Contribution to big data analytics in water networks
- Autores: Diego García Valverde
- Directores de la Tesis: Vicenç Puig (dir. tes.), Joseba Jokin Quevedo Casin (dir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2017
- Idioma: español
- Tribunal Calificador de la Tesis: Ramón Pérez Magramé (presid.), Xavier Litrico (secret.), Enrico Creaco (voc.)
- Programa de doctorado: Programa de Doctorado en Automática, Robótica y Visión por la Universidad Politécnica de Catalunya
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
The Critical Infrastructure Systems (CIS), as the water distribution networks, are operated and managed by utilities by means of the information gathered from the water network. The water utility (WU) uses supervision, control and data acquisition systems (SCADA), to monitor on-line parameters (e.g. reservoir levels, flows, pressures, ...) at different points of the water distribution network. This information allows the WU to define the control strategy to supply water to the consumers. Hence, the safe and efficient operation of the water distribution network is directly linked to the quality of the information gathered.
The WU is also responsible of guaranteeing the safety and sanitation of the water supplied to the consumers. A common water disinfection technique to provide safe drinking water is chlorination (consists in injecting chlorine in the water distribution network). Thus, aside from monitoring hydraulic parameters, WU monitors also water quality (e.g. conductivity, pH, chlorine and temperature) in order to make the chlorination work properly and to detect quality events that could pose a risk to the consumers health.
However, the chlorine is a complex parameter to measure on-line with reliability. The sensor used is composed of degradable elements over time (e.g. the membrane and the electrolyte). This results in additional anomalies in the on-line measurements (e.g. loss of sensitivity and miscalibration), producing several incidences and requiring corrective maintenance actions.
In general, the sensors, whose function is to transform a physical parameter into a digital signal, can be affected by different problems degrading the quality of the gathered information. In addition, the communication system used to collect this information is not immune to problems (e.g. transmission faults, electric power outages, ...).
Moreover, the volume of data to gather and handle is another challenge, because in many cases cannot be managed with traditional storing and processing tools. Emergent technologies such as Big Data provides tools to acquire, store, query and visualize large data volumes.
In particular, automatic meter reading (AMR) systems produce large volumes of data. This technology allows the WUs to know remotely the end-user water consumptions, and with higher frequency (e.g. hourly). These systems allows to reduce the time between meter readings (traditionally: bimonthly or quarterly) and to improve the readings reliability. This information enables to improve the management and operation of the water distribution network: better demand characterization, better planning and better leak detection precision; and to provide new services: new tariff systems and real-time feedback to the user (e.g. consumption alarms, bill forecasting and recommendations).
In this thesis, a set of methodologies to transform unreliable raw data into valuable information to the WU have been proposed. With this aim, firstly, a consolidation strategy for sensors data has been defined.
Regarding the water quality management, it has been defined a methodology to detect and predict quality sensors malfunction. This methodology allows the WU: to decrease the number of corrective actions and to define a predictive maintenance planning.
With regard to the AMR data management and exploitation, a segmentation has been defined to extract homogeneous groups of consumers based on the consumption behavior, using Big Data technologies. This allows the WU to extract and analyze demand patterns and to detect abnormal consumptions. Finally, an early warning system (EWS) using AMR data has been defined to detect anomalies in the water distribution network. And an individual analysis of the consumption to identify the meter causing the problem.
