Discipline: Technology and Engineering
Subcategory: Computer Engineering
Room: Exhibit Hall A
Oshin Wilson - University of the District of Columbia
Co-Author(s): Dr.Nian Zhang, University of the District of Columbia, 4200 Connecticut Ave NW, Washington, DC 20008
Land development activities inevitably change watershed conditions, primarily due to an increase in the impervious area through paving, construction, adding drainage systems and removal or alteration of vegetation which results in water quantity and quality problems of local receiving bodies. The examples of water quantity problems include flooding, stream bank erosion, while examples of water quality problems include pollution loading and receiving water impairments. The impact of this type of activity is more pronounced for highly urbanized areas and the associated receiving waters such as the Potomac River and Anacostia River within the Chesapeake watershed. In addition, it has been recognized that climate change can have severe impacts on our streams and rivers due to extreme weather events such as frequent flooding. In this regard, reliable estimation of stream flows at various locations is very important from the water resources management viewpoint. Engineers, water resources professionals, and regulatory authorities need this streamflow information for planning, analysis, design, and operation & maintenance of water resources systems (e.g., water supply systems, dams, and hydraulic structures). Currently USGS provides the streamflow data at various locations in the form of gage height and discharges at specific locations, and I used this input to design a reliable prediction model. In this research, I am proposing different efficient computational intelligence methods that forecast the future streamflow discharge using the past streamflow data and gage height. The first one is a predictive model based on least squares support vector machine (LS-SVM). The training process of LS-SVM involves the selection of kernel parameters and the regularization constant. Then the prediction capability with different kernel functions will be explored to evaluate the impact of the streamflow discharges and gage height for long-term prediction of flow rates and its accuracy. The second option is a hybrid learning algorithm incorporating particle swarm optimization and an evolutional algorithm, which takes the complementary advantages of the two global optimization algorithms. The neural networks model will be trained by particle swarm optimization and evolutional algorithms to forecast the storm water runoff discharge. The USGS real-time water data at various locations will be used as time series input. These computational methods are not limited to runoff data, and once I established the precision and effectiveness of the LS-SVM algorithm I can branch out to a wide variety of applications and apply this algorithm to many time-series based simulation and prediction problems.
Funder Acknowledgement(s): This study was supported by a grant from the University of the District of Columbia (NSF/HBCU-UP/ HRD #1622811), Washington, D. C. 20008
Faculty Advisor: Dr.Nian Zhang, email@example.com
Role: In this research, I am proposing different efficient computational intelligence methods that forecast the future streamflow discharge using the past streamflow data and gage height. And as such I carried out different procedures to get answers to the problem statement and I was also assisted by my mentor in making the poster.