Predicting Water and Wastewater Consumption for Prague Water Supply and Sewerage Company
Prague Water Supply and Sewerage Company (PVK), the largest water utility in the Czech Republic, manages more than 117,000 connection points in Prague and its surroundings. Accurate forecasting of water and wastewater consumption is essential for financial planning, accrual calculations, and operational network management. Sophia Solutions designed and implemented a predictive system for PVK that today achieves a billing error rate below 1%.
Requirements
PVK needed to replace existing estimates with a more precise tool for forecasting consumption and billing at the level of individual connection points – across more than 117,000 connections with varying degrees of smart meter coverage.
The goal was a solution enabling monthly forecast updates, simulation of the impact of price changes, early anomaly detection, and clear visualizations to support both financial and operational decision-making.
Solution
Sophia Solutions designed and implemented a comprehensive system made up of five interconnected modules. At its core are approximately 10,000 predictive models (GAMs) optimized for individual connection points, using both billing data and smart meter data; connection points without smart meters are covered through extrapolation from a representative sample.
The solution also includes a strategy for installing an additional approximately 2,500 smart meters to maximize statistical representativeness, causal modeling for scenario analysis of price-change impacts, configurable anomaly detection at both the individual connection-point and entire catchment-zone levels, and interactive Power BI dashboards for billing, consumption, and alert management.
Benefits
- Billing forecast error below 1%
- Consumption forecast deviation of approximately 1.6–3.8%
- More than 10,000 individual predictive models covering 117,000 connection points
- Monthly forecast updates with version archiving for retrospective comparison with actual results
- Real-time scenario analysis of price-change impacts
- Configurable alerts when consumption threshold values are exceeded
- Recommended deployment plan for an additional approximately 2,500 smart meters
- Solution validated retrospectively on 2023–2025 data
- Modular architecture ready for deployment at other water utility companies
Technology: Predictive modeling (GAM), causal and scenario analysis, Power BI.
Customer feedback
"The greatest benefit of the project, in my view, isn't just the accuracy of the forecasts themselves, but the fact that we managed to turn a large volume of operational and billing data into a practical tool for both strategic and operational decision-making. Today we can identify deviations much faster, evaluate scenarios, and respond to changes throughout the year with greater confidence."
