Hydrological forecasting
Water-quality forecasts from sensor data.
Visit watermap.pl (opens in a new tab), watersense.pl (opens in a new tab)
Facts
- for
- Water monitoring
- data
- Monitoring station readings
- output
- Forecast API
- status
- In production
How it works
- station readings
- prediction model
- parameter forecasts
The need
Water-quality readings vary by time and place. A useful forecast needs consistent time-series processing, evaluation suited to each parameter and a dependable path to daily use.
What we built
We delivered the predictive layer: from data analysis and model development, through training, to a forecast API. We also contributed to the data-management design around the deployment.
- Multivariate time series · Readings from the stations are processed consistently.
- Model comparison · We compared Transformer and TimesNet approaches. Quality is measured by the forecast error for each parameter.
- Training and forecasts · One path from model development, through training, to forecasts served by an API.
What changed
A custom deep-learning model runs in production and produces parameter forecasts through the inference layer (a forecast API).
Production deployment and scale as reported by ArtNovation in September 2026. The work covers a stochastic environment.
A similar task?
Do you collect readings from sensors, machines or other systems and want to know earlier what they will show? We start with the data you already have and how we will check forecasts against what actually happens.
