EECS Theses and Dissertations

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 357
  • Item
    Verification of protection blinding in a real-world simulation model
    (Wichita State University, 2024-05) Rahman, Mohd Abrar; Aravinthan, Visvakumar
    The addition of renewable energy into the distribution comes with a large number of environmental benefits but does give rise to some risks associated with the protection system. Issues such as protection blinding, reverse flow, and sympathetic tripping are the most common that have been researched for the time over current relays. The coordination of these relays is at risk and verification of the relay settings on real world systems needs to be studied. This paper will build an urban Midwest utility Feeder distribution model in OpenDSS with the help of a local utility company and referencing one of their urban distribution models and verify if the standard protective relay settings are at risk of misoperating at the substation level. The study will be conducted on 3 different sized distributed energy resources (DER) and on two different locations on the same feeder, close to the substation and further away from the substation. Three different testing scenarios are conducted: normal flow with no faults on the system, faults only on the feeder that has the DER connected to it, and faults on neighboring feeders. Adjusting the DER size and location along with different fault types will allow to find the worst combination for protection blinding on the simulated model. The results show that the urban system in study is not at risk of protection blinding but is susceptible to sympathetic when the DER is large enough to back feed into the system. However, the placement of DER on the main feeder branch will avoid sympathetic tripping as the Switchgear feeder is the first protective device and is set high enough to avoid a misoperation. Placement of the DER on the Feeder branches does pose a threat and will cause a misoperation.
  • Item
    Application of artificial neural networks in low-thrust cislunar estimation, prediction and control
    (Wichita State University, 2024-05) Thulaseedharan Pillay, Yrithu; Watkins, John Michael; Steck, James E.
    There is an increased interest in cilsunar missions, i.e., missions in the Earth-Moon space due to their strategic importance. Many of these missions will use Solar Electric Propulsion (SEP) due to its lower propellant usage. However, SEP has significantly lower thrust compared to traditional chemical propulsion resulting in a significantly higher flight time. Moreover, cislunar spacecraft experience the three-body dynamics due to the presence of the Earth and the Moon. The three-body dynamics are extremely non-linear, chaotic, and computationally expensive to solve. This dissertation explores the use of Machine Learning and Artificial Neural Networks (ANN) to aid in cislunar path planning and control while using easier-to-solve low-fidelity spacecraft dynamics. First, an online single-layer neuro-adaptive estimator called Modified State Observer (MSO) is used to capture the unmodeled perturbations acting on a spacecraft operating in the cislunar space. The performance of the developed algorithm is investigated through numerical simulations, demonstrating significant reduction in modeling errors. Then a multi-layer off-line ANN is used to learn and predict the unmodeled dynamics in low-trust multi-revolution cislunar missions. The ANN-based dynamics prediction scheme results in significantly lower errors for subsequent revolutions. Finally, the MSO is combined with a Non-Linear Dynamic Inversion (NDI) controller to create a neuro-adaptive control scheme for cislunar spacecraft. Exploring the effectiveness of the control law, it is seen that while the control scheme works, it cannot realistically provide control for a low-thrust spacecraft with reference trajectory designed in the very low-fidelity CR3BP model. However, it demonstrated good reference-tracking with low thrust, in the case of trajectory designed in high-fidelity dynamics experiencing unmodeled perturbations due to a Missed Thrust Event (MTE).
  • Item
    Energy storage optimization using modified cuckoo search
    (Wichita State University, 2023-12) Wiebe, Kyle Garrett; Aravinthan, Visvakumar
    Renewable generation is a topical area of research due to its low environmental impact compared with conventional generation and the current push to reduce fossil fuel emissions. However, inconsistencies in renewable power generation can cause a variety of power system problems. Energy storage technologies, such as batteries, are able to be used to help smooth out the inconsistent generation patterns of renewable resources. This ability to, partially or completely, remedy the unpredictability inherent to renewable generation makes battery storage an important topic for academic progress and optimization to increase industry utilization. There are numerous parameters pertaining to batteries and the power grid that can be optimized, one of which being economic optimization, which is the subject of this thesis. This optimization goal enhances battery systems’ feasibility of installation in the industry. This thesis presents a modified version of Cuckoo Search Optimization [1] and validates it with a more conventional optimization technique, Dynamic Programming [2]. It is then shown how this optimization technique can be applied to optimize a battery’s charge and discharge schedule and to implement economic battery sizing for various battery capital cost estimates.
  • Item
    Mapping business entities to intersections for network simulaiton
    (Wichita State University, 2023-12) Nawal, Nowshin; Asaduzzaman, Abu
    Business entities (i.e., shops) in a region can be considered as network nodes, where the profit of a shop is related to the vehicular traffic at the adjacent intersections. Argonne National Laboratory (ANL) and UChicago Argonne have developed Portable Extensible Toolkit for Scientific Computation (PETSc) and Data Management Network (DMNetwork) libraries to conduct network analysis. However, PETSc and DMNetwork do not provide a way to map a business entity to a network node. In this work, a methodology has been developed for mapping shops to the associated intersections using their (longitude, latitude) values for profit analysis through network simulation. OpenStreetMap (OSM) is used to generate the initial network containing traffic nodes and edges. Then, Google Maps Platform (GMP) and LocationIQ (LIQ) Application Programming Interface (API) tools are used to generate the locations’ (longitude, latitude) values of shops. The developed methodology then assigns a node to a shop using the shortest distance. As a result, each and every shop is mapped to its nearest intersection. Unlike OSM, both GMP and LIQ offer dedicated services to provide geo-location data of shops. While both GMP and LIQ follow the same file structure, LIQ provides a better source for shops. The newly generated files with shop information are used for network simulation and profit analysis using PETSc and DMNetwork libraries. According to the simulation results, PETSc reduces the computation time by 50% if two processors are used instead of one. This work can be extended to automate the mapping process and incorporate shops of multiple regions simultaneously.
  • Item
    Sensitivity analysis in gyrator synthesis
    (Wichita State University, 1968-01) You, Chen-nan; Hoyer, Elmer A.
    The purpose of this thesis is to apply both classical sensitivity function and pole-zero sensitivity to a method of active network synthesis developed by Hoyer.
All items in SOAR are protected by copyright, with all rights reserved, unless otherwise indicated.