Validation
Benchmark your solar forecasts.
Solar forecast performance changes with location, season, plant behaviour, aggregation, lead time and operational events. GraphSight uses a documented evaluation method and avoids universal accuracy claims that remove this context.
Proof model
How GraphSight builds evidence.
- 01
Product facts
Resolution, horizons, scale, coverage and delivery specifications.
- 02
Methodology
Definitions, issue-time alignment, metrics, quality control and exclusions.
- 03
Event studies
Qualitative comparisons during curtailment, snow, ramps or changing cloud.
- 04
Public zonal comparison
GraphSight, actual generation and a public TSO reference.
- 05
Private customer benchmark
Asset and portfolio evaluation on customer-provided data.
Benchmark methodology
Agree the rules before looking at the result.
A credible comparison fixes the forecast issue time, target horizon, actual source and quality rules before calculating the score.
Event studies
Look closely at the days that expose the model.
Aggregate scores can hide operating conditions that matter most. Event studies show how forecast vintages respond during curtailment, snow, ramps and fast-changing weather.
- What happened and why was it difficult?
- What was known at each issue time?
- How did potential and operations-aware forecasts differ?
- What limitation remains?
Potential and delivered generation
A known curtailment event changes expected operational output while weather-driven solar potential remains high.
Put GraphSight beside the forecast you use today.
Tell us the assets, portfolio or zone you need to forecast. We will review the available data and propose a fair evaluation scope.