Abstract:  In order to improve the convergence speed and generalization ability of BP neural network and prevent it from falling into local optimal value, the traditional particle swarm optimization algorithm is improved in three aspects based on the previous research, including the limit of the maximum speed, the changes of the inertia weight factor and the improvement of the fitness function. Then it is used to optimize the weight and threshold of the BP neural network. And the dynamic prediction on reservoir parameter is realized by the improved PSO-BP neural network, the whole process is determining the input and output neurons, quantitating the time parameter, constructing the neural network model with the training samples and testing it. Finally, the simulation results of the average training error is analyzed, and it proves that the convergence and generalization ability of the improved PSO-BP algorithm are better than the BP algorithm and PSO-BP algorithm.

Key words: improved PSO-BP neural network, inertia weight factor, reservoir parameter, predication

摘要： 为提高BP神经网络的收敛速度和泛化能力，防止其陷入局部最优值，在前人工作基础上对传统粒子群算法进行了改进，具体包括：设定最大限制速度、改变惯性权重因子和改进适应度函数，并把改进粒子群算法应用于BP神经网络权值和阈值的优化。之后利用改进粒子群算法优化的BP神经网络实现对储层参数的动态预测，具体步骤为：确定神经网络的输入、输出神经元，定量化时间参数[T]，利用训练样本构建神经网络模型并进行检验。最后通过平均训练误差对仿真过程进行分析，结果表明改进PSO-BP算法的收敛性与泛化能力均优于BP算法和PSO-BP算法。

关键词: 改进PSO-BP神经网络, 惯性权重因子, 储层参数, 预测