Quantifying Solar Irradiance Variability Using Self-Organizing Map (SOM) Method

Chua, Shu Yuan (2017) Quantifying Solar Irradiance Variability Using Self-Organizing Map (SOM) Method. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Quantifying Solar Irradiance Variability Using Self-Organizing Map (SOM) Method.pdf - Submitted Version Download (384kB)

Abstract

Among the renewable energy generation in Malaysia, the use of solar energy has increased exponentially throughout the years. The Solar PV generation depends on seasonal variation and clouding which causes variability in the output. The current solution is the discharge of electrical energy from battery storage. However, battery storage is expensive and has a life expectancy. The research here focuses on proposing a classification scheme which can be used for forecasting. Solar irradiance data is clustered into groups for further analysis. The scheme involves the use of unsupervised learning, the self-organizing map (SOM) to cluster the variability profile using solar PV system data from FKE, UTeM. The MATLAB software will be used for SOM simulation. The data used is taken from January to December 2016. The data was tested by varying the map size. After running a few map sizes, the map size 35x35 was chosen. It had a quantization error of 0.844 and error percentage of 10.929%. From the 35x35 map size, a total of 186 days were clustered. Out of the 186 days, 104 days were grouped into 6 categories of variability. The 6 categories were clear sky, high value variability, low value variability, morning to noon variability, noon to evening variability, and overcast. The days clustered were compared to manual clusters and the self-organizing map (SOM) was determined to be highly accurate in clustering high variability irradiance values.

Actions (login required)

View Item