Clustering Irradiance Values Using Unsupervised Machine Learning

Yeoh, Yee Jun (2019) Clustering Irradiance Values Using Unsupervised Machine Learning. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Clustering Irradiance Values Using Unsupervised Machine Learning.pdf - Submitted Version Download (527kB)

Abstract

Clustering is an unsupervised machine learning that works by splitting a large dataset into multiple distinctive groups. As a fast developing renewable energy, output of photovoltaic is prone to fluctuation due to some factors. The purpose of clustering solar irradiance is to determine the daily pattern of irradiance and possibly group those having similar profile. Through this grouping, we can use the clusters obtained as a precursor for solar energy forecasting. The focus of this project is on both Self organizing map(SOM) and K-Means clustering. SOM utilizes plots for visualization purpose and to aid in manual classification. K-Means, on the other hand, makes use of silhouette analysis, elbow analysis and gap statistics analysis to determine the number of cluster. With the help of MATLAB software, a series of supporting detail and evidence is produced with minimal issue. In preliminary clustering, both SOM and K-Means are able to show similarity in the outcome, leading to a high confidence conclusion. In final clustering, an additional software, Weka is used alongside Matlab utilising only K-Means. The final outcome is the same as preliminary result where the optimum number of cluster is three. Irradiance profiles are plotted for categorization consisting of “ clear sky”, “cloudy” and “overcast”.

Actions (login required)

View Item