Local maximum trapping in a standard P&O inverter can produce afternoon telemetry that resembles a partial string fault. The diagnostic task is separating an actual fault from MPPT behavior that follows the inverter's design. Under partial shade, algorithm responses vary widely, and inverter data can mislead without knowing which algorithm is active.
This article examines the main MPPT algorithm families, their partial shade behavior, and the patterns they produce in string and inverter telemetry available through your SCADA system.
Why Partial Shade Complicates MPPT
With uniform irradiance, a PV string's power voltage, or P-V, curve has one maximum power point. MPPT algorithms locate and follow it by changing operating voltage and measuring the power response. Standard algorithms handle this condition adequately.
Partial shade creates multiple local maxima on the P-V curve of a string with standard bypass diodes. In a 6-module string with 2 modules shaded, one peak occurs with the bypass diodes inactive, at higher voltage and lower current. A second appears with the shaded modules' bypass diodes active, at lower voltage and higher current from the unshaded modules. Either peak can be global, depending on shade depth and module characteristics.
Standard MPPT algorithms usually settle on the closest local maximum rather than the global one. Small voltage steps can lead the algorithm to a nearby peak, where it stops even though a higher peak exists elsewhere on the P-V curve. This local maximum trapping causes partial shade underperformance that resembles a fault without being one.
Perturb and Observe: Default and Partial Shade Failure
Perturb and Observe, or P&O, is widely used across inverter manufacturers. It changes operating voltage by a small step, measures the power response, continues in the same direction after a power gain, and reverses after a loss. With uniform irradiance, it converges quickly on the MPP.
Under partial shade, P&O can settle at the nearest local maximum. After convergence, small movements in either direction reduce measured power, so the algorithm has no built in way to seek the global maximum. The inverter may operate 15 to 30% below the global maximum, depending on shading fraction and bypass diode configuration.
Telemetry may then show stabilized string voltage below the expected level for the irradiance, string current below that of peer strings facing the same irradiance, and DC power below the irradiance normalized baseline. Without context on the shade and active MPPT algorithm, the pattern resembles a high resistance connection or partial string degradation. Static threshold monitoring will often flag this false positive at shaded plants.
Incremental Conductance and Partial Shade Behavior
Incremental Conductance, or InC, is the other major standard algorithm. It compares incremental conductance, dI/dV, with instantaneous conductance, I/V, rather than following the direction of power change. At the MPP, dI/dV + I/V = 0. Under stable conditions, InC generally converges faster and oscillates less at steady state than P&O.
InC shares P&O's core limitation under partial shade: it uses local gradient information and can settle at a local maximum. Its failure mode is therefore the same. Some implementations periodically sweep the full voltage range for a global scan, but that feature is added separately and is not part of the base algorithm.
Under the same partial shade, InC and P&O systems from one manufacturer may settle at different steady state voltages because their convergence paths differ. Mixed firmware versions can therefore leave two inverters with identical string configurations at different operating points. String comparisons may show this as relative underperformance, though it reflects algorithm behavior rather than a fault.
Global MPPT Algorithms: What Changes
Several inverter manufacturers now offer global MPPT algorithms to reduce local maximum trapping. These periodically sweep the full operating voltage range, identify the local maxima, and select the highest one. Examples include Fronius's Dynamic Peak Manager, SMA's OptiTrac Global Peak, and similar systems in commercial inverter lines from Huawei and Sungrow.
Global MPPT produces two characteristic telemetry patterns. During a sweep, output power briefly falls, then returns at another operating point. In 5-minute SCADA data, this appears as a short dip followed by a step change in output power. The dip typically lasts 1 to 3 minutes. Without this context, monitoring may flag a transient fault.
A successful move from a local maximum to the global maximum produces a step increase in output power. When that increase occurs at a specific time of day and aligns with partial shade onset or cloud clearing, it indicates that the algorithm is functioning as intended.
For fault detection, a plant using global MPPT will show more short power transients in SCADA data that reflect algorithm activity rather than faults. A model tuned with data from P&O plants may over alarm on global MPPT plants unless it accounts for algorithm type.
Reading MPPT Behavior in Inverter Data
Finding MPPT behavior in your data means looking for distinct patterns in inverter telemetry.
Operating voltage under partial shade: P&O and InC systems may stabilize at either the full string MPP voltage or the bypass diode MPP voltage, usually separated by 15 to 20% based on string configuration. Voltage consistently below the irradiance expectation, but not at the full string MPP, points to local maximum trapping.
Power vs. irradiance scatter: A trapped MPPT string produces less power at a given irradiance than it does under uniform irradiance. A consistent downward cluster during partial shade indicates algorithm behavior rather than a fault. Its shape will follow the shading geometry, which remains deterministic for fixed shading objects.
Recovery patterns after clouds: After clouds clear following partial shade, global MPPT systems often step up in power as they locate the global maximum. P&O systems may stay at the local maximum if the operating point did not move enough to escape it. Slow recovery after intermittent clouds can indicate standard P&O that is slow to find the global MPP again.
Calibrating Fault Detection Models for MPPT
MPPT differences mean one anomaly detection model will not perform equally across inverter types. A model trained on standard P&O plants may produce false positives on global MPPT plants because scan related power dips resemble transient faults. A global MPPT model may instead under alarm on P&O plants with local maximum trapping, treating low voltage as an algorithm state rather than an anomaly.
Effective monitoring starts by recording each inverter's MPPT algorithm type in the plant configuration, then applying detection logic suited to it. This segmentation matters especially in mixed inverter plants. Comparing a P&O inverter with a global MPPT inverter during the same partial shade event can reveal a difference caused entirely by the algorithms.
Solarvyew's detection models include algorithm aware MPPT context because this pattern appears often in pilot data from plants with mixed inverter vintages. The aim is to identify string faults, connection problems, and inverter degradation without false positives whenever global MPPT sweeps voltage or P&O remains at a local maximum on a partly cloudy afternoon.
For related context on inverter telemetry in performance analysis, read our article on what PR monitoring tells you and what it hides and our discussion of weather normalization for solar generation data.