Date of Award
1-1-2019
Document Type
Dissertation
Degree Name
Ph.D.
Organizational Unit
Daniel Felix Ritchie School of Engineering and Computer Science, Electrical and Computer Engineering
First Advisor
Mohammad Matin, Ph.D.
Second Advisor
Vijaya Narapareddy, Ph.D.
Third Advisor
Jun Zhang, Ph.D.
Fourth Advisor
David Wenzhong Gao, Ph.D.
Fifth Advisor
Amin Khodaei, Ph.D.
Keywords
Photovoltaic based, Solar photovoltaic, Long-short term memory
Abstract
The world of energy sustainability landscape is witnessing high proliferation of smartgrids and microgrids, it has become significant to use intelligent tools to design, operate and maintain such crucial systems in our lives. Solar energy is an intermittent source and purely Photovoltaic (PV) based, or PV and storage based smartgrids require characterization and modelling of PV resources for an effective planning and effective operations. This dissertation familiarizes briefly the existing tools for design, monitoring, forecasting and operation of a solar system in smart electric grids infrastructure and proposes a unique application-based infrastructure to monitor, operate, forecast and troubleshoot a working PV of a smartgrid. A resilient smartgrid communication is proposed which enables monitoring and control of different elements in any PV system. This communication architecture is used to facilitate a feedback-oriented monitoring of different elements in a microgrid ecosystem and investigated thoroughly. This integrated architecture which is a combination of sensors, network elements, database and computation elements is designed specifically for solar photovoltaic (PV) powered grids on modular basis. Apart from this, the network resilience and redundancy for smooth and loss less communication is another characteristic factor in this research work.
Subsequently, a deep neural network algorithm is developed to diagnose the underperformance in the generation of a PV system connected to a smartgrid. As PV generation is predominantly dependent on climatic parameters, it is necessary to have a mechanism for understanding and diagnosing performance of the system at any given instance. To address this challenge, this deep neural network architecture is presented for instantaneous performance diagnosis. The proposed architecture enabled modeling and diagnose of soiling and partial shade conditions prevalent with an accuracy of 90+%. Features of monitoring and regulating the generation and demand side of the grid were integrated through network along with feedback-based measures for effective performance in the PV system of a smartgrid or microgrid using the same network. The novelty in this work lies in real-time calculation of ideal performance and comparison for diagnosing critical performance issues of solar power generation like soiling and partial shading.
Furthermore, long-short term memory (LSTM), which is a recurrent neural network model, is created for forecasting the PV solar resources, in which can assist in quantifying PV generation in various time intervals (hourly, daily, weekly). PV based smartgrids often experience expensive or inaccurate resources planning due to the lack of accurate forecasting tools where the projected methodology would eliminate such losses.
This research work in its whole provides a different proposition of vertical integration which can transform into a new concept called Internet of Microgrid (IoMG). Planning, monitoring and operation form the core of smartgrids administration and if intelligent tools intertwined with network are being used as integral part in each of these aspects, then it forms a holistic view of smartgrids.
Publication Statement
Copyright is held by the author. User is responsible for all copyright compliance.
Rights Holder
Ahmad Almadhor
Provenance
Received from ProQuest
File Format
application/pdf
Language
en
File Size
121 p.
Recommended Citation
Almadhor, Ahmad, "Proactive Monitoring, Anomaly Detection, and Forecasting of Solar Photovoltaic Systems Using Artificial Neural Networks" (2019). Electronic Theses and Dissertations. 1640.
https://digitalcommons.du.edu/etd/1640
Copyright date
2019
Discipline
Engineering