Computational Intelligence and Operational Water Management

IHE Delft digital water

Target Audience

The course is designed for engineers and scientists involved in operational water management and control and interested to broaden their knowledge of modern approaches and modelling tools. The course could be also interesting to PhD and Master students conducting research in real-time control of water systems and/or use of data-driven models in hydrology, hydraulics or environment. Pre-requisites are a basic knowledge of mathematics, hydrology and systems analysis.


The course provides information on real-time control systems for operational water management and the use of computational intelligence methods to build data-driven models.

Course content
The course includes three main parts:

1. Introduction to optimisation.
Classical optimisation. Linear and non-linear optimisation. Derivative-based and direct methods. Dynamic programming. Global (multi-extremum) optimisation. Genetic and evolutionary approaches. Multi-objective optimization. Applications in water sector. Exercise: optimal water allocation; automatic model calibration.
2. Real time control of water systems.
Introduction to Real-Time Control. Modelling hydrological systems and optimal control problems with AQUARIUS and SWMM. Control-systems functions and techniques. Hardware and software components. Control systems in industry. Identifying control system components. Use of Kalman filters. One day field trip to North-West Netherlands.
3. Data driven modelling and computational intelligence.
Modelling in the framework of Hydroinformatics. Data-driven and physically based models. Overview of machine learning and computational intelligence. Decision, regression and M5 model trees. Artificial neural networks. MLP and RBF networks. Instance-based learning. Fuzzy logic and fuzzy rule-based systems. Hybrid models combining simulation and data-driven models. Error correction and data assimilation techniques.

Learning Objectives

Upon completion, the participant should be able to:
– Understand and be able to formulate and solve an optimisation problem in relation to water systems (model calibration, reservoirs, urban pipe networks)
– Understand and explain how real-time control systems work
– Identify the potential of control to solve hydrological problems
– Sketch a general plan for a regional real-time control system
– Appreciate and apply the main techniques of data-driven modelling (machine learning): neural networks, model trees, instance-based learning, and select proper methods and tools for building data-driven models
– Correctly classify a modelling problem as a physically-based, data-driven, or hybrid