在油气藏的组分模型中，需要进行大量的相平衡计算。其中，平衡常数K值的计算是非常 关键的。根据相平衡公式精确计算K值需要大量的计算时间。提出采用BP神经网络来计算 K值，并根据不同的情况需要可以建立简化的或完整的BP网络。指出用BP网络计算K 值具有计算速度快、精度高的优点。

A compositional model usually performs a lot of phase equilibrium calculations. During this process, the calculation of equilibrium const K is a key point. Calculating K value according to the phase equilibrium equations takes a lot of time. This paper presents a fast algorithm of calculating K value using BP neural networks. The BP network could be complete or abbreviated depending on the situation. The result shows that using BP network to calculate K value can save a lot of calculation time.