With Quadrical Digital Twins. Commissioning models set your plant's production baseline — then, by layering in real-world conditions and recalibrating in real time, Quadrical builds a live, personalized benchmark down to every sensored device on your plant.
Project Finance used PVSyst to set commissioning numbers for investors. Those contracted numbers are now the O&M benchmark — and your team needs to attribute losses accurately against them for the next 25 years. Some losses stay invisible to SCADA and performance reports, yet still become expensive problems.
A string producing 3–5% below its neighbors doesn't trigger an alarm — but it compounds.
Shading at 7am lasting until 9am. Unusual soiling patterns tied to string orientation.
The gap between working and working at full capacity — measurable only with a device-specific benchmark.
An inverter getting slightly worse each month never crosses an alarm threshold — visible only across years of twins.
A battery pack degrading faster than its design baseline — no alarm, just gradual cycle-cost increase.
These losses weren't hidden. They were just invisible to the tools used to find them.
In a 9.26 MW plant audited over 3 months, Quadrical's Digital Twin AI found 620 MWh of undetected generation capacity — $126,078 of recoverable yield. This older, high-PPA plant had been continuously monitored the entire time.
A Quadrical Digital Twin is a unique model created for every physical device in your portfolio — inverter, string, battery cell node. It learns exactly how that specific device behaves, using peer behavior, age, degradation history, and operational characteristics.
A typical 50 MW solar site may run hundreds of individual twins — one per device, per job.
When an inverter starts clipping earlier than it should, a string's performance drops below its peers, or a battery's round-trip efficiency slips — Quadrical Digital Twins catch it, classify the deviation, quantify the revenue impact, and generate a ticket. Because our model always knows what that exact device's "normal" looks like.
This is the difference between just watching a number and understanding what it means. Quadrical builds this intelligence directly into your Asset Management platform — it's not something you pay for separately. Manage your whole fleet and optimize each asset. Continuously. Automatically. In real time.
No single model can catch every type of fault. Long-term degradation requires peer comparison over months. Acute faults require real-time deviation from an asset's own expected curve. Quadrical runs both simultaneously — two independent models per device, for every data interval.
Dynamic Environment – Action – Response. Trained on that asset's own data — not fleet-average shortcuts or generalized guesses.
Every raw stream passes through a quality pipeline before training. Curtailment periods, sensor dropouts, and communication glitches are identified and removed so the twin baseline stays uncontaminated.
Data ingested at 15-minute resolution, configurable to 1-second for high-sensitivity assets. Structural Twins retrain monthly, Temporal Twins weekly — with an immediate refresh triggered by maintenance events.
Two models. Every asset. Every interval.
For asset managers, gradual degradation is both commercially significant and the hardest to see. A module cluster losing a fraction of output monthly triggers no alarms — just silent, compounding losses multiplying across the portfolio.
How it catches it: by watching how one device performs relative to its peers — comparable devices at the same plant, under the same conditions. Underperformance against that group is a structural signal. The baseline recalculates monthly to absorb seasonal effects.
Sustained divergence from peers. Caught before it compounded.
Where the Structural Twin catches degradation over time, Temporal Twins catch faults in minutes. Running at 15-minute intervals, they compare device production against what it should be — based on the asset's own learned behavior and current peer performance.
If gaps sustain across consecutive intervals, an automatic Temporal Performance Ticket is issued. Extra care around consecutive-interval requirements prevents false tickets from transient shading, meter glitches, or momentary grid events.
Example: a device with a partial bypass diode failure six months ago would trip a fleet-average model every single day. Because Quadrical trains on post-failure baselines, only genuine new deviations generate a ticket — so your team can fully trust every alert.
Fault detected in the same interval it occurred.
An inverter's job is to convert DC power from strings into AC power for the grid. The right measure of whether it's doing that job is efficiency — AC output divided by DC input — not normalized power.
This matters: if an upstream string fault reduces DC input, a normalized-power model would flag the inverter as underperforming when it's actually just doing its job. By measuring efficiency, the Inverter Twin evaluates only what the inverter itself is doing — alerts are double-checked against upstream SCB deviations, so your team gets clean, correctly attributed tickets.
The right metric for the right fault.
Deviation thresholds are configurable per asset, per plant, per asset class. High-value assets may warrant tighter thresholds; assets in high-variability environments can use widened thresholds to prevent alarm fatigue. If you carry generation guarantees or PPA obligations, sensitivity can be tuned to protect your exposure. Every threshold change is version-controlled and timestamped.
Every ticket is a complete evidence chain — from raw data to its triggering deviation.
Asset ID, SCB or string reference, deviation % from peer baseline, peer group reference, historical trend chart, suggested inspection actions.
Asset ID, timestamp, deviation % from ML prediction, predicted vs. actual generation values, intraday trend chart at the moment of fault occurrence.
From deviation to actionable ticket. Every step traceable.
Quadrical doesn't replace your O&M workflow — it makes every step of it accurate.
Extensible, transferable technology to model, predict, and simulate — across Solar, Storage, and Wind.
8 months of Digital Twin–driven O&M
Loss reduction trend was continuous — each O&M cycle informed by Digital Twin tickets drove total losses lower, month over month.
Total loss trend: 8 months of continuous decline.
3-month engagement · 920 strings · 13 inverters · 33,877 panels
The audit produced a severity-ranked list of every string with data inconsistencies, long-term underperformers, and strings uncorrelated with the Digital Twin benchmark.
"Quadrical's Digital Twin approach was instrumental in setting realistic expectations and then pinpointing areas of underperformance."
Quadrical Digital Twin Analytics is not an add-on cost — it's built into the platform. From the moment you're live, every device in your portfolio is ready for its own twin, tracking deviations and generating revenue-prioritized tickets. Quadrical doesn't just monitor. It manages your whole fleet and optimizes each asset. Continuously. Automatically. In real time.
15,000+ MW of Solar, Storage, and Wind are already running on it.