Abstract: A quantum neural networks(QNN) method for forecasting oil recovery ratio in molecule deposition oil-displacement was proposed on the basis of the information processing of biology neuron and quantum computing theory.A quantum neuron compositing weighting,aggregating,activating and prompting parts was constructed and used to develop a three-step quantum neural networks model.Both the input and output of the model are real vectors.The linked weight and activation values are qubits and updated by the quantum rotation gates.According to the gradient descent algorithm,a learning algorithm of the QNN was developed.The availability of the model applied to forecast oil recovery rate of molecule deposition oil-displacement showed that the convergence rate and functional capacity of this model are prior to those of the normal three-step BP network.

Key words: quantum neural networks, quantum neuron, learning algorithm, molecule deposition oil-displacement, oil recovery, forecast method

摘要： 提出一种用于预测分子沉积(MD)膜驱原油采收率的量子神经网络方法。基于生物神经元信息处理机制和量子计算原理构造出一种量子神经元,该神经元由加权、聚合、活化、激励四部分组成。再由量子神经元构造出三层量子神经网络模型,其输入和输出为实值向量,权值和活性值为量子比特。权值和活性值调整由量子门实现。基于梯度下降法构造了该模型的学习算法。将该模型应用于MD膜驱原油采收率的预测实验结果表明,该模型在收敛速度和泛化能力方面明显优于普通三层BP网络。

关键词: 量子神经网络, 量子神经元, 学习算法, 分子沉积膜驱油, 原油采收率, 预测方法