Abstract: Using porosity prediction method based on seismic multiattributes can comprehensively balance individual attribute parameters so that more objectively and effectively reflect the variation of porosity, develop the relationship between the well log porosity and the seismic attribute and determine the optimal type and amount of attribute by means of multielement regression, error analysis and cross-validation techniques, and finally establish the mapping relations between the attributes and well log porosity and predict the lateral and vertical distributions of porosity. This paper first applies this method to the eastern Sichuan A gas field and predicts the porosity distribution of its super-dense sandstone reservoir. And good result is gained.

Key words: porosity, prediction, multiattribute, neural network, multielement regression, cross validation