基于神经网络的A330多因素油耗模型研究

发布时间：2018-02-20 07:30

本文关键词： 油耗模型 BP神经网络影响结构分析平均影响值算法敏感度分析法　出处：《中国民航大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着民航运输业的快速发展以及国际社会对环境问题的关注,航空运输节能减排已成为热点。因此,建立油耗模型,准确预测航线飞行油耗量,是节油的重要环节之一。由于影响航线油耗的因素众多且内在关联紧密,相对预测变量航线飞行油耗量呈显著非线性,其难点在于表达各影响因素间的强耦合关系,关键在于模型的预测精度。基于以上问题,本文的主要工作是:首先,对影响航线油耗的关键因素进行研究,并完成QAR数据的飞行阶段划分和参数提取工作。然后,采用两种方法构建了基于BP神经网络的各阶段油耗模型。第一种建模方法:以各阶段影响参数为输入,各阶段油耗量为输出,构建了一个三层BP网络油耗模型。第二种建模方法:以单参数油耗贡献为加强因子的基础上,采用重入设计方法,构建了一种新型的BP网络油耗模型。最后,基于模型采用平均影响值算法进行影响结构分析,给出各影响因素对相应阶段油耗量的影响方向和影响程度排序。接着,利用敏感度分析法对各阶段影响因素进行分析并量化其影响大小,进而确定了各因素微小变化对相应阶段油耗量的贡献程度。实验结果表明:实验建立的油耗模型精度均在自主加油量范围之内,可为实际航线飞行油耗估计提供技术支持;平均影响值算法和敏感度分析法对影响因素的分析并量化对航线油耗影响因素评估具有参考价值。
[Abstract]:With the rapid development of civil aviation transportation and the attention of the international community to the environmental problems, the energy saving and emission reduction of air transportation has become a hot spot. It is one of the important links of fuel saving. Because there are many factors affecting the fuel consumption of the route and the internal relationship is close, the relative prediction variable flight fuel consumption is significantly nonlinear, the difficulty of which is to express the strong coupling relationship among the influencing factors. The key is the prediction accuracy of the model. Based on the above problems, the main work of this paper is as follows: first, the key factors affecting the fuel consumption of the route are studied, and the flight stage division and parameter extraction of the QAR data are completed. Based on BP neural network, the oil consumption model of each stage is constructed by two methods. The first modeling method is based on the input of each stage's influence parameter and the output of each stage's oil consumption. A three-layer BP network oil consumption model is constructed. The second modeling method: based on the single parameter fuel consumption contribution as a strengthening factor, a new BP network oil consumption model is constructed by using the re-entry design method. Finally, Based on the model, the average influence value algorithm is used to analyze the influence structure, and the influence direction and degree of the influence factors on the oil consumption in the corresponding stage are given. The influence factors of each stage are analyzed by sensitivity analysis method and the magnitude of the influence is quantified. The experimental results show that the accuracy of the fuel consumption models established by the experiment is within the range of self-refueling quantity, which can provide technical support for the actual flight fuel consumption estimation. The algorithm of average influence value and sensitivity analysis have reference value to the analysis of influencing factors and quantification to the evaluation of influencing factors of airline fuel consumption.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V323;TP183

【参考文献】