“Productive curtailment” — the trick that lets agrivoltaic farms comply with grid export limits by helping the plants below instead of wasting power.
Here's a scene playing out across the grid industry right now. A solar developer wants to connect a new system. The utility runs its interconnection study and delivers the verdict: your local substation is at capacity. To connect, you either pay for grid upgrades — or you agree to curtail.
"Flexible interconnection agreements" are the compromise mechanism of the moment: the developer gets to connect faster and cheaper, in exchange for agreeing that the system's power exports get capped when the grid is stressed. California's "Limited Generation Profiles," which set time-varying export limits based on grid conditions, are the flagship example.
There's a catch, though. When the cap bites, a conventional solar system's inverters simply waste the energy — power is curtailed into nothing. The developer absorbs the loss. The sun keeps shining; the value evaporates.
But what if the energy didn't have to be wasted? What if, when the grid says slow down, the system responded in a way that increased its other output — the one nobody was counting?
That's the question Marcus Wu and Anna Stuhlmacher (Michigan Technological University) just answered in a new technical note, and the answer is a genuinely delightful piece of systems thinking: a solar farm over farmland can comply with export limits by turning its panels away from the sun — letting the extra light fall on the crops instead.
They call it productive curtailment. Curtailment that grows lettuce.
The Setup: A 150 kW Solar Farm That Farms
The team built their study on the Alliant Energy agrivoltaic research site in Ames, Iowa — a real, publicly documented installation: five rows of single-axis tracking panels over cropland, rated at about 150 kW, with lettuce (a shade-tolerant crop) growing below.
The framework is elegant:
- The system has a power export limit — a cap, expressed as a percentage of rated capacity, that it must never exceed (mimicking a flexible interconnection agreement).
- An optimizer solves for the time-varying panel tilt that maximizes energy production across the entire growing season — subject to that export cap.
- The optimized tilt schedule is then fed through a shading analysis and the EPIC crop model to see what happens to the lettuce.
The math trick: the raw problem (tilt angles, irradiance geometry, power) is non-convex — nasty to solve over a 60-day horizon at 10-minute resolution. By expressing tilt deviations as (cos θ, sin θ) pairs and relaxing the unit-circle constraint, the team reformulated it as a Second-Order Cone Program (SOCP) — solvable in seconds by modern optimizers.
The Results: Less Power Out, More Food Out
Here's the table that tells the whole story. As the export limit tightens from 100% down to 40% of nameplate capacity:
| Export limit | Energy produced (kWh/season) | Crop yield (t/ha) |
|---|---|---|
| 100% (no cap) | 73,501 | 1.874 |
| 90% | 72,090 | 1.885 |
| 80% | 68,534 | 1.937 |
| 70% | 63,706 | 2.017 |
| 60% | 57,583 | 2.129 |
| 50% | 50,382 | 2.276 |
| 40% | 42,257 | 2.455 |
Read that bottom row carefully. At a 40% export cap, the system loses 43% of its energy production — but the crop yield rises 31% compared to the uncapped case.
Why? Because the only mechanical way to produce less power is to tilt the panels away from the sun. And every degree tilted away from the sun is a degree of sunlight that falls on the field instead. The "curtailment" isn't wasted — it's redirected. The energy that can't enter the grid enters the food system instead.
A conventional PV system under the same agreement would have simply burned off that 43% — no food, no value, just a smaller check at the end of the year. The agrivoltaic system converts the loss into a different revenue stream.
There's a physical limit, naturally: below a 35% export cap, panel reorientation alone can't comply — diffuse and reflected irradiance keep pushing power out regardless of tilt, so the optimization becomes infeasible. At that point, you need inverter curtailment as a supplement. The paper also quantifies compliance quality: at caps above 50%, export violations are negligible (0–0.13% of the season); below 50%, they creep to 0.93–1.88% — manageable, and fully fixable by letting inverters backstop the mechanical curtailment.
One more subtle finding: the optimizer slightly beats the standard sun-tracking schedule even with no cap at all — because the classic sun-tracking algorithm optimizes direct beam only, while the SOCP accounts for diffuse and reflected irradiance. This is the familiar "backtracking" effect: flatter panels in the early morning and late evening capture more diffuse light. Free lunch, in both directions.
Why This Matters for Interconnection Due Diligence
Interconnection is the #1 bottleneck for distributed solar in the US right now — waitlists, upgrade costs, and curtailment agreements are reshaping project economics everywhere. This paper changes the math in three ways:
- Curtailment risk gets a silver lining. In an agrivoltaic project, a tight export cap is no longer a pure penalty. The developer can present it as: "we'll accept your limit — and convert the curtailed energy into crop value." That's a conversation-changer in interconnection negotiations.
- Land gets a second income stream — exactly when exports get capped. The more the grid constrains the energy output, the more the agriculture output matters. The two systems are anti-correlated by design: grid stress boosts food value. That's a hedge most assets simply don't have.
- *Flexible agreements become an agrivoltaic feature, not a bug.* Utilities offering Limited Generation Profiles can now treat agrivoltaics as the model citizen: compliant by construction, with a productive use for every curtailed watt-hour. Meanwhile, hosting-capacity-constrained feeders can admit agrivoltaic capacity that would be uneconomic as plain PV.
- The framework is practical now. Single-axis tracking is the standard configuration in utility-scale solar — no exotic hardware required. The optimizer runs in seconds, and the inputs (irradiance, sun position) are standard public data.
The authors are candid about what's next: embedding the crop model directly into the optimizer (currently it's evaluated post-hoc), realistic tilt-rate limits, forecast uncertainty, and tighter convex relaxation. The trajectory is clear — this note is a first step toward agrivoltaics as a first-class grid resource, not a research curiosity.
The Bottom Line
The idea at the heart of this paper is so clean it feels obvious in hindsight: when the grid refuses your power, hand it to the plants. The panel that can't export becomes the panel that feeds. Under a 40% export cap, the Ames system lost 43% of energy but gained 31% in crop yield — converting a curtailment penalty into agricultural revenue.
In an era where interconnection is the scarcest resource in renewable development, "productive curtailment" might be the best deal in solar: the grid gets compliance, the developer gets connection, and the lettuce gets lunch.









