﻿ 基于深度学习的船舶航线数据趋势性预测方法
 舰船科学技术  2023, Vol. 45 Issue (10): 164-167    DOI: 10.3404/j.issn.1672-7649.2023.10.033 PDF

1. 北部湾大学，广西 钦州 535011;
2. 南宁学院 智能制造学院，广西 南宁 535200

Optimization of trend prediction method for ship route data based on deep learning
WEI Li-lan1, QIN Jian-bo2
1. Beibu Gulf University, Qinzhou 535011, China;
2. Intelligent Manufacturing College, Nanning University, Nanning 530200, China
Abstract: By predicting the trend of ship route data to achieve intelligent route scheduling, a deep learning based method for predicting the trend of ship route data is proposed. Constructing a combination of centralized access and distributed sensing to achieve sampling and fusion processing of ship route data, using load balancing scheduling and data block grouping feature distance methods for trend related feature detection of ship route data, and using deep learning and time series label feature detection methods to achieve distribution reconstruction of multiple time series of ship route data, According to the reconstruction results and cluster analysis trend, the trend prediction of ship route data is realized. The simulation results show that this method has good clustering performance and high prediction accuracy in predicting the trend of route data.
Key words: deep learning     warship     route data     trendliness     prediction     clustering
0 引　言

1 数据采样和特征分析 1.1 船舶航线数据采样

 图 1 船舶航线数据分布结构模型 Fig. 1 Distribution structure model of ship route data

 ${x_{ij}}^\prime = {x_i} - \rho + j\frac{{2\rho }}{n} 。$ (1)

 $e(x) = \sum\limits_i {{d_{ei}}} {D_s}B(x).{d_e} 。$ (2)

 $B = \left[ {\begin{array}{*{20}{c}} { - 1}&0&0&{ - 1}&0&0 \\ 0&{6\xi /L}&{3\xi - 1}&0&{6\xi /L}&{3\xi - 1} \end{array}} \right] 。$ (3)

 $d\left( t \right) = \left\{ {\begin{array}{*{20}{c}} {{\rm{arctan}}\left( {\dfrac{{X_2' \left( t \right)}}{{X_1' \left( t \right)}}} \right),} & {X_1' \left( t \right) > 0}，\\ {{\rm{arctan}}\left( {\dfrac{{X_2' \left( t \right)}}{{X_1' \left( t \right)}}} \right) + \pi } ，& {X_1' \left( t \right) < 0,t = 1,2,\cdots , T}，\\ {\pi /2}，& {X_1'\left( t \right) = 0}。\end{array}} \right.$ (4)

1.2 船舶航线数据聚类处理

 $\begin{split} {d_{m + 1}}(m) =& {d_{k + 1}}(m) \pm \\ &\sqrt {{{({d_m}(0){e^{{\lambda _1}}} + 1)}^2} - \sum\limits_{i = 1}^{m - 1} {{{[{d_{m + 1}}(i) - {d_{k + 1}}(i)]}^2}} } \end{split} 。$ (5)

 $K = \left(\frac{1}{2}(u_A^ + - u_A^ - ) - \frac{1}{2}{F_{{A_1}}},\frac{1}{2}(u_A^ + - u_A^ - ) - \frac{1}{2}({F_B} + {F_{{A_2}}})\right) 。$ (6)

2 船舶航线数据趋势性预测 2.1 预测深度学习模型

 ${d_i} = \left[ {\begin{array}{*{20}{c}} {{k_b}}&{ - {k_b}} \\ { - {k_b}}&{{k_b} + {k_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{x_b}} \\ {{x_b}} \end{array}} \right\} + \left[ {\begin{array}{*{20}{c}} {{m_b}}&{} \\ {}&{{m_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{u_b}} \\ {{u_s}} \end{array}} \right\}{d_{ei}} 。$ (7)

 $p\left( r \right) = \frac{1}{{\sqrt {2\pi } {\sigma _s}}} = - 2E{\left({X_1}\cdot\frac{{\partial H}}{{\partial \tau }}\right)^{\rm{T}}} 。$ (8)

 $\begin{split} & \left[ {\begin{array}{*{20}{c}} {{m_b}}&{} \\ {}&{{m_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{{\ddot x}_b}} \\ {{{\ddot x}_s}} \end{array}} \right\} + \left[ {\begin{array}{*{20}{c}} {{c_b}}&{ - {c_b}} \\ { - {c_b}}&{{c_b} + {c_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{{\ddot x}_b}} \\ {{{\ddot x}_s}} \end{array}} \right\} + \\ & \left[ {\begin{array}{*{20}{c}} {{k_b}}&{ - {k_b}} \\ { - {k_b}}&{{k_b} + {k_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{x_b}} \\ {{x_b}} \end{array}} \right\} + \left[ {\begin{array}{*{20}{c}} {{m_b}}&{} \\ {}&{{m_s}} \end{array}} \right]\left\{ {\begin{array}{*{20}{c}} {{u_b}} \\ {{u_s}} \end{array}} \right\} = 0 。\end{split}$ (9)

2.2 船舶航线数据重组与预测

 $F = {X_2} - {X_1}\cdot H = \min \left(\sum\limits_i^N {{R_i}} \right) = \left\{ \begin{array}{cc} \dfrac{{{s_{ij}} - s(i,j)}}{{{s_{ij}}}}，& s(i,j) < {s_{ij}} ，\\ e(i,j)，& s(i,j) \geqslant {s_{ij}} 。\end{array} \right.$ (10)

 \begin{aligned} {\tau _{k + 1}} =&{\tau _k} + \mu ( - {{\hat G}_k}) = {t_{ACK}} + \\ &\sum\limits_{i = 0}^N {\left( {DIFS + C_R^{\left( i \right)} \times {t_{slot}} + {t_{DATA}} + SIFS + {t_{T - start}}} \right)}。\\ \end{aligned} (11)

3 仿真实验与结果分析

 图 2 船舶航线数据采样 Fig. 2 Sampling of ship route data

 图 3 船舶航线数据趋势性预测收敛曲线 Fig. 3 Convergence curve of trend prediction for ship route data

 图 4 舰船航线数据趋势预测精度对比 Fig. 4 Comparison of trend prediction accuracy for ship route data
4 结　语

 [1] SALINAS A, MORENO-VALENZUELA J, KELLY R. A family of nonlinear PID-like regulators for a class of torque-driven robot manipulators equipped with torque-constrained actuators[J]. Advances in Mechanical Engineering, 2016, 8(2): 1-14. [2] 代森强, 贺鹏. 水面舰队防空配置问题研究[J]. 信息工程大学学报, 2016, 17(5): 613-616. DAI Senqiang, HE Peng. Research on air defense configuration of surface fleet[J]. Journal of information engineering University, 2016, 17(5): 613-616. DOI:10.3969/j.issn.1671-0673.2016.05.018 [3] 蔡芝明, 金家善, 陈砚桥. 舰艇编队备件携带方案的混合约束问题研究[J]. 海军工程大学学报, 2016, 28(5): 80-85+111. CAI Zhiming, JIN Jiashan, CHEN Yanqiao. Research on the mixed constraint problem of spare parts carrying scheme in ship formation[J]. Journal of Naval Engineering University, 2016, 28(5): 80-85+111. [4] 岳光, 潘玉田, 张华君, 等. 多神经元神经网络算法的DSP无人侦察车伺服控制系统[J]. 北京理工大学学报, 2019, 39(2): 203-208. YUE Guang, PAN Yutian, ZHANG Huajun, et al. DSP unmanned reconnaissance vehicle servo control system based on multi neuron neural network algorithm[J]. Journal of Beijing University of Technology, 2019, 39(2): 203-208. DOI:10.15918/j.tbit1001-0645.2019.02.016 [5] 龙腾, 刘震宇, 史人赫, 等. 基于神经网络的防空武器目标智能分配方法[J]. 空天防御, 2021, 4(1): 1-7. LONG Teng, LIU Zhenyu, SHI Renhe, et al. Neural network-based intelligent allocation method for air defense weapon targets[J]. Aerospace Defense, 2021, 4(1): 1-7. DOI:10.3969/j.issn.2096-4641.2021.01.001