Solar Radiation Time Series Prediction release_vetwrw76lncizdeyygatnuhwgy

Abstract

A model was constructed to predict the amount of<br> solar radiation that will make contact with the surface of the earth in<br> a given location an hour into the future. This project was supported<br> by the Southern Company to determine at what specific times during<br> a given day of the year solar panels could be relied upon to produce<br> energy in sufficient quantities. Due to their ability as universal<br> function approximators, an artificial neural network was used to<br> estimate the nonlinear pattern of solar radiation, which utilized<br> measurements of weather conditions collected at the Griffin, Georgia<br> weather station as inputs. A number of network configurations and<br> training strategies were utilized, though a multilayer perceptron with<br> a variety of hidden nodes trained with the resilient propagation<br> algorithm consistently yielded the most accurate predictions. In<br> addition, a modeled direct normal irradiance field and adjacent<br> weather station data were used to bolster prediction accuracy. In later<br> trials, the solar radiation field was preprocessed with a discrete<br> wavelet transform with the aim of removing noise from the<br> measurements. The current model provides predictions of solar<br> radiation with a mean square error of 0.0042, though ongoing efforts<br> are being made to further improve the model's accuracy.
In text/plain format

Archived Files and Locations