Improving Data for Real-Time Control of Urban Sewer Systems: Sensor Validation and Data Assimilation
Author
Brødbæk, Daniel
Term
4. term
Publication year
2013
Submitted on
2013-05-31
Pages
88
Abstract
Formålet med denne kandidatafhandling er at forbedre datagrundlaget for realtidsstyring af urbane kloaksystemer, så beslutninger bygger på pålidelig viden om systemets aktuelle tilstand. For at øge tilliden til målinger fra in-situ sensorer (sensorer installeret i systemet) er der udviklet en metode, som validerer sensordata i realtid; metoden er afprøvet i et casestudie. Når styringen baseres på en deterministisk hydraulisk model (en computersimulation af vandstrømmen i nettet), kan forudsigelsesfejl forringe ydeevnen. Til gengæld beskriver modellen hele systemet, mens sensorer kun måler få steder. For at få det bedst mulige billede af systemets tilstand sammenkøres sensordata med den kørende model via data-assimilering. Konkret anvendes Ensemble Kalman Filter (EnKF), som beregner den mest sandsynlige tilstand ved at afveje usikkerheder i både målinger og model. Filteret opdaterer hele modellen, ikke kun der hvor sensorerne sidder. Afhandlingen beskriver, hvordan EnKF kan konfigureres til dette formål.
This thesis aims to improve the data used for real-time control of urban sewer systems, so that decisions are based on reliable information about the system’s current state. To increase confidence in measurements from in-situ sensors (sensors installed in the system), a method that validates sensor data in real time was developed and evaluated in a case study. When control relies on a deterministic hydraulic model (a computer simulation of how water flows through the network), prediction errors can impair performance. However, the model represents the whole system, whereas sensors provide observations at only a few locations. To obtain the best possible picture of the system state, sensor data are combined with the real-time model using data assimilation. Specifically, the thesis applies the Ensemble Kalman Filter, which estimates the most probable state by balancing the uncertainties of measurements and model predictions. The filter updates the entire model, not only the locations with sensors. The thesis explains how to configure the Ensemble Kalman Filter for this purpose.
[This abstract was generated with the help of AI]
Documents