Solar power generation prediction by using k-nearest neighbor method

Abstract

The increasing of global energy demand by 2.1% in 2017 which is more than twice the previous year’s rate resulting in increasing of carbon dioxide emissions by 1.4% in the previous year after three years of remaining flat. Energy demand can be supplied by renewable energy which is more clean and help reducing carbon emissions. Solar energy has become the dominant renewable energy in Malaysia since it is situated at the equatorial region with an average solar radiation of 400-600 MJ/m2 per month. In this paper, factors that affected solar power generation are studied. All data from these factors are collected and the correlation analysis is done to determine which factor has strong correlation with solar power generation. The factors that have strong correlation with power generation will be used to predict solar power generation for next month. The results from this study showed that k-nearest neighbor method provides a better prediction result than artificial neural network since its root mean square error is the lowest value.The increasing of global energy demand by 2.1% in 2017 which is more than twice the previous year’s rate resulting in increasing of carbon dioxide emissions by 1.4% in the previous year after three years of remaining flat. Energy demand can be supplied by renewable energy which is more clean and help reducing carbon emissions. Solar energy has become the dominant renewable energy in Malaysia since it is situated at the equatorial region with an average solar radiation of 400-600 MJ/m2 per month. In this paper, factors that affected solar power generation are studied. All data from these factors are collected and the correlation analysis is done to determine which factor has strong correlation with solar power generation. The factors that have strong correlation with power generation will be used to predict solar power generation for next month. The results from this study showed that k-nearest neighbor method provides a better prediction result than artificial neural network since its root mean square err...