基于不同采样率的短航程油耗估计

针对多因素影响下的短航程油耗呈现双峰分布,提出了使用高斯混合聚类(Gaussian mixture model,GMM)和随机森林(random forest,RF)相结合的方法对短航程油耗进行估计。该算法先使用GMM对短航程油耗数据聚类,得到两个不同形状的聚类簇。以不同的采样率对两个聚类簇进行采样,构造子数据集,并对每个子集使用回归树进行训练。将CART回归树并行得到RF用于短航程油耗估计。在同一机型和航线,不同的航班数据上进行对比实验,结果验证了所提算法的有效性。

Estimation of short-voyage fuel consumption based on different sampling rates

Abstract] In view of the bimodal distribution of short-voyage fuel consumption under the influence of multiple factors, this paper proposes a method combining Gaussian mixture clustering (GMM) and random forest (RF) to estimate the short-range fuel consumption. The algorithm first uses GMM to cluster short-range fuel consumption data to obtain clusters of two different shapes. After that, the two clusters are sampled at different sampling rates, the sample subset is constructed, and the regression tree is used for each subset. Finally, the CART decision tree is used in parallel to obtain random forest for short-range fuel consumption estimation. The comparison experiments were carried out on the same model and route, and different flight data. The results show the effectiveness of the proposed algorithm.