Moon Power Grid: Scientists Optimize a Megawatt Solar Farm for the Lunar South Pole

Engineers have long dreamed of powering a permanent human presence on the Moon, but building a reliable electrical grid on another world is one of the hardest design problems in space exploration. Now, a pair of researchers at Shell Global Solutions has taken a major step toward that goal with a detailed design optimization of a megawatt-scale solar array farm near the lunar south pole. Their study, published in the journal Space and Planetary Resources, demonstrates that a carefully arranged field of vertically oriented, sun-tracking solar panels could sustainably deliver more than one megawatt of average power on the Connecting Ridge near Shackleton Crater, exceeding NASA’s own estimate of the power needed to support lunar missions over the coming decade.

The technology at the heart of the study is the Vertical Solar Array Technology, or VSAT, a concept NASA has been advancing for lunar polar missions. Unlike conventional flat panels, a VSAT features an elevated, vertically oriented photovoltaic panel that rotates to always face the Sun. This design is ideally suited to the lunar poles, where the Sun never rises high into the sky. Because the Sun skims just a few degrees above the horizon, a vertical rotating panel can receive sunlight at nearly normal incidence, maximizing the solar energy it captures. A single large VSAT measuring roughly fifty square meters, with a solar conversion efficiency of thirty percent, can produce about fifteen kilowatts of power in full sunlight. That is impressive for one unit, but early lunar activities will quickly outgrow such bespoke power sources, so the real challenge is combining hundreds of panels into a farm capable of generating power in the megawatt range.

That is where the geometry turns hostile. The same low Sun angle that makes vertical panels so effective at the poles also creates a serious problem: shadows. Because the Sun sits near the horizon, every VSAT casts an enormous shadow directly behind it, and a dense field of panels can shade one another severely, slashing the total energy output. On a hillside where the terrain itself can also block the Sun, self-shading between panels becomes a critical design constraint. Previous studies, including work by Ross and colleagues, had shown that lunar polar hills offer near-constant solar visibility but suffer abrupt shading from nearby structures, and that a megawatt farm would require hundreds of panels whose interactions had to be modeled carefully.

Earlier illumination studies of the lunar poles, such as those by Noda, Speyerer, Gläser, Mazarico and their colleagues, modeled the local horizon as a slowly changing elevation function around a site. These approaches revealed that a ridgeline on the Connecting Ridge beside Shackleton Crater could receive over eighty percent illumination across an eighteen-year precessional cycle. But they were not designed to handle the abrupt, discrete shadows cast by neighboring solar panels, and they were computationally far too expensive to evaluate the thousands of possible panel arrangements needed to optimize a farm of hundreds of units.

The new study introduces two novel ideas to break that computational bottleneck. The first is a binary sky bitmask technique, borrowed from terrestrial solar energy estimation. On Earth, sky imagers use bitmasks to capture clouds and scattered light, which vary continuously in brightness. The airless Moon is far simpler: sunlight is either blocked or not. The researchers built their terrain model from a cascade of digital elevation models, including a high-resolution five-meter DEM for the Connecting Ridge and coarser regional models from the Lunar Orbiter Laser Altimeter. For each candidate panel location, the narrow corridor of sky within a few degrees of the horizon where the Sun can appear is mapped onto a rectangular bitmask with an angular grid resolution of 0.05 degrees, fine enough that the solar disk spans ten pixels and eighty active bits. Computing the insolation at any moment then reduces to a fast bitwise AND operation between the sun bitmask and the sky bitmask, with bits packed into sixteen-bit integers for speed. The team validated the method by comparing it against a traditional ray-tracing calculation, finding the two power curves for a hundred-panel farm almost perfectly overlapped, and by checking their terrain registration against actual photographs from the Lunar Reconnaissance Orbiter.

With the fast computation in hand, the researchers turned to the design problem. They considered a grid of 6,166 candidate VSAT locations spaced twenty meters apart above the 1,900-meter elevation contour on the Connecting Ridge. Each panel was modeled as five meters wide and ten meters tall, with its bottom edge five meters above the surface, delivering up to fifteen kilowatts in full sun. A greedy optimization algorithm then built the farm panel by panel: at each step, the candidate location that added the most total energy to the farm, after accounting for the shading it cast on all previously placed panels and received from them, was permanently selected. Starting from the single best-lit spot on the ridge, the algorithm generated an ordered sequence of panel positions.

The first hundred panels in the energy-maximizing configuration produced a total of 10,695 megawatt-hours during the year 2030, an average of 1.22 megawatts, comfortably above NASA’s one-megawatt requirement. But the design had a hidden flaw. Twice per lunar day, the Sun’s direction aligned with the long diagonal of the panel layout, and most panels shaded one another simultaneously, causing dramatic dips in output known as brownouts. The energy-optimized arrangement had naturally exploited alignments that concentrated shading at times when other panels were already well lit, but these transient power troughs posed a real problem for a grid that must supply life-support systems without interruption.

The second novel idea addressed the brownouts directly. Instead of maximizing total integrated energy, the researchers optimized a logarithmic value function that rewards marginal power most when the farm is generating little of it. In effect, an extra kilowatt is worth far more during a power trough than during peak production, so the algorithm steers panel placement toward times and places that fill in the gaps. Empirically, the result proved insensitive to the scaling constant chosen for the value function. The reoptimized hundred-panel farm lost only two percent of its total annual output, delivering 10,522 megawatt-hours or an average of roughly 1.2 megawatts, yet the brownout troughs vanished entirely, leaving only the unavoidable blackouts that occur when the entire hill is plunged into darkness by the shadow of Mons Mouton during part of the year.

Those terrain-driven blackouts are the one problem no arrangement of solar panels can solve, and the study is explicit about the implication: survival power during dark periods must come from stored energy or a nuclear source. The researchers also quantified the diminishing returns of farm expansion. While the first panel contributes an average of 13.7 kilowatts, the hundredth adds only 10.8 kilowatts, a 21 percent reduction, and the five-hundredth adds just 7.7 kilowatts, down 44 percent, as self-shading and less favorable sites take their toll. Even so, the Connecting Ridge could sustain an average of 4.8 megawatts with five hundred panels, and the analysis indicates the chosen boundary comfortably accommodated the first five hundred placements, though larger farms would need more space. The economics of panel deployment versus marginal benefit will ultimately determine the optimal size of such an installation.

The study opens several avenues for future work. Real deployments will face slope constraints and other siting restrictions, which can be incorporated by simply removing candidate locations from the grid. More intriguingly, the authors suggest that if individual VSAT panels could move by meters over hours, coordinated motion could further suppress self-shading, a far harder optimization problem that might one day be tamed by deep learning or autonomous distributed computation, or even solved by a simple emergent behavior at each station. For now, the work provides a practical blueprint: by combining a fast bitmask illumination model with value-based optimization, lunar planners can design solar farms that grow gracefully from a handful of panels into a megawatt-class power grid, bringing permanent human habitation at the Moon’s south pole a significant step closer to reality.

The bitmask approach also carries practical implications for how lunar power infrastructure might be planned and expanded over time. Because pre-existing installations can simply be added to the initial sky bitmask at every candidate location, the method naturally accommodates incremental growth: a farm that begins with a handful of bespoke panels can be extended later without redesigning the whole site. This aligns with the staged reality of lunar exploration, in which early missions arrive with modest power needs and only later demand megawatt-scale generation as surface activity intensifies.

The study site itself was chosen for good reason. The Connecting Ridge, on the rim of Shackleton Crater near the south pole, is considered a strong candidate for high-activity landing zones, partly because the hypothesized presence of water ice and other frozen volatiles in permanently shadowed regions nearby makes the area attractive for resource utilization. The Sun’s path there is confined to a narrow swath only a few degrees above the horizon, shifting between solstice extremes and bounded by a margin of about two degrees to ensure no part of the solar disk escapes the modeled corridor from any vantage point in the study area.

There is also a strategic dimension to the work. By estimating how much power a single, well-chosen lunar hill can deliver before self-shading becomes excessive, the analysis offers long-term guidance on the eventual mix of solar and nuclear power that sustained lunar presence will require. If the best hilltops can supply several megawatts but no more, planners can determine at what point stored energy or fission systems must take over, informing architecture decisions years before hardware is deployed.

Subject of Research: Design optimization of a megawatt-scale vertical solar array farm for power generation at the lunar south pole

Article Title: Design optimization of a megawatt scale lunar VSAT solar array farm

Article References: Butler, W. H., & Freudenreich, Y. P. (2026). Design optimization of a megawatt scale lunar VSAT solar array farm. Space and Planetary Resources, 2(1), Article 4. https://doi.org/10.1007/s44461-026-00009-8

Image Credits: AI Generated

DOI: 10.1007/s44461-026-00009-8

Keywords: lunar south pole, VSAT, solar array farm, self-shading, sky bitmask, illumination modeling, Shackleton Crater, Connecting Ridge, lunar power systems, photovoltaics, space exploration, megawatt power