Splitting a Roof Two Ways Without Splitting the Performance: A Verified 15.64 kWp Case Study
September 10, 2026
The Engineering Challenge
A residential rooftop in Herford, Germany presented a layout problem that is more common than most feasibility spreadsheets admit. The available roof surface split naturally into two mounting planes: one facing northeast at 59 degrees azimuth and one facing southwest at 239 degrees azimuth, both pitched at 35 degrees. Neither orientation is optimal on its own, and combining them behind a single inverter raises a real question that a quick 2D layout tool cannot answer with confidence, how much energy is actually lost to the compromise, and where exactly on the roof does that loss concentrate.
The system totaled 15.64 kWp across 34 modules, split 11 modules on the northeast face and 23 modules on the southwest face, feeding a single 15 kW three-phase inverter through three separate MPP trackers, plus a 9.6 kWh lithium iron phosphate battery on a DC-coupled architecture. In our projects we consistently see this exact scenario, a homeowner or developer who wants to use the whole roof, not just the best-oriented section, and the only way to make that decision responsibly is through full 3D simulation rather than a rule-of-thumb derate.
The Approach and Methodology
We modeled the installation in PV*SOL using hourly climate data sourced from PVGIS-SARAH3 for the Herford-Stadt location, with the Hofmann model for diffuse horizontal irradiance and Hay and Davies for irradiance on the tilted plane. Both roof planes were built as separate module areas inside the same 3D scene, each carrying its own orientation, tilt, and independent shading calculation, then tied together through three MPP inputs on one Sungrow SH15T inverter, wired as 1×11, 1×11, and 1×12 strings.
The critical step was not the string design itself but the shading disaggregation. Rather than applying one blanket shading factor across the whole 67.9 square meter array, we simulated obstruction on a per-module basis. The 3D shading render showed the vast majority of modules losing less than 0.3 percent of annual irradiance to nearby obstructions, but one specific table position on the southwest face showed a 5.7 percent loss, driven by a rooftop vent pipe sitting close to that group of panels. Isolating that number meant the rest of the array’s performance was not falsely dragged down or falsely inflated by an average.
This is where PV*SOL design work for EPCs and developers earns its place in the workflow. As part of our PV*SOL design, PV planning and solar simulation services, we routinely rebuild the roof geometry down to individual obstructions, chimneys, vent stacks, satellite dishes, before assigning any shading loss, because a single averaged percentage across a split-orientation roof can either overstate or understate the real production of specific strings.
The Results
The completed simulation returned a specific yield of 837.57 kWh per kWp per year, a system-wide performance ratio of 88.20 percent, and an overall shading-related yield reduction of just 3.8 percent, despite the presence of that one 5.7 percent hotspot. Annual PV generator energy delivered to the AC side, including battery interaction, came to 12,731 kWh. Of that, 3,210 kWh was consumed directly on site and 9,521 kWh was exported to the grid, for a self-consumption share of 25.0 percent and a self-sufficiency rate of 79.5 percent against a 4,038 kWh total household demand including inverter standby losses.
The full energy balance breakdown confirms how the losses stack. Orientation and tilt losses versus an ideal plane accounted for 10.86 percent of the theoretical resource, a number driven largely by the northeast-facing string. STC conversion, the fixed physical limit of module efficiency, took the largest single share at 76.97 percent, as expected. Module-specific partial shading losses, the figure directly tied to that vent-pipe hotspot and other local obstructions, came to 1.88 percent of the post-STC energy, a figure our simulation could isolate precisely because each roof plane and each string position was modeled independently rather than averaged.
Why Lion Solar is Better Than Reonic and Solarmonkey for Complex Roof Geometries
Quick-proposal platforms such as Reonic or Solarmonkey are built for speed on straightforward, single-orientation roofs, and for that use case they perform well. Their workflows are optimized around fast 2D sales proposals: a satellite image, a rough tilt and azimuth assumption, and a blanket derate for shading and soiling. That approach breaks down precisely on projects like this one, where two roof planes with different orientations feed a shared inverter and a shared battery, and where a single rooftop obstruction affects one string disproportionately.
A 2D tool applying one shading percentage across the entire 67.9 square meter generator would either overstate losses on the well-exposed southwest strings or understate the real 5.7 percent hit taken by the shaded table near the vent pipe. Neither outcome is acceptable when a client is deciding how to size the inverter’s three MPP trackers or whether the northeast face is worth installing at all. Full 3D hourly simulation, run at the individual module and string level, is where engineering-grade design consistently outperforms quick sales tools. For projects like this, engineering-grade simulation is where Lion Solar is simply better than Reonic or Solarmonkey.
What This Means for EPCs and Investors
For EPC contractors quoting split-orientation roofs, the lesson is direct, a single average shading or orientation derate is not sufficient once more than one roof plane feeds a shared inverter. The per-string and per-module simulation approach we used here is the same one we apply across our PV*SOL design, PV planning and solar simulation services for EPCs and developers, because it protects the contractor from underperformance disputes and gives the investor a production forecast that reflects the real roof, not a simplified stand-in for it. With an 88.20 percent performance ratio and a 79.5 percent self-sufficiency rate confirmed against hourly climate data, this project shows that dual-orientation roofs, modeled correctly, do not have to mean compromised performance.
