Power generation from photovoltaic systems has been the focus of many research studies. The main aim has been to tackle environmental and financial issues of typical power resources. Unstable fossil fuel prices and a considerable portion of environmental pollutions and greenhouse emissions are major concerns when it comes to industrialized countries. Ability to produce electricity for a long time with minimal maintenance and reduction in capital costs are factors to be considered while integrating photovoltaic systems into the electrical grid.
Unfortunately, output power of photovoltaic system is dependent on ambient temperature and irradiation level. In addition, fluctuations in the output power could be experienced owing to shadowing or power quality interference. Therefore, it is important to study the effect of these output power fluctuations before photovoltaic systems installation. To achieve this, simulation implementing historical data and extensive analysis should be done.
Handling this data is computationally expensive and time consuming. Therefore, developing solutions that can ease the burden of extensive studies and simulations relating to integrating photovoltaic systems into the electrical grid is of outmost importance. Clustering methods can be used to group photovoltaic power patterns with similar properties. Thus, a representative power pattern for every group can be integrated in the simulations.
Amr Munshi and Yasser Mohamed from the University of Alberta presented the outcomes of an in-depth analysis of Bat clustering algorithms based on a number of objective functions in a bid to establish the grouping mechanism of photovoltaic power patterns. Their main objective was to enhance the clustering formation of the former clustering algorithm, Bat J. Their research work is published in Solar Energy.
The authors performed and in-depth analysis of the performance of Bat clustering algorithms dictated by a number of integrated objective functions to validate the clustering of photovoltaic power pattern process. They then compared the performance of the K-means and Bat J clustering algorithms with the new Bat clustering on the clustering process of photovoltaic power patterns data.
The researchers also illustrated the original Bat clustering algorithm methods to undertake photovoltaic power patterns grouping. They adopted the principle component analysis in a bid to reduce the dimensionality of the photovoltaic power patterns data.
Bat clustering algorithms were comparable or surpassed K-means in the validity index, compactness and separation values. The within-cluster-sum-of-squares validity index values of Bat were observed to have improved as opposed to K-means by approximately 14.10% and 14.71% over the knee-points for the first and second datasets, respectively. The authors observed that Bat within-cluster-sum-of-squares posted the best outcomes and was capable of enhancing Bat J algorithm that exhibited the best cluster data.
Nevertheless, this corresponded to more complexity since the number of parameters ought to have been priori calibrated. The preferable combination presenting the optimum number of clusters was observed to be Bat within-cluster-sum-of-squares clustering and within-cluster-sum-of-squares validity index. They presented considerably high separated and compact clusters.
Lower within-cluster-sum-of-squares values at a selected partition presented the most preferable combination of separation and compactness. Therefore, Munshi and Mohamed study on the Bat within-cluster-sum-of-squares could offer well-defined photovoltaic power pattern clusters as well as cluster representatives that can be used in photovoltaic output power analyses.
Amr A. Munshi and Yasser A.-R.I. Mohamed. Comparisons among Bat algorithms with various objective functions on grouping photovoltaic power patterns. Solar Energy, volume 144 (2017), pages 254–266.Go To Solar Energy