Planning a Rooftop Solar Array

The year is the unit of design, not the day

Every rooftop array has twelve answers, one per month, and the annual figure on the proposal is only their sum. A system that covers 104 percent of yearly consumption can still leave a house importing heavily from November through February while spilling surplus into the grid in June. That mismatch is invisible in an annual total and it drives almost every dispute that surfaces after commissioning — the customer who was told the array 'covers the whole house' and then opens a January bill.

Before anything is drawn on the roof plan, get twelve months of metered consumption. Utility interval data or twelve bills, not an estimate. Then plot them. Heating-dominated properties peak in the months the sun is weakest; cooling-dominated properties peak when the sun is strongest and are far easier to serve. A property with a heat pump, an EV charger, or a workshop with a compressor has a load shape that changes the array conclusion more than any panel specification will.

Production follows the same calendar from the other direction. Solar resource on a fixed plane varies with declination, day length, atmospheric path length and local weather, and the ratio between best and worst month widens sharply as you move away from the equator. Near the tropics the swing may be modest; at high latitude the darkest month can deliver a small fraction of the brightest. Published irradiance datasets for the site's coordinates govern this — do not carry a number from a previous job in a different climate zone.

Reading the roof before you read the datasheet

Survey work sets the ceiling on everything downstream. Measure each roof plane's azimuth and pitch, note the ridge and hip lines, and record every penetration: vents, stacks, skylights, chimneys, satellite mounts, dormers. The usable rectangle after fire-service setbacks and obstruction clearances is usually far smaller than the gross plane area, and that shortfall is best discovered with a tape rather than at rack layout.

Shade is the item that most often gets logged loosely and punishes hardest. A shade assessment taken at one moment in one season tells you very little, because a deciduous tree that is transparent in March is opaque in July, and a neighbouring roofline that clears the array at midsummer noon will cut across it at midwinter. Take a full-year shade profile at multiple points on the plane — typically the corners and the centre of the lowest row, since the bottom of the array is shaded first. Record the result month by month so it can be applied to the production curve rather than as a single annual derate.

Structure and covering matter next. Confirm rafter or truss dimensions, spacing and span, and whether the framing can carry the added dead load plus the wind and snow loads the array introduces. Structural adequacy is determined by the governing building code and the local snow and wind criteria for the site — those criteria differ substantially between jurisdictions and are the engineer's call, not the installer's assumption. Note the covering type and its remaining service life. Mounting a twenty-five-year array on a roof with eight years left in it commits somebody to a removal and re-install, and that conversation belongs in the survey, not the handover.

Attachment detailing is where roofs leak. Flashings must be integrated into the water-shedding plane in the manner the covering demands, which is different for asphalt shingle, standing seam, tile and low-slope membrane. Standing seam allows clamped attachment with no penetration at all, and where the seam profile suits it, that is the least risky path. Tile requires either a flashing that replaces a tile or a hook detail that does not load the tile itself. Membrane roofs generally need the roofing contractor involved so the warranty survives.

Sizing the array against the twelve-month curve

Now the two curves can be put on the same axes. Take the monthly consumption from the meter data, take monthly production from the irradiance dataset applied to the surveyed plane with the month-by-month shade profile, and subtract. Positive months are export. Negative months are import. The shape of that residual is the design brief.

The array sizing calculator turns a target annual yield into a module count and a footprint, and the useful move is to run it three or four times rather than once — once for full annual offset, once for a size that fits the unshaded rectangle you actually measured, and once for whatever cap the interconnection rules or the available roof imposes. Comparing those against the residual curve shows you what each option buys in real months.

Sizing up to close a winter gap is usually a poor trade on a grid-tied system. Chasing the darkest month with array capacity means large summer surplus, and what that surplus is worth depends entirely on the local export arrangement — full retail credit, a lower export rate, or nothing at all. That arrangement is set by the utility tariff and the jurisdiction's net metering or export rules, and it should be established in writing before the array is sized, because it changes the correct answer by a wide margin.

Orientation trades sit here too. Splitting an array across east and west planes lowers annual yield relative to an equator-facing plane but flattens the daily production curve, which suits a household with morning and evening loads and a low export rate. A steeper tilt gives up some summer output and recovers some winter output, moving the curves closer together in exactly the months that were short. Neither is universally right; both are decided by the residual curve rather than by maximising a single annual number.

Running the array size

Feed the calculator the annual consumption from the bills, the site's specific yield, and a system derate that reflects what the survey found rather than a generic figure. Soiling, temperature, wiring, inverter conversion and module mismatch all sit inside that derate, and the shade profile you took should be applied per month rather than folded into a single annual percentage — otherwise a heavily shaded winter disappears into an average that looks acceptable.

Record the output as module count, module footprint and total DC capacity. That footprint has to be reconciled against the usable rectangle from the survey, with row spacing, walkways and setbacks subtracted. If it does not fit, the honest response is to revise the target rather than shave clearances that a fire official or an inspector will restore later.

Strings, voltage and the cold morning that trips the inverter

String design is an annual-extremes problem, which is why it belongs in a guide organised around the calendar. Module open-circuit voltage rises as cell temperature falls. The string that sits comfortably inside the inverter's window on a mild afternoon can exceed the maximum system voltage on a clear, still, sub-freezing morning at first light — the coldest condition the array will ever see with irradiance on it. Design the upper bound against the record low or the design low temperature for the site, taken from the recognised climatic data the local electrical code points to. Guessing that value is how an inverter gets damaged and a warranty claim gets denied.

The lower bound is the mirror image. On the hottest afternoon, cell temperature climbs well above ambient and maximum power point voltage sags. A string sized only against the cold limit can drop below the inverter's minimum MPPT voltage in high summer, and the array quietly under-produces in the months it should be strongest. Both limits have to hold simultaneously, and the window between them is what sets the permitted range of modules per string.

Conductor and overcurrent sizing follow from the same extremes: ampacity is derated for the temperature inside a conduit on a hot roof, and rooftop conduit runs get significantly hotter than the shaded ambient. Article requirements for PV source and output circuits, grounding, rapid shutdown and disconnecting means are set by the electrical code adopted in the jurisdiction — in North America that is NFPA 70 National Electrical Code as amended locally, elsewhere the national wiring rules. Confirm which edition the inspector is enforcing before the design is finalised, because PV requirements have changed materially between editions and an out-of-date detail is a failed inspection.

Checking the string against both extremes

Enter the module's open-circuit voltage, maximum power voltage and their temperature coefficients from the manufacturer's datasheet for the specific model being installed, not a similar one. Pair those with the site's design low and design high temperatures and the inverter's maximum system voltage and MPPT range.

The result is a permitted band of modules per string. Where the band is narrow, the array layout may need reworking — different string counts, a different inverter, or module-level electronics that remove the constraint entirely. Resolve that before racking is ordered, because a string that does not fit the window forces either a re-layout on the roof or a second site visit with different equipment.

Closing the gap the array cannot close

Some part of the residual curve will remain. What closes it depends on why it exists. If the gap is a daily timing mismatch — production at noon, consumption at seven in the evening — storage addresses it, and storage is sized against the evening load and the number of days of autonomy the customer actually wants, not against the annual shortfall. If the gap is seasonal, storage does almost nothing: no domestic battery bank moves energy from July into January, and selling one on that premise is a complaint waiting to happen.

Seasonal gaps are closed by the grid, by a supplementary heat source, or by accepting them. That acceptance is easier when the customer has seen the monthly chart during the sales conversation rather than discovering the shape of it a year later. Hand over the twelve-month production and consumption plot with the commissioning documents, alongside the string voltage calculations and the shade survey. It is the single document that makes the system's behaviour predictable to whoever owns the building next.

Commissioning should verify the design against measurement: string open-circuit voltages against the calculated values at the measured cell temperature, insulation resistance, polarity, and torque on every attachment and electrical connection. Log the readings. A string reading noticeably low against its siblings on a clear day is a module or connector problem that is cheap to find on the day of commissioning and expensive to find in year three, after the scaffold has gone and the roof has weathered around every flashing.