Real-time distribution system modeling: Development, application, and insights

Water distribution system models and computational aids in general are becoming ever more commonplace in the planning and decision-making practices of utilities. Moreover, these models are being used for exceedingly complex tasks involving water quality prediction, sensor design, and disaster management. To trust a hydraulic model without proof of its accuracy would be foolish, but models are seldom verified or scrutinized on a continuous, operational basis. To be sure, a professionally calibrated model will emulate the behaviour of a distribution system for a certain time span under certain conditions, but there exists no open and accessible framework for validating a model in real-time operational scenarios, or for extended time periods of several years or more. The current landscape of interconnected and open information architectures, along with the availability of vast quantities of SCADA data and computing resources, gives a researcher little excuse for delaying this validation step. Indeed, models must be subject to such scrutiny if they are to be trusted for critical decision-making. The work described here builds on prior efforts in the development of a real-time hydraulic modeling framework by field-testing such a system. This "Real-Time Extension" to EPANET (so-named EPANET-RTX) is installed at a water utility, and key personnel are given basic means of interaction with it. The software connects a model's controls, demands, and boundary conditions to real-time SCADA data, and gives visual output of the model's predictions and statistical accuracy metrics. In addition to viewing error statistics and time series, personnel are capable of adjusting the hydraulic model's parameters dynamically and exporting historical scenarios (as EPANET input files) for offline analysis. The steps taken to field-validate critical model details and implement real-time simulation are outlined. Key visualization and statistical analysis steps are also presented. Further, this research attempts to document the utility's experience and its recommendations for future development and deployment of the software tools. Close collaboration between the researchers and water utility personnel has generated substantial enthusiasm for sharing knowledge and improving both the model and instrumentation. The development of the RTX platform and subsequent pilot-scale installation is a first step in the path to creating an extensible and open-source framework for real-time hydraulic modeling.