Photovoltaic Power Pattern Clustering Based on Conventional and Swarm Clustering Methods

  • Author / Creator
    Munshi, Amr A
  • Among renewable energy resources, solar energy is promising and has recently become an area of interest in research. Photovoltaic (PV) systems have the capability of converting solar energy into electrical power. The advances in PV technology, such as the reliability and the continuous reduction in capital costs, motivate the integration of PV systems into the electrical grid. The power output of PV systems is mainly influenced by the level of irradiation and ambient temperature. This leads to operational problems and instability in the power output generated from PV systems. Accordingly, the integration of these systems requires extensive study and simulations of lengthy historical data with sub-hourly time steps. However, dealing with such data is time consuming and computationally expensive. Photovoltaic power pattern (PVPP) clustering is fundamental in providing enhanced knowledge on the impacts of integrating PV systems into the electrical grid without extensive analysis and simulations. Therefore, this research aims to develop solutions that can reduce the burden of extensive studies and simulations related to the integration of PV systems into the electrical grid. This research investigates a set of clustering methods from different clustering categories to determine the optimum number of clusters and to produce cluster representatives for PVPP data. Furthermore, the introduction of bio-inspired swarm optimization methods, such as the Ant Colony and Bat methods in clustering power patterns is presented. For the purpose of clustering and achieving efficient cluster representatives, six clustering algorithms from five different clustering categories are involved: K-means from partitional clustering, Hierarchical Ward’s minimum variance (WMV) from agglomerative clustering, Fuzzy C-means (FCM) from fuzzy clustering, self-organizing maps (SOM) from neural network based algorithms, and Ant Colony and Bat from bio-inspired swarm optimization methods. In order to evaluate the clustering methods in a comprehensive manner, the following nine internal validity indices were employed: Davies Bouldin (DBI), Dunn, Silhouette (SI), Bayesian information criterion (BIC), Xie-Beni (XB), mean square error (J), clustering dispersion indicator (CDI), mean index adequacy (MIA), and ratio of within-cluster sum-of-squares to between-cluster variation (WCBCR). The clustering results show that swarm clustering methods are comparable to conventional methods. Moreover, the Bat method was the most efficient and outperformed the other clustering methods. Therefore, five Bat algorithms with various objective functions: Bat based on Davies Bouldin Index (Bat DBI), Bat based on Dunn index (Bat Dunn), Bat based on clustering dispersion indicator (Bat CDI), Bat based on mean index adequacy (Bat MIA), and Bat based on within-cluster sum-of-squares to between-cluster variation (Bat WCBCR) are proposed to enhance the clustering results. The clustering results on two data sets show that the bio-inspired swarm clustering algorithm Bat based on WCBCR, as an objective function, produces significantly highly separated and well-compacted clusters that can be utilized in PV system simulations. In order to test the efficiency of the produced PVPP representatives in PV system simulations, a short-term PV power prediction model is presented. The results of the prediction model verify the efficiency of the PVPP clustering methodology in PV system studies.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Electrical and Computer Engineering
  • Specialization
    • Computer Engineering
  • Supervisor / co-supervisor and their department(s)
    • Mohamed, Yasser A.-R. I. (Electrical and Computer Engineering)
  • Examining committee members and their departments
    • Musilek, Petr (Electrical and Computer Engineering)
    • Niu, Di (Electrical and Computer Engineering)