Centram Centre for Applied Artificial Intelligence

Ai automation course for Water and Sewerage Regulation

Target audience:

Regulators, compliance officers, utility analysts, asset managers, lab/data engineers, policy analysts, data scientists working in water/wastewater, and technical staff implementing automation for regulatory workflows.

Course description

This course empowers professionals how AI, ML and automation can be applied safely and practically to water and sewerage regulation. It combines domain fundamentals (hydraulics, treatment, water quality, sampling and permits) with applied analytics: time‑series forecasting, anomaly detection, predictive maintenance, hydraulic and process optimization, RPA and orchestration, explainability, governance and cybersecurity. Students build end‑to‑end automation workflows for common regulatory tasks: compliance reporting, pollution event detection, non‑revenue water/leak detection, demand forecasting, asset risk scoring, and automated inspections and lab result processing.

Learning outcomes

By course end participants will be able to:

1. Translate regulatory questions into data, model and automation requirements
2. Build and evaluate models for demand forecasting, anomaly/leak detection, and asset failure prediction
3. Design automated ETL/orchestration and RPA workflows for monitoring and reporting
4.  Apply explainability, validation and governance appropriate to regulatory contexts
5. Implement basic MLOps practices (versioning, monitoring, retraining triggers)
6. Identify and mitigate legal, ethical and cyber risks in water automation systems

Introduction & Use Cases in Water Regulation

1. Topics: Role of analytics and automation in water/wastewater regulation; high‑value regulatory use cases (compliance automation, CSO/SSO detection, non‑revenue water, contamination alerts, permitting/inspection automation); success stories & risks
2. Lab: Map 3 regulatory questions to data and automation solutions
3.  Deliverable: Capstone project idea proposal

Data Foundations & Water Sector Datasets

1. Topics: Data types: SCADA/telemetry, AMR/AMI meter reads, customer billing, lab test results, hydraulic model outputs (EPANET), GIS layers (pipes, valves, manholes), rainfall/stormwater, complaints, inspection logs; data quality, sampling frequency, censoring, laboratory turnaround & chain‑of‑custody
2. Tools/skills: SQL, Pandas, time alignment, handling censored/left‑/right‑censored lab data, anonymization
3. Lab: Ingest and clean SCADA + sample lab/water quality datasets; build ETL skeleton
4.  Datasets: EPA Water Quality Portal, USGS streamflow, open municipal AMI samples, EPANET sample networks

Time‑Series Analysis & Demand Forecasting

1. Topics: Load/demand/time‑series basics for consumption forecasting, seasonality and calendar effects, ARIMA/SARIMA, exponential smoothing, feature engineering (temperature, holidays, billing cycles), evaluation metrics and time‑series CV
2. Lab: Baseline short‑term demand forecasting for distribution planning and leak detection baselines
3. Deliverable: Forecast baseline report for a service area

ML for Water/Wastewater Applications

1. Topics: Supervised learning for regression/classification, tree ensembles (Random Forest, XGBoost), feature engineering for spatial/temporal data, imbalance handling (rare events), hyperparameter tuning
2. Lab: Build models for meter-level consumption forecasting; customer segmentation for conservation programs

Process Models, Hybrid ML & Hydraulic Integration

1. Topics: Combining physics/hydraulic models (EPANET) with ML (residual learning, surrogate models), digital twins for network hydraulics and treatment plant processes, emulation and speedups
2. Lab: Create an ML surrogate for hydraulic head/flow predictions using EPANET outputs; evaluate surrogate accuracy and speed gains

Anomaly Detection: Leaks, Sensor Faults, Pollution Events

1. Topics: Unsupervised & semi‑supervised methods (isolation forest, autoencoders, clustering), statistical control charts, change‑point detection, differentiating sensor faults from true events, prioritization for inspections
2. Lab: Implement anomaly detection to identify leaks/non‑revenue water and abnormal treatment plant effluent parameters; build an alerting workflow
3. Deliverable: Anomaly investigation playbook

Predictive Maintenance & Asset Risk Scoring

1. Topics: Failure modes in pumps/valves/lines, survival analysis, classification for failure prediction, condition indicators from SCADA and inspection logs, prioritization under budget constraints
2. Lab: Develop asset risk scores and a maintenance scheduling optimizer (simple constrained scheduling)

Optimization & Operational Automation

1. Topics: Pump scheduling and energy optimization (minimize energy cost subject to pressure/level constraints), valve control for pressure management, optimization approaches (linear, mixed integer, heuristics), scenario analysis for droughts or floods
2. Lab: Pump scheduling optimization using historical tariff/energy price data and network constraints

Orchestration, RPA & Automated Reporting

1. Topics: Workflow orchestration (Airflow/Prefect), RPA for regulatory reporting and permit workflows (UiPath concepts), integrating lab results into compliance dashboards, automated incident notification, real‑time vs batch choices
2. Lab: Build an automated ETL + model scoring pipeline that generates compliance reports and triggers RPA to populate submission templates

 MLOps, Versioning & Model Governance for Regulators

1. Topics: Model lifecycle, model registries (MLflow), testing and validation, drift detection, audit trails and lineage, governance requirements for regulatory artefacts
2. Lab: Package a model with versioning and test suites; create a factsheet and retraining criteria

Explainability, Legal & Ethical Considerations

1. Topics: Explainable ML (SHAP/LIME), communicating model outputs to regulated entities/public, fairness (impacts on vulnerable customers), data privacy (GDPR/PII), environmental justice considerations, appeal mechanisms for automated enforcement

Capstone project ideas

1. Automated leak detection and inspection prioritization pipeline (AMI + pressure/flow + weather)
2. Real‑time pollutant exceedance detection and notification system with explainable alerts and compliance workflow
3. Predictive maintenance for pumps/valves: risk scoring + scheduling optimizer
4. Automated lab result ingestion + compliance report generation and RPA submission to regulator portal
5. Demand forecasting + drought scenario simulator to inform temporary restrictions and tariff adjustments
6. Non‑Revenue Water dashboard combining hydraulic models, sensor analytics and customer billing reconciliation

Tools & technologies recommended

1. Languages: Python (Pandas, scikit‑learn, XGBoost, TensorFlow/PyTorch), R optional
2. Hydraulic/process: EPANET (and toolkit), SWMM (stormwater), HEC models where applicable
3.  Time‑series: statsmodels, Prophet, tsfresh
4. Orchestration & MLOps: Airflow/Prefect, MLflow, Docker, Kubernetes (intro)
5. Datastores: PostgreSQL/PostGIS, Parquet, cloud storage (S3/Azure Blob)
6.  GIS: QGIS, GeoPandas
7.  BI/visualization: Power BI, Tableau, Dash/Streamlit
8.  RPA/Automation: UiPath concepts, Python API automation, webhook/alerting integrations

Data sources & reading resources

1. EPA Water Quality Portal, USGS streamflow, WHO/EPA drinking water guidance
2. EPANET and SWMM documentation and sample networks
3. IWA (International Water Association) publications on analytics and asset management

• Case studies from municipal utilities on leak detection and NRW reduction

• Papers on sensor fault detection, hydraulic model‑ML hybrids, and water quality anomaly detection

Ethics, governance and stakeholder engagement

1. Emphasize transparency to utilities, customers and the public (explainability, appeals)
2. Model factsheets, impact assessments and public consultation mechanisms
3. Policies for human oversight before automated enforcement or penalties

Environmental justice and equity considerations in targeting interventions or restrictions