Abstract: Dempster-Shafer（D-S） evidence theory is an effective mathematical tool for modeling and processing uncertain information without considering prior probability. However, when there is a high degree of conflict between the two bodies of evidence, Dempster’s combination rule will produce unreasonable results. To solve this problem, a multi-sensor data fusion method based on divergence difference measure and belief entropy is proposed. Firstly, a new distance based on the trigonometric divergence of single focal element is proposed to measure the distance between two bodies of evidence and calculate the reliability of each evidence. Then, in the framework of D-S evidence theory, a novel belief entropy measure is proposed to quantify the uncertain information contained in each evidence. Thereafter, the weight of each evidence is calculated by integrating the credibility and uncertainty of each evidence, and the weighted average evidence is obtained. Finally, Dempster’s combination rule is used to fuse the weighted average evidence to obtain the final fusion result. The proposed method is used to solve the practical application of transformer on-line monitoring fault diagnosis. Experimental results show that the proposed method has faster convergence speed and higher diagnosis accuracy, which is superior to other methods.

Key words: multi-sensor fusion algorithm, D-S evidence theory, belief entropy, triangulation divergence measurement, evidence conflict

摘要： Dempster-Shafer（D-S）证据理论是在不考虑先验概率的情况下建模和处理不确定信息的有效数学工具。当两证据之间高度冲突时，Dempster组合规则会产生不合理的结果。针对这一问题，提出了一种基于散度差异值测度和信念熵的多传感器数据融合方法。提出了一种距离公式-单焦元三角散度来衡量两证据之间的距离，并由此计算出各证据的可信度。在D-S证据理论框架中提出了一种信念熵测度来度量各证据中包含的不确定信息，综合各证据的可信度和不确定度来计算各证据的权值，由此得到加权平均证据。利用Dempster组合规则融合加权平均证据得到最终融合结果。利用所提出的方法来解决变压器在线监测故障诊断实际应用问题，实验结果表明，提出的方法具有较快的收敛速度和较高的诊断精度，优于其他方法。

关键词: 多传感器融合算法, D-S证据理论, 信念熵, 三角散度测量, 证据冲突