Design Solution · Water Management
Design Solution · Dream about it
ML-powered short-term wastewater level forecasting to prevent sewer backups and CSO events.
CANN Forecast's InteliFlow applies machine learning to precipitation and distributed flow-meter data to generate 0–3 hour ahead predictions of wastewater levels. This enables municipal water authorities to operate gates and pumps proactively rather than reactively, reducing combined sewer overflow (CSO) events and basement flooding during heavy precipitation. The system integrates real-time sensor networks with historical hydraulic behavior to outperform traditional physics-based models in operational deployment.
CANN Forecast's InteliFlow applies machine learning to precipitation and distributed flow-meter data to generate 0–3 hour wastewater level predictions, enabling municipal operators to pre-emptively manage gates and pumps before a combined sewer overflow event occurs rather than reacting to it — which is exactly the operational window that matters for CSO mitigation and basement flooding prevention. The claimed cost advantage relative to grey-infrastructure expansion (tunnels, storage tanks) is plausible and, if validated, compelling for municipalities facing regulatory CSO compliance pressure without capital budgets for physical expansion. However, no provided evidence exists and the inndex of 72 sits alongside high-severity risks on accuracy data and BIM fidelity — the 40+ deployment claim is asserted rather than independently documented, and the public performance record for false-positive rates and CSO reduction verification across diverse drainage typologies is thin. Two structural dependencies constrain the system in ways that are not resolvable by software: a dense, reliable distributed sensor network is a prerequisite, and retrofitting legacy municipal systems with adequate flow-meter and telemetry coverage is capital-intensive and operationally disruptive in itself; and the 3-hour forecast horizon is only actionable if the downstream SCADA system and operating staff can respond within it, which depends on operational readiness the technology cannot provide. ML performance also degrades under unprecedented precipitation patterns — exactly the climate-shifted extreme events that CSO stress cases are built around. High cybersecurity risk for live municipal infrastructure is correctly flagged and requires explicit governance. Worth pursuing as part of a funded feasibility programme with defined sensor network investment and operational readiness assessment; not a drop-in solution for a municipal drainage authority without that infrastructure already in place.
Company origin (2016 AquaHacking win) and Montreal client (2017) are verifiable. Named deployments in 7+ municipalities are plausible and largely publicly documented. Claim of '40+ clients' across three geographies is unverified by public registries; no third-party audit of deployment count found. 'Outperformed hydraulic models' claim lacks peer-reviewed publication or independent validation; Imagine H2O selection is credible but competitive selection is outcome-dependent, not product-proof. Website source is code-only (no published case study or performance data). No evidence on model accuracy in edge conditions (snow, extreme multi-day rain, sensor failure), cost per deployment, integration friction, or retention rate.
#machine-learning #stormwater-management #cso-prevention #real-time-forecasting #smart-water-infrastructure #municipal-adaptation