Agentic AI for solar plant operations

Utility-scale solar loses energy in many small, distributed ways that plant-level KPIs hide. SanchAI finds and quantifies each loss, then tells you which fixes pay for themselves.

Why solar losses go unnoticed

A 100 MW solar plant can contain hundreds of thousands of modules, thousands of strings, and hundreds of inverters. No monitoring team can review that surface area daily, so in practice teams watch plant-level KPIs and investigate only when the performance ratio dips noticeably. By then, individual string faults, tracker misalignments, and soiling hotspots have often been bleeding energy for weeks. Individually each loss is small, but they add up to several percent of annual generation.

SanchAI closes that gap by reviewing everything, continuously. Its agents benchmark each string and inverter against comparable peers, then attribute each deviation to a cause and rank it by recoverable energy, drawing on 14+ years of solar PV research and field experience. Your team starts each morning with a ranked list of losses worth fixing.

Where a solar plant's energy goes

Plane-of-array potential 100%
Module temperature −3.5%
Soiling −2.5%
Inverter clipping −2.5%
Grid curtailment −2.0%
Tracker misalignment −1.5%
Shading −1.5%
Auxiliary consumption −1.0%
Net delivered 85.5%
80% 85% 90% 95% 100%

Of the 14.5% lost, 5.5% sits with the O&M team this week: soiling, trackers, and shading. The rest is design, weather, or grid.

Yesterday's generation from plane-of-array potential to net export, one step per loss mechanism. Illustrative split; the real one depends on plant and season. Amber steps are the ones the O&M team can act on.

What SanchAI does for solar plants

Loss waterfall, priced

Yesterday's generation is broken down from irradiance potential to net export: module temperature, soiling, inverter clipping, grid curtailment, tracker misalignment, shading, and auxiliary consumption, each sized in kWh and revenue. Budget versus actual on generation, irradiation, and PR is attributed to weather, equipment, operations, or grid.

String and SCB deviation

Every string-combiner box is normalised and compared to its best-performing peer on the same inverter, so a failing string, connector fault, or shading issue is isolated in hours instead of surfacing in a quarterly review, with the likely cause and the matching SOP attached.

Soiling and cleaning

SanchAI separates soiling from other loss modes, tracks its trend per block, and measures the gain after each clean, so cleaning is scheduled when the recovered energy is worth more than the cleaning cost.

Inverter underperformance

Every inverter is benchmarked against its block peers, and each deviation is tagged as hardware, DC-side, or AC-side by correlating alarms, derating events, and efficiency drift with temperature, grid, and string conditions.

Tracker fault detection

Stuck or misaligned trackers are detected from production asymmetries across the day, catching losses that look like normal cloud variation in daily summaries.

Equipment reliability

Transformer OTI and WTI trends and DC ground faults are tracked toward their thresholds before a trip, assets that fail twice in 90 days are flagged with a repair-versus-replace case, and MTBF and MTTR are benchmarked across the fleet.

Frequently asked questions

What data does SanchAI need from a solar plant?
Standard operational data: inverter and string-level measurements from your SCADA or monitoring system, meteorological data (irradiance, module temperature), and meter data. It connects over OPC UA, Modbus, and vendor APIs, and no new hardware is required.
Can SanchAI detect issues at string level, not just inverter level?
Yes, wherever string or string-combiner box monitoring exists. That is where most recoverable solar losses hide. For plants with only inverter-level data, the same peer benchmarking runs at inverter granularity.
Does SanchAI connect to our maintenance system?
Yes. It integrates with SAP Plant Maintenance for work-order status, backlog aging, PM compliance, and spares consumption, so a finding can be traced through to the ticket that closed it.
How is this different from our monitoring portal's reports?
Monitoring portals visualize data and flag threshold breaches; the investigation is left to you. SanchAI performs the investigation itself. It finds statistically significant underperformance and traces the root cause across electrical and environmental signals, then explains what it found.

Learn how the SanchAI platform works, see the definitions behind terms like performance ratio and soiling loss in the glossary, or explore solutions for wind and BESS.

Find your plant's hidden losses

Run a 2-week pilot on your solar plant's data. Full deployment takes about 4 weeks.

Start a Pilot