- Calibsun | NEXT
PRÓXIMO
Como funciona
- Rigorous data quality control before any forecast is produced
- NWP + satellite imagery + real-time on site data, dynamically weighted by time horizon
- Deterministic and probabilistic outputs
From data to accurate intraday and day-ahead forecasts
NEXT by CalibSun delivers the predictive solar power that PV operators, traders, aggregators, and grid managers need: dynamic intraday and day-ahead forecasts continuously updated with satellite data at 15-minute intervals, NWP models updated every 6 hours, and on-site irradiance and production data in real time.
This multi-source, multi-horizon forecasting method provides the accuracy required to optimize trading strategies, reduce imbalance costs, support storage, and generate measurable revenue, whatever the climate or grid regime. Data confidentiality is guaranteed throughout. Contact our team to explore what NEXT can deliver for your business.
- PRÓXIMO
Previsão process
We process your data with care
From understanding your needs to providing forecasts, our team of experts makes the most of your energy data to offer you forecasts of unrivalled accuracy, tailored to your scenarios.
Sensor Quality Check & data validation
On an operational PV plant, corrupted sensor data rarely looks corrupted. A pyranometer affected by soiling, frost, or misalignment continues to deliver measurements with valid timestamps and plausible values. But is silently distorting every downstream process: forecasting models, satellite calibration, performance ratio calculations.
CalibSun’s Sensor Quality Check is the first stage of NEXT’s pipeline, applied systematically to every dataset before any forecast model is trained or recalibrated. It is one of the core differentiators of our technology, because data quality issues at sensor level have cascading effects on solar irradiance forecasting accuracy, generation estimates, and grid planning reliability.
When an issue is detected, CalibSun’s expert team provides direct recommendations to help operators resolve it, protecting both data integrity and long-term forecast performance.
Integração de dados de múltiplas fontes
NEXT integrates continuously:
- Numerical weather prediction (NWP) (ECMWF, GFS, AROME, ICON, ARPEGE), refreshed every 6 hours, structuring day-ahead forecast and capturing large-scale weather conditions
- Geostationary satellite imagery (MSG, IODC, GOES, Himawari), updated every 15 minutes + lag for intraday input
- On-site SCADA & sensor data : global horizontal irradiation (GHI), temperature, humidity and tailored inputs integrated in real time at up to 1-minute resolution
The algorithm dynamically reweights each source by forecast horizon and meteorological context, prioritizing satellite imagery and on site data for short-term intraday prediction, shifting to NWP for day-ahead planning.
Site-specific model calibration
Every solar power plant has a unique relationship between incoming solar irradiance and actual power output, depending on panel technology, layout, local soiling patterns, or grid constraints.
At onboarding, NEXT trains on your historical data to learn your plant’s irradiance-to-generation behaviour. The machine learning model then recalibrates continuously as new measurements flow in to integrate local weather conditions.
This continuous adaptation, driven by statistical forecasting methods, ensures forecast accuracy increases over years of operation.
Forecast delivery & operational outputs
Full forecast uncertainty envelope in 5% quantile increments, natively calibrated on your plant’s observed variability, not modeled from parametric assumptions. Designed for risk-aware trading, storage dispatch, and decision-making under imbalance penalties.
Delivery:
- API segura (push or pull, JSON/CSV) — integrates directly into your trading platform, EMS, or SCADA
- Web dashboard — intraday, day-ahead, and historical views, with real-time vs. actual tracking, accuracy KPIs, and tailored inputs
- Accuracy follow-up: RMSE, MAE, skill score by forecast horizon and by site for reliable follow-up of your solar forecast
- PRÓXIMO
Proven accuracy
Proven results, consistently validated: our approach delivers measurable accuracy gains that translates directly into financial gains.
- PRÓXIMO
Proven accuracy
Proven results, consistently validated: our approach delivers measurable accuracy gains that translates directly into financial gains.
Operational Reliability & Data Protection
No hardware installation
NEXT connects to your existing sensors and SCADA via secure API. No on-site equipment is needed. Integration is handled entirely by CalibSun’s technical team.
Forecast continuity when data is interrupted
When on-site measurements are flagged or temporarily unavailable, NEXT automatically falls back on satellite and NWP inputs, maintaining solar forecast delivery without interruption. The model resynchronizes as valid data resumes.
Isolated, secure processing
Each power plant operates within a dedicated processing environment. All data flows are encrypted end-to-end. Scalable from single plant to multi-GW portfolio. Each site has its own dedicated pipeline and calibration model.
Full data ownership
Your production data and irradiation data are used exclusively for your own forecasting service, never shared with third parties, pooled across clients, or used outside your own pipeline.
Precisa de mais informações?
Entre em contato!
- Perguntas
Frequently Asked Question about NEXT
How are solar forecasts produced by NEXT?
Solar forecasts rely on a combination of satellite data, weather models, and ground-based measurements – integrating both historical data and real-time inputs across multiple time series. Satellite data provides detailed information on current cloud cover and solar irradiance, while weather models predict the large-scale movement of weather systems and their impact on future solar irradiance. Ground-based measurements provide insight into local weather conditions and the real-time output of the solar plant. These input data are continuously updated to ensure forecast accuracy, minimize forecast errors, and maintain optimal model performance.
What are the different types of forecasts?
To ensure optimal operation of solar plants, there are three types of solar forecasts, classified by their time horizon: near real-time, short-term, and long-term forecasts.
- Very short-term forecasts, also known as Nowcasting, cover a time horizon of a few minutes and typically require a sky observation system using hemispherical cameras. They complement standard on-site sensor measurements and satellite imagery, which is less precise. This type of forecast is particularly useful for hybrid microgrid systems in isolated locations, where the alternative to PV is not instantly available, or when there are critical energy constraints – for example, at industrial sites with storage systems requiring precise generation targets.
- Short-term forecasts (intraday) predict solar irradiance for the upcoming hours of the day at regular hourly and sub-hourly intervals. They are crucial for enabling solar energy developers to optimize operations in real time, support intraday trading activity in electricity markets, and help grid operators manage solar energy integration into the network under changing market conditions.
- Long-term forecasts (day-ahead) predict solar irradiance over several days to weeks. These are essential for solar energy stakeholders to plan operations, anticipate daily generation trends, and make decisions to balance energy supply and demand — ensuring a reliable power supply and more predictable electricity prices.
Why does NEXT use on-site data?
Weather models and satellite imagery are excellent sources of data for solar forecasting and are available worldwide. However, their temporal and spatial resolution is insufficient to detect cloud behavior at very short timescales and site-specific levels – limiting their usefulness where on-site analytics and historical data reveal finer generation patterns.
Monitoring sensors typically available on-site provide highly valuable insights into the real-time status of the system with much greater precision. CalibSun’s forecasting algorithm is able to integrate these measurements to dynamically adjust forecasts based on actual data – reducing forecast errors, especially when other data sources fail to anticipate short-term variability accurately.
CalibSun retrieves real-time data through secure pipelines at the required frequency. Whenever possible, real-time data collection is strongly recommended, as it significantly improves forecasting performance and key forecast metrics over short-term horizons.
