分布式主客观证据融合模型及其应用

为了有效地提高油田剩余油分布预测的准确率和可靠性,通过BP神经网络联合模型与两级D-S证据推理模型的优势互补进行主客观证据融合,实现了剩余油分布多属性特征的准确分类．提出了将BP神经网络分类结果的可信度及专家系统推理结论的可信度作为D-S证据推理模型输入证据基本概率赋值的有效方法．实际应用与结果分析表明了上述方法的有效性,为各类多源信息融合系统的研究和工程实现提供了示例、途径和有益的经验．

Distributed Fusion Model of Subjective Evidence Fusing Objective Evidence and Its Application

In order to improve effectively the accuracy rate and reliability of remaining oil distribution forecast in oil field,a new model is proposed,which utilizes the merits of BP neural network combination model and two-level D-S evidence reasoning model and avoids their demerits.The exact classification with many characte- ristics is implemented about remaining oil distribution.An effective method is proposed,namely,the classifica- tion output reliability of each BP neural network and the reasoning result reliability of each domain expert system are regarded as basic probability assignment of input evidence in D-S evidence reasoning model.Practical appli- cation and result analysis on system show that the proposed model is effective and can be applied widely.The re- suits provided examples,approaches and useful experiences to the research of multiple sources information fu- sion system in different types.