Grantee Research Project Results
Final Report: An Artificial Neural Network and Optimization Methodology for Detecting and Managing Terrorist Attack Against Water Distribution Systems
EPA Contract Number: EPD04026Title: An Artificial Neural Network and Optimization Methodology for Detecting and Managing Terrorist Attack Against Water Distribution Systems
Investigators: Coppola, Emery A.
Small Business: Neural Optimization Applied Hydrology LLC
EPA Contact: Richards, April
Phase: I
Project Period: March 1, 2004 through August 31, 2004
Project Amount: $69,913
RFA: Small Business Innovation Research (SBIR) - Phase I (2004) RFA Text | Recipients Lists
Research Category: Watersheds , SBIR - Homeland Security , Small Business Innovation Research (SBIR)
Description:
This research project was conducted by NOAH, LLC, with participation from American Water (the largest privately owned water supplier in United States) to assess the feasibility of using artificial neural networks (ANNs) in conjunction with mathematical optimization for developing a “smart” water distribution security system. The developed system would be used to detect potential terrorist attacks against a water distribution system and, in the event of such an attack, identify the optimal operational responses to mitigate it to the extent possible. Using hydraulic and water quality data provided by a mid-sized city’s water distribution system, hundreds of different ANNs were developed and tested for predicting hydraulic states and water quality conditions at points of measurement within the system. The predictive accuracy of different ANNs was assessed under a variety of operational and data set conditions, and data requirements and other critical modeling issues were identified. In addition, several hydraulic ANN models were combined with mathematical optimization to quickly (i.e., within seconds) identify the optimal operational controls for a given management objective (e.g., maximizing tank water levels for fire protection).
Summary/Accomplishments (Outputs/Outcomes):
Research results demonstrated that the ANN technology can be used to accurately predict dynamic hydraulic conditions in complex water distribution systems. It was surprising that even with relatively large dynamic changes in the systems, such as tank water levels, the ANNs were able to accurately predict final states over relatively long periods of time (e.g., 8 hours). The mean absolute prediction errors were significantly smaller than the mean absolute changes. For example, for Tanks 1 and 2, the mean absolute prediction errors for the 8-hour prediction periods were 0.54 and 0.73 feet, respectively, while the corresponding mean absolute changes in water levels in these tanks were 5.6 and 7.5 feet. The ANNs also accurately predicted highly variable and dynamic pressure states in the water distribution systems.
The water quality models accurately predicted quasi-steady-state chlorine levels, but were unable to predict sudden large changes (e.g., 0.5 ppm in 5 minutes). This is attributed to insufficient characterization of highly time-variable chlorine levels at critical water source/sink (i.e., tank) locations, which could possibly be achieved through mass balance calculations with existing data, but probably would require additional sensors. One potential benefit of the ANN models is that they can help improve monitoring strategies for the system and can provide quality assurance/quality control (QA/QC) for data; for example, the ANNs in this research identified likely spurious hydraulic data for one of the months considered.
For conducting mathematical optimization, a variety of objective functions were either maximized or minimized subject to imposed management constraints. The ANN-derived state transition equations were coupled with LINDO proprietary optimization software, and 140 different optimization problems were solved. Because of the complexity of the water distribution systems—not just in terms of physics, but in terms of the water balance constraints, minimum and maximum constrained flow rates, minimum and maximum constrained tank water levels, and minimum and maximum constrained pressures—the solutions were evaluated for accuracy using a number of heuristic criteria. First, the computed optimal solutions for conflicting objectives, such as maximizing a certain tank’s water level and minimizing the same tank’s water level given the same initial state, were compared in terms of values of the decision variables and final states. Second, it was verified that the final values of the state and decision variables were all within the imposed minimum and maximum constraint values. Third, it was verified that the two water balance constraints imposed for the higher and lower pressure zones were met. In all cases, the optimal solutions were within the bounds of the constraint set. In addition, the computed decision variables within the context of the objectives displayed numerical consistency and conformed to physical intuition.
These research results demonstrated great potential for developing and commercializing a robust water security system that can be multifunctional in purpose. The ANNs already have been used in the water industry for reducing treatment costs and for monitoring water quality trends for anomalous conditions. The methodology developed here, however, is different from other reported work in at least two major areas. First, it represents the first known case in which the ANNs were trained with real-world data to explicitly compute relevant hydraulic states, such as pressures and tank water levels, for a water distribution system. Second, the ANN-derived state transition equations, which constitute a highly efficient yet accurate physical system simulator, were combined with mathematical optimization to quickly compute (i.e., seconds) the optimal values for the decision variables (i.e., pumping rates) that maximized (i.e., benefit) or minimized (i.e., cost) the specified objective without violating any imposed constraints.
The ANN-based prediction and management methodology is an ideal complement to the Supervisory Control and Data Acquisition (SCADA) type of systems that are increasingly utilized by water utilities. These automated data collection systems measure and record system conditions such as water levels, flow rates, and pressures at high frequencies. These data sets are vastly underutilized by the industry, making the returns on investments on data collection systems questionable in many cases. The ANNs, often referred to as “data-driven,” thrive on large data sets, and represent the ideal complement to these data sets. The ANNs can perform important QA/QC analyses of the data, identify important cause-and-effect relationships, and perform real-time prediction and optimization. By interfacing the ANN-optimization management tool with the SCADA systems, the ANN-derived state transition equations can be initialized to existing conditions, increasing the accuracy of the predictions and computed optimal decisions.
One of the powerful advantages of the ANNs approach is that different models can easily be integrated together for conjunctive management of complex, multicomponent water systems. For example, NOAH, LLC, has demonstrated in earlier work that ANNs can accurately predict complex hydrologic behavior, such as transient groundwater responses to pumping and climate conditions. It is envisioned that the multifunctional ANN-optimization methodology can be used in an integrated manner so that utility operations can be optimally coordinated and managed. The ANNs could compute a variety of system outputs, such as energy costs as a function of water source and treatment, and collectively, the ANNs system could be used to minimize operational costs without violating regulatory and/or environmental constraints, such as wetland impacts or salt-water intrusion. The flexibility and adaptability of the technologies lend themselves to solutions for multiobjective problems that may even be conflicting.
For water security, the multistate and redundant model approach developed in this research provides an added element of safety for water utilities concerned with terrorist attacks. This type of real-time, computer-automated assessment can be used to help determine whether a potentially anomalous condition has occurred and reduces the human element of uncertainty. The system would make use of the SCADA data to continuously assess hydraulic and water quality conditions. Simultaneously, the system could be performing other important functions, such as minimizing energy consumption costs or enhancing other daily operational decisions.
Conclusions:
In this project, NOAH, LLC, demonstrated the potential for developing multistate ANN-based water security systems that continuously assess both hydraulic and water quality conditions in real time for identification of a possible terrorist attack (or accident). A number of possible methods for improving prediction of sudden and significant water quality changes with relatively limited water quality data require additional research. In addition, by coupling the technology with mathematical optimization software, appropriate crisis responses can be identified quickly and accurately. The research results to date, particularly with regard to hydraulics, demonstrate strong promise for protecting water distribution systems with the ANNs system as well as performing other important functions, such as minimizing energy consumption.
In this research, NOAH, LLC, with participation from American Water, also has identified a number of ways to improve the methodology. Although Phase I was limited to in-house development and testing, implementation of software at the participating water distribution system in Phase II will allow for further development and improvement of the system and enhance its adaptability to other water industry problems. During implementation, it will be benchmarked against actual system operations, providing quantifiable measures of benefits provided by the ANN optimization system.
As data collection systems and sensor instrumentation evolve and become more prevalent, the need for a tool that can process this data in real time to facilitate accurate and important management decisions will grow. In addition, the increasing scarcity of water resources and growing conflict among competitive users will require superior prediction and management tools. The convergence of data acquisition with ANN-optimization modeling will represent a major advance in this area.
Supplemental Keywords:
artificial neural network, ANN, terrorist attack, water distribution systems, mathematical optimization, hydraulic states, water quality conditions, water security system, monitoring, Supervisory Control and Data Acquisition system, SCADA, monitoring, SBIR,, RFA, Scientific Discipline, INTERNATIONAL COOPERATION, Water, Ecosystem Protection/Environmental Exposure & Risk, Chemical Engineering, Environmental Chemistry, Chemistry, Monitoring/Modeling, Analytical Chemistry, Environmental Monitoring, Environmental Engineering, Drinking Water, Engineering, Chemistry, & Physics, artificial neural network, homeland security, monitoring, detection, environmental measurement, field portable monitoring, water distribution security, biopollution, drinking water regulations, community water system, field monitoring, chemical detection techniques, analytical methods, environmental contaminants, crisis management, measurement, drinking water contaminants, drinking water securityThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.