Study of forecasting solar irradiance using neural networks with preprocessing sample data by wavelet analysis

摘要: Artificial neural network is a powerful tool in the forecast of solar irradiance. In order to gain higher forecasting accuracy, artificial and wavelet analysis have been combined develop new method this paper, data sequence irradiance as samples mapped into several time-frequency domains using transformation, recurrent back-propagation (BP) established for each domain. The forecasted equals algebraic sum components, which were predicted correspondingly by networks, all domains. A discount coefficient adopted updating weights biases networks so that late forecasts play more important roles. On basis principle combination analysis, model completed fore-casting Based on historical day-by-day records Shanghai an example total presented. results indicate makes much accurate than without with analysis.