石油とガスの研究

抽象的な

Prediction of Solar Radiation using hybrid algorithm of Random Forest and Particle Swarm Optimization

Sunayana Gupta

Due to increased pollution, greenhouse effect and global warming
resulting from power production using fossil fuels, there
is increased penetration of renewable energy sources into the
power production system. Over the last few years, solar radiation
has become a significant means of power production
using solar panels and the concept of microgrids has made
solar power an indispensable source of power in the distribution
system. The power production using solar energy is highly
variable and weather dependent which creates a power imbalance
into the system when it is penetrated without forecasting.
Therefore, solar power prediction plays a critical role in the
proper usage of solar energy while keeping the system stable.
For automating the power system the forecast needs to be very
accurate and thus, it is needed to improve the existing forecasting
techniques. In this study, we have proposed a solar radiation
scheme based on various meteorological factors, including
temperature, humidity, wind speed, and others and used this
data for building a machine learning model for prediction. We
introduced a hybrid model for prediction which optimizes the
parameters of Random Forest using Particle Swarm Optimization
technique. The results show empirically that the hybrid
RF-PSO model significantly improves the prediction accuracy
and reduces the MAE error.