-magazine.com/2026/07/14/amsterdam-tightens-
conditions-around-sustainable-solar-panel-subsidy/
[39] Housing & Development Board, "Sustainability Report 2023/2024." [Online]. Available:
https://www.hdb.gov.sg/-/media/hdb-pulse/reports/annual-reports-and-financial-
statements/HDB-SR-FY23.pdf
[40] "Housing & Development Board Green Bonds: Second Party Opinion," 2025. [Online].
Available: https://www.hdb.gov.sg/-/media/hdb-pulse/reports/green-finance-framework-and-
reports/02-SPO-Report-2025.pdf
[41] Housing & Development Board, "HDB Green Towns Programme," Annex C, 2022. [Online].
Available: https://www.hdb.gov.sg/-/media/hdb-pulse/news/2022/greener-more-sustainable-
homes/HDB-Green-Towns-Programme_Annex-C.pdf
[42] IEA PVPS, "IEA PVPS Insights 2025," 2026. [Online]. Available: https://iea-pvps.org/wp-
content/uploads/2026/03/IEA-PVPS-Insights-2025.pdf
[43] IEA PVPS, "Trends in photovoltaic applications 2025," 2025. [Online]. Available: https://iea-
pvps.org/wp-content/uploads/2025/10/IEA-PVPS_Trends_2025-.pdf
[44] Directive (EU) 2024/884 of the European Parliament and of the Council of 13 March 2024
amending Directive 2012/19/EU on waste electrical and electronic equipment (WEEE),
CELEX: 32024L0884. [Online]. Available: https://eur-lex.europa.eu/eli/dir/2024/884
[45] U.S. Environmental Protection Agency, "Universal waste." [Online]. Available:
https://www.epa.gov/hw/universal-waste
[46] Washington State Department of Ecology, "Solar panels." [Online]. Available:
https://ecology.wa.gov/waste-toxics/reducing-recycling-waste/our-recycling-programs/solar-
panels
[47] Washington State Legislature, "Senate Bill Report SSB 5175," 2025. [Online]. Available:
https://lawfilesext.leg.wa.gov/biennium/2025-26/Htm/Bill%20Reports/Senate/5175-
S2%20SBR%20APS%2025.htm
21Page 22
[48] P. Dias, "The role of AI in solar circularity," SOLARCYCLE, 2026. [Online]. Available:
https://www.solarcycle.us/resources/the-role-of-ai-in-solar-circularity
[49] A. Franzoni, C. Leggerini, M. Bannò, M. Avanzini, and E. Vitto, "A multi-scale approach to
photovoltaic waste prediction: Insights from Italy's current and future installations," Solar, vol.
5, no. 3, Art. no. 32, 2025, doi: 10.3390/solar5030032.
[50] U.S. Environmental Protection Agency, "Solar panel recycling." [Online]. Available:
https://www.epa.gov/hw/solar-panel-recycling
[51] International Energy Agency, "Potential contribution of module recycling to solar PV material
demand under the Net Zero Scenario for selected materials, 2022–2050," chart. [Online].
Available: https://www.iea.org/data-and-statistics/charts/potential-contribution-of-module-
recycling-to-solar-pv-material-demand-under-the-net-zero-scenario-for-selected-materials-
2022-2050
[52] International Energy Agency, "Modernising grids in the age of electricity: The need for modern
grids." [Online]. Available: https://www.iea.org/reports/modernising-grids-in-the-age-of-
electricity/the-need-for-modern-grids
[53] International Energy Agency, "World energy investment 2025: Executive summary," 2025.
[Online]. Available: https://www.iea.org/reports/world-energy-investment-2025/executive-
summary
[54] Dutch Water Sector, "Blue-green roofs to increase energy yield of solar panels," 2023. [Online].
Available: https://www.dutchwatersector.com/news/blue-green-roofs-to-increase-energy-yield-
of-solar-panels
[55] Soren, "Rapport d’activité Soren 2025," 2026. [Online]. Available:
https://www.soren.eco/rapport-dactivite-soren-2025/
[56] Soren, "Rapport d’activité Soren 2024," 2025. [Online]. Available:
https://www.soren.eco/rapport-dactivite-soren-2024/
[57] IEA PVPS Task 12, "Advances in photovoltaic module recycling: Third update to empirical life
cycle inventory data," Apr. 2026. [Online]. Available: https://iea-pvps.org/key-topics/t12-
advances-module-recycling-3rd-edition-2026/
22Suggested Citation
Shipulin, S. Y. (2026). Photovoltaic Module Recycling in Urban Solar Energy Systems: Technologies, Economics, Regulation, and Artificial Intelligence. Smart World Journal, 1(3), Article v1i3-031. https://smartworldjournal.com/issues/vol1/issue3/articles/article31.html





















