有监督SOM神经网络在油气预测中的应用

Application of the supervised SOM neural network to oil and gas prediction

用于油气预测的方法基本上可分成两类：一类是有监督预测方法，另一类是无监督预测方法。80年代传统模式识别方法（统计、句法和模糊模式识别）得到广泛应用；90年代以来神经网络理论异军突起，以BP为代表的有监督神经网络和以SOM为代表的无监督神经网络广泛应用于油气预测。本文介绍了应用有监督SOM神经网络进行油气预测的方法原理，经两个工区的实际资料试算结果证明本方法性能良好，可以成为油气预测的一种可选方法。

The methods used for oil and gas prediction can be classified as two kinds on the whole. The one kind is the supervised prediction method, the other kind is the non-supervised prediction method. In 1980s, the conventional pattern recognition method (statistical, syntactic, and fuzzy pattern recognition) found extensive application. Since 1990s the neural network theory comes to the fore suddenly.The supervised neural network taking BP as representative and non-supervised neural network taking SOM as representative have been used widely in oil and gas prediction. In the paper, we introduce the principle of oil and gas prediction using supervised SOM neural network. The real data trial computation results for two work areas prove that the method behaves well and can be an alternative for predicting oil and gas.