Research Project

L-BEST: Advanced Learning-Based supervision for Efficiency and Safety in smart infrasTructures


National Project

Start Date


End Date


Project Code


Project illustration


Project Description

Project PID2020-115905RB-C21 funded by MCIN/ AEI /10.13039/501100011033

Smart infrastructures (SIs), such as water and energy networks, buildings, among others, are most important in modern society because of the services they provide and the essential resources they manage. In SIs, safety and efficiency must be taken into account in all steps of their life cycle, including design, operation and validation of the resulting performance.

L-BEST considers research in supervision and fault-tolerant control of SIs by means of two related methodologies: advanced learning and optimization. The use of advanced learning from data is supported by the fact that SIs include sensors and smart-grid technology providing a continuous flow of operational data. First-principles models alone may not capture the complex behaviours of these SIs, so that a combination with learning from operational data is proposed. Optimization-based control is concerned with computing strategies which improve specific performance indicators, such as those related to efficiency and safety. Advanced optimization and learning are proposed for this type of problems in large-scale and complex SIs.

The specific goals of L-BEST are grouped in five complementary axes:

(i) Designing learning-based approaches for monitoring, forecasting and control of SIs taking into account interaction of multiple sensors and actuators as well as uncertainty.

(ii) Developing reconfiguration methodologies for fault-tolerant control schemes applied to complex SIs.

(iii) Designing distributed control techniques for SIs based on advanced learning methods and evolutionary game theory.

(iv) Contributing to ontologies definition for platform development in order to provide a flexible solution capable to cover different nature of SIs.

(v) Implementing demonstrators for the design and implementation of the proposed learning-based supervision approaches and analyse the economic, social and environmental impact of the demonstrator results.

The research team has a strong background in two key areas: methodology developments towards optimization-based control of large-scale SIs (especially energy and water infrastructures, e.g., drinking water and sewer networks, canals and rivers, electric networks, clusters of microgrids); and fault diagnosis and fault-tolerant control towards the supervision of complex industrial systems.
The team also includes two researchers from Cetaqua Water Technology Centre, a foundation of the water company Aguas de Barcelona, CSIC and UPC, and a reference centre in the application of scientific knowledge to sustainable management of water and environmental systems. This collaboration is essential for the orientation of research towards real industrial and market needs, as well as for maximizing the impact and knowledge transfer towards industry. Realistic demonstration case studies will be provided by Cetaqua and industrial partners (EPO). Specifically, the case studies in smart water networks will be related to: monitoring for leak detection and localization in water networks, monitoring water quality, water and energy efficiency in water networks and network reconfiguration after fault, such as bursts or contamination events.

The expected impact of L-BEST has several aspects; namely: a scientific impact for the automatic control and SI management communities, dissemination in top journals, communication actions to more general public and, most importantly, technology transfer to the water industry.