Literature review: Agrivoltaics adoption

▼Western University's Free Appropriate Sustainability Technology (FAST) Research Group

Wanted! Students to make a distributed future with solar-powered open-source 3-D printing and recycling. Contact Dr. Joshua Pearce - Apply here

Home · Projects & Publications · Methods · Literature reviews · People · Sponsors · News · Free Software · YouTube · LinkedIn · Github · For kids!

▼This page is a literature review.

Feel free to edit this page and add sources and summaries.

The "What" of the topic.

The "How" of the topic.

The "Why" of the topic.

The "When" of the topic. Review current state with an emphasis on the development of the field over time.

The "Who" of the topic.

The "Where" of the Topic

• Coined the term ‘agrivoltaics’ to describe the combination of solar panels and food crops on the same land.
• Compared conventional options (agriculture-only and energy-generation-only) with two agrivoltaics systems at different solar panel densities using Land Equivalent Ratio (LER).
  • LER = (Y_cropinAV / Y_monocrop) + (Y_electrictyAV / Y_{electricityPV}) ; monocrop refers to agriculture-only, PV refers to conventional/stand-alone PV plant, AV refers to agrivoltaic system
  • If LER > 1, AV is an effective use of land than agriculture-only and energy-generation-only
• Models show that agrivoltaic systems may increase overall land productivity compared with using the same land for agriculture-only and energy-generation-only separately.

• Identified Ontario Canada policies that limit agrivoltaics development:
  • Provincial Policy Statement (PPS) (2020) regulates land use and development in Ontario; Note: The Provincial Planning Statement (PPS) 2024 replaced the Provincial Policy Statement (PPS) 2020.
  • Growth Plan for the Greater Golden Horseshoe (2020)
  • Greenbelt Plan (2017)
• Agrivoltaics development on farmland meets the criteria for permitted uses in prime agricultural areas specified in PPS 2020 (now PPS 2024). Table 1 and Table 2 below is copied from paper for ease of reference:

• Author argues that provincial and municipal policies must be aligned to increase agrivoltaic system deployment in Ontario; policy language must be updated to account agrivoltaics which is a dual-use system.
• Author recommends that provincial energy policy could incentivize agrivoltaics.

• While the paper demonstrated that agrivoltaics can meet Canada’s energy demands using only 1% of agricultural land, the authors suggest the need to understand the effect of having a large number of solar farms (through agrivoltaics) affect the power grid.

• Participants expressed that long-term land viability (land quality) is important in their work and should not be compromised.
• Participants also proposed that having land-use agreement or long-term plan relieves them from concerns about the future of their farms and reduce uncertainties in land-use planning.
• Economic concerns appear second to technical considerations. Participants want to be assured of financial gain.
• Participants expressed the importance of a flexible agrivoltaic system that can adapt to changing market conditions (which looks like: “If we’re looking at a 25-year kind of investment with the solar panels and when you’re talking about integrating them within the livestock species too, that market for livestock might look totally different within 10 years. So, implementing some flexibility there that if we’re not going to run rabbits, maybe we’re running something else in there in 20 years. But having some flexibility in the system that you could respond to the livestock markets in there as well, I think is important.”).
• Farmers also looked for just compensation and tangible benefits should they particiapte in agrivoltaic projects.
• Their likelihood of adoption is also based on whether agrivoltaics could benefit them and their operation—compatible and easily integrated with their current practice.
• This paper identified 3 barriers and their corresponding opportunities and future work, as presented in Table 3 below (copied direct from the paper):

• Findings suggest that the "current state of the diffusion of agrivoltaics is advancing towards wider implementation and has surpassed initial phases of information gathering and persuasion."
• The study participants were early majority adopters.

• The authors defined social acceptance in the context of renewable energy innovation as the triangle of socio-political acceptance (of technologies and policies by the public, key stakeholders, and policy makers), community acceptance (procedural justice, distributional justice, and trust), and market acceptance (consumers, investors, intra-firm). A lack of acceptance in these areas may hamper the successful innovation implementation.
• The popularity of an innovation does not necessarily translate to innovation acceptance, like how people may support something (say, agrivoltaics) but oppose it when confronted to set it up in their backyard (The Not-In-My-Back-Yard, NIMBY effect).
• "Trust is a key issue in all facility siting issues."
• For micro-generation technologies to get socially accepted, they must be accepted by the homeowners expressed in attitudes, behavior, and investment.

• How a farmer perceives agrivoltaics influences the farmer’s intention to adopt agrivoltaics.
• Figure 1 summarizes the main themes the authors identified in their analysis.
• There are mixed perceptions about agrivoltaics. While it is seen as a solution to the land-use conflict and contribute to a sustainable energy transition, agrivolatics was also described as complicated, expensive, visually unacceptable, and may increase conflict of interest among stakeholders.
• The authors also conceptualized a model called “the Pentagon tower of social innovation diffusion” presented in Figure 2 which combined the "two theoretical lenses of innovation diffusion theory (horizontal) and different social acceptance levels (vertical)."

• The nationawide quantitative survey revealed that 85.8% of Canada supports agrivoltaics. The authors identified that stilt-mounted agrivoltaics is the most favored configuration, followed by greenhouse agrivoltaics, solar grazing, vertical or single axis, and agrivoltaics over trees and perennial crops.

• Section 4.3.2.1 states that in prime agricultural areas, only three kinds of land use are allowed: agricultural uses, agriculture-related uses, and on-farm diversified uses (OFDUs) which should be “based on provincial guidance.”
• OFDUs “shall be compatible with, and shall not hinder, surrounding agricultural operations.” Municipalities can set the criteria using provincal guidance or their own approach. Agrivoltaics arguably meets this more easily than conventional solar (italics inferred by JIQuesada).
• In Chapter 8 (Definitions), the OFDU definition now includes “electricity generation facilities and transmission systems, and energy storage systems."
• OFDUs has to be "secondary to the principal agricultural use” and located on a farm property actively in agricultural use and must be limited in area.
• In Section 4.3.5.2, impacts on the whole agricultural system must be avoided or else minimized and mitigated through an agricultural impact assessment. An agrivoltaic project could use this assessment to show it keeps land productive (italics inferred by JIQuesada).
• Some supportive energy and climate policies:
  • Section 3.8.1: planning authorities “should provide opportunities” for renewable energy systems, energy storage, and electricity generation and transmission to accommodate current and projected needs.
  • Section 2.9: planning should cut greenhouse gas emissions and build resilience to climate change. This could be used to argue for agrivoltaics’ climate-adaptation benefits (italics inferred by JIQuesada).
  • Section 2.5.1 (e) and (h): rural areas should diversify their economic base, including economic activities in prime agricultural areas under Section 4.3.
  • Section 2.6.1: on rural lands (not prime), OFDUs and “other rural land uses” are allowed which gives agrivoltaics much more room (italics inferred by JIQuesada).

• OMAFRA Publication 851 is the “provincial guidance” the PPS 2024 refers to.
• In it PPS Criteria for On-Farm Diversified Uses (Section 2.3.1.3), OFDUs are limited to 2% of the property, to a maximum of 1 hectare (10,000 m²). (A 1ha of land roughly fits 0.5–1 MW of fixed-tilt PV and since agrivoltaic layouts are spaced more widely, the estimate could be lower).
• Publication 851 dates from 2016 and at that time, solar was approved under the Green Energy Act, 2009. However, the Green Energy Act was repealed in 2019 so solar now needs municipal approval under the Planning Act.
• Publication 851 already excludes “land producing harvestable crops” from the area count for agri-tourism (Section 2.3.1.3). And by analogy, it can be argued that the cropland or pasture under agrivoltaic panels should not count toward the 2% or 1 ha cap. Only the parts that take land out of production would count such as posts, inverters, and access lanes (italics inferred by JIQuesada).

• In Section 2.1(d)(ii), where a project is a “Non-Rooftop Solar Project,” the project site “may not include any lands located in areas designated as Prime Agricultural Areas.” Agrivoltaics may not be mentioned specifically but since agrivoltaic systems (elevated, vertical, or grazing-based), it fall under this ban regardless of whether the land stays in agricultural production (italics inferred by JIQuesada).
• Section 2.1(d)(i) states that no project of any type may be located in Specialty Crop Areas.
• In Section 2.1(d)(iii), other energy projects like wind and storage may be located in Prime Agricultural Areas if the Local Municipality confirms that an Agricultural Impact Assessment (AIA) Component One was completed to “the satisfaction of the Local Municipality,” with Components Two and Three completed after selection. However, solar projects have no equivalent pathway (italics inferred by JIQuesada).
  • Need to read more about what Components One, Two, and Three are.
• This document is a draft for discussion (August 14, 2026).

• The paper asks: When a group could settle on two different normal ways of doing things (two Nash equilibria), which one does it actually end up with?
• How agents behave in the model: when group plays the same game over and over:
  • Not everyone changes strategy every period (inertia hypothesis), because switching is costly and there are also uncertainties.
  • When agents change, they choose what works best right now (myopia), justified by a substantial presence of inertia because they know that only a small fraction of the population changes their behavior at any given point in time. They do not consider the long run impact of their choices and they don’t plan ahead.
  • Occasionally, someone switches at random (mutation), for example a newcomer who knows nothing about the game.
• Without mutations, the group gets stuck wherever it started but with mutations, the group occasionally gets knocked from one equilibrium to the other. And so over a long period of time, where does the group spend most if its time? … which the authors call the long-run equilibrium.
• (example on technology choice, p.32): People switch technology only when their old equipment needs replacing because new equipment is expensive.
• (another example, pp.34-35): Ten students each use one of the two computer systems. They can only work together if they use the same one, and one system is better.

• • If 4 or more students use the better system, everyone gradually switches to it. If fewer do, everyone drifts to using the worse system.
  • The situations split into two “basins of attraction,” each being the set of starting points that all end up at the same outcome.
  • Escaping the ‘worst system’ basin takes 4 students switching by accident at the same time. Escaping the ‘better system’ basin takes 7 students.
  • And because 7 random switches (mutation) at once is much rarer than 4, the group spends most of its time on the 'better system.’ And if the mutation rate is 10%, the group leaves the 'worse system' after about 78 periods on average, but takes 100,000 periods before the group leaves the ‘better system’.
• Main finding: In the long run, the group settles on the option with the bigger basin of attraction because it takes the most random switches / mutations to know people out of it (long-run equilibrium).
  • For large populations, the equilibrium with the larger basin is the risk-dominant equilibrium (Harsanyi & Selten, 1988): the option that pays more if a player assumes others are equally likely to choose either option (Corollary 1, p.46). This need not be the option that pays best when everyone coordinates (the Pareto-dominant equilibrium). When both options pay the same under miscoordination (equal security levels), the Pareto-dominant equilibrium is selected (Corollary 2, p.46).
• Random switches can take a very long time. In large groups, enough randowm switches may practically never happen. The authors used QWERTY keyboards as the example (p.36). The result is most realistic for small groups.
• And with two separate groups that adjust at different speeds, the details of learning start to matter again (Section 9).
• In coordination games with a single population, the long-run equilibrium depends only on the payoffs (through the relative sizes of the basins of attraction), and not on the details of how agents learn or how fast they adjust, as long as better-performing strategies gain users (Section 8, p.51). The result does not extend to two separate populations adjusting at different speeds (Section 9).

• A convention is an equilibrium that everyine expects: everyone conforms, expects others to conform and wants to conform if others do (p.57). It is a pattern of behavior that is customary, expected, and self-reinforcing (p.57).
• Examples from p.57: Driving on the right. Lunch at noon. Accepting dollar bills. Conventions can be asymmetic: sailboats on one tack yield to the other, and in some regions, tenant farmers get 1/3 of the harvest while landlords get 2/3 (or reverse elswhere).
• The paper asks: When a game has several equilibria, how does one become the convention? Which one survives in the long run? (p.58)
• How agents behave in the model (adaptive play) pp.59-62:
  • Each role in the game is filled from its own population; one person is drawn from each period (p.61)
  • Society remembers only the last m plays (memory). Each player learns about only k of them (sample), by “asking around” (p.61)
  • Players best-respond to their sample; occasionally, they experiment or make mistakes (pp.66-67)
  • Agents do not learn individually; each plays once and is replaced by a newcomer (p.59). Society can learn even when its members do not (p.77)
• Main finding #1 – conventions form (Theorem 1, p.64): If nobody makes mistakes, the group eventually settles on one way of doing things and sticks with it for good. Young calls this a convention (p.62). Two conditions are needed to arrive into a convention. (1) The payoffs must not send people chasing each other in circles. From any starting point, people switching one at a time to their best option must eventually reach an outcome that everyone is happy with. Young calls it a weakly acyclic game. (2) Each person must hear about only a small part of what the group remembers. In a two-option game, what a person hears about can be at most 1/3 of the remembered history (k≤ m/3, p.70).
  • Incomplete sampling breaks miscoordination. If there is full information (k = m), players can miscoordinate forever (Example 2, pp.65-66)
  • Games with built-in best response cycles may never settle (Example 1, pp.62-63)
• Main Finding #2 – stochastic stability (Theorem 2, p.69): Suppose people occasionally slip up and do something unusual. If enough slip-ups happen close together, the whole group can be knocked from one convention into another. Over a long time, the group spends almost all its time in the convention that is easiest to drift into from all the others. Young measures this by counting slip-ups (resistance): add up the fewest slip-ups needed to bring the group into a convention from every other one, and the convention with the smallest total (stochastic potential) wins. The winner is stochastically stable.
  • The result does not depend on how often each population makes mistakes or which mistakes they make (pp.67, 69)
  • How to compute (p.70): (1) list the conventions; (2) find the path of least resistance from every convention to every other; (3) find the tree of least resistance for each convention.
• In 2x2 games (Theorem 3, p.72): The selected convention is the (weakly) risk-dominant equilibrium, matching Kandori, Mailath, & Rob (1993) by a different process. This needs large enough k and m/k because small samples cannot tell close payoffs apart.
• In 3x3 games (Section 8, pp.73-75): The selected convention can be neither risk-dominant nor Pareto-optimal.

• Text

• Text

• Text

• Text

• Text

• Text

• Text

• Text

• Text

Zotero citation field (using “Cite -> Basic) with the URL (DOI preferred). Use APA Style 7th Edition.

• Each top-level point should be a clear and concise key item from the source (methodology, info, design, gap, etc.)
  • Sub points are to be concise explanations of critical aspects of the key item
  • Should not be a copy and paste of info but rather an interpretation of what parts are relevant and why, selective copy & paste of relevant snippets is fine.

Zotero citation field (using “Cite -> Basic) with the URL (DOI preferred). Use APA Style 7th Edition.

• Each top-level point should be a clear and concise key item from the source (methodology, info, design, gap, etc.)
  • Sub points are to be concise explanations of critical aspects of the key item
  • Should not be a copy and paste of info but rather an interpretation of what parts are relevant and why, selective copy & paste of relevant snippets is fine.

• Text

• Text

• Text

• Text

• Text