【作者】 邢明海； 陈祥光； 王渝；

【Author】 Xing Minghai Chen Xiangguang Wang Yu (Beijing Institute of Technology, Beijing 100081, China)

【机构】 北京理工大学；

【摘要】 准确预测油田未来原油产量对油田的开发和生产管理具有重要的指导意义。对于具有非线性、不确定和开放特性的多变量系统进行预测,使用传统的统计学方法或静态模型预测通常不能满足精度要求。这里把BP神经网络、模糊神经网络以及基于神经网络的组合预测方法应用于多变量系统。采用了两层预测系统:第一层包含两个神经网络;第二层是把第一层的两个网络输出进行组合。研究了3种不同的组合算法:平均法、最小平方回归法和前馈神经网络法。实验结果表明,采用组合方法比采用单一的预测方法的预测精度有了进一步的提高,特别是应用人工神经网络(即BPNN)的组合预测优于其他的预测方法,具有较好的适应性。更多还原

【Abstract】 The prediction results usually can’t satisfy the precision demands using the traditional statistics meth ods for nonlinear and uncertain multi-variable system. The artificial neural network methods are used to predict the multi-variable system, the combined forecast methods are used to predict the multi-variable complex system based upon the routine BP neural network and fuzzy neural network. By the training for the neural network, the connotative relation among the impact factors and forecast factors is established in order to better prediction results. The test’s results show that the prediction precision using the combined forecast methods are better than using the traditional statistics methods, it has smaller modeling workload, these methods are practical, flexible and steady. The forecast results are better using the combined methods than using the single method.更多还原