Solar Power Prediction for Smart Community Microgrid

Urban areas host more than 50% of the world's populations, are responsible for 75% of energy consumption in the world, and they emit almost 80% of global carbon dioxide. There is an urgent need to develop "low carbon" cities that are smart and efficient and use renewable energy to foster the growth of the green economy. Smart grids are being developed to tackle these challenges through integration of renewable and green energy as well as energy efficiency. They are moving toward a concept of networked microgrids. Microgrids will enable the integration of distributed renewable energy such as roof top solar panels within smart city communities. For these microgrids to operate reliably and efficiently, prediction algorithms are important because of the fluctuation of solar energy and its dependence on weather. Prediction of energy is a component of microgrids energy management systems to optimize their operation. This paper presents a machine learning based algorithm, which learns a regression tree model with time of the day and humidity as main parameters. The regression tree model presents a promising accuracy. This work shows that solar panel prediction in Houston is heavily dependent on humidity of the region.