Methods of Forecasting Solar Radiation

Abstract

Extreme demands on the methods used for forecasting solar radiation has been the driving force behind the efforts to find the best method available. An extensive study of different techniques available was conducted. Methods studied in this research can be classified as time series and neural network approach. Time series approaches considered are autoregressive (AR), moving average (MA), autoregressive moving average (ARMA), and autoregressive integrated moving average (ARIMA). In neural network approaches, multi-layer perceptron networks are used. The error back-propagation learning algorithm is utilized in the training process. Comparison of methods and performance of different methods are presented in the result and discussion section of this chapter. The solar radiation data used were a collection of past data acquired throughout the US continent for 10 years period. These data were used to forecast future solar radiation based on the past trend observed from the database using both time series and neural network approaches. Finally, this chapter makes general comparison among the methods used and outlines some advantages and disadvantages of using the neural network approach.