Understand data coverage
Map telemetry sources, update intervals and periods of missing data. Distinguish unavailable measurements from actual equipment underperformance.
Energy & Solar
Solar fleet monitoring and revenue protection AI
A solar plant rarely fails loudly. It underperforms quietly, for months, and the revenue is gone before anyone opens a ticket.
- Fleet
- 100+ MW
- Pipeline
- 1 GW
What is broken before we arrive
Two to five percent of production leaks away unseen
Industry analysis puts revenue leakage at 2% to 5% of annual production. On a 100 MW fleet that 2% is roughly €500,000 to €1,000,000 a year. It does not arrive as an outage. It arrives as a number slightly below the model, every day.
The loss has four separate causes and one dashboard
On a 500 MW portfolio the split measured across the industry runs inverter downtime 2.1%, curtailment 1.4%, pricing and settlement 1.2%, soiling and degradation 0.6%. A monitoring tool that only watches inverters is watching half the money.
Compliance is a grid-by-grid problem
PREPA imposes some of the strictest interconnection requirements in the world. A generic SCADA dashboard does not know that, and a compliance breach is not a performance number, it is a contractual one.
AI Monitoring & Revenue Protection: Puerto Rico Solar
AI monitoring deployed on live solar IPP operations: predictive maintenance across a 100+ MW fleet, with a 1 GW pipeline behind it.
Curtailment detection and revenue leakage analytics separate the losses that are the grid's from the losses that are the asset's, which is the distinction that decides whether the answer is a truck roll or a contract conversation.
BESS optimisation schedules battery storage against the same signals, and compliance monitoring runs continuously against PREPA grid requirements rather than at audit time.
How the build runs
Model expected production per asset
Without a physics-backed expectation there is no such thing as underperformance, only a number with no baseline.
Split the gap by cause
Downtime, curtailment, settlement and soiling are four different problems with four different owners. The system attributes the loss before it raises anything.
Schedule storage against the same data
BESS dispatch uses the production and price picture the monitoring already holds, rather than a second disconnected model.
Keep compliance continuous
Grid requirements are monitored the same way performance is, so a breach surfaces as an alert and not as a finding.
Implementation considerations
What to establish before automating operational alerts
Evaluate alerts against events
Compare alerts with operator-reviewed events and account for weather, curtailment and maintenance windows. Track false alarms alongside missed incidents.
Keep intervention accountable
Assign each alert a responsible operator and record its resolution. Define escalation and authorization separately from detection, especially for any control action.