改进天牛群搜索算法及其在船舶纵摇运动预测中的应用

Improved Beetle Swarm Optimization and Its Application In Ship Pitching Motion Prediction

【目的】针对天牛群搜索算法易陷入局部最优及搜索精度较低等缺陷,提出一种基于二阶振荡自适应变异的天牛群搜索算法。【方法】该算法引入二阶振荡环节增加算法的全局探索能力和局部开发能力。采用正余弦搜索思想对天牛个体进行位置更新,使得天牛个体可充分的利用自身与最优位置的差值信息逐渐趋近最优解。同时引入自适应t分布变异算子来增加种群的多样性,避免算法陷入局部最优。【结果与讨论】对高维单峰函数、高维多峰函数的仿真实验结果表明,改进的算法有效地提高其搜索精度、收敛速度及鲁棒性,克服其易陷入局部最优的缺陷。将改进天牛群算法应用于BP神经网络权值及阈值优化对船舶纵摇运动姿态进行预测,并于BP网络、BSO-BP网络及PSO-BP网络的预测结果进行比较,精度分别提升85.7%、74.6%和77.2%。改进天牛群搜索算法在实际工程应用中具有一定的优越性。

【Objective】To resolve the shortcomings of the beetle swarm optimization(BSO)easily fall into local optimum and the low accuracy in search,a new BSO based on second-order Oscillatory adaptive mutation is proposed.【Method】Initially the algorithm introduces a second-order oscillation element to increase the algorithm's global exploration capabilities and local exploitation capabilities.The idea to use sine cosine search to update the position of each beetle so that we can make full use of the difference information between itself and the optimal position to gradually approach the optimal solution.At the same time,an adaptive t-distribution mutation operator is introduced to increase the diversity of the population and avoid falling into the local optimum for the algorithm.【Result and Conclusion】Through simulation experiments on high-dimensional unimodal functions and high-dimensional multimodal functions,the results show that the improved algorithm effectively improves its search accuracy,convergence speed and robustness,and overcomes its shortcomings of falling into local optimal.Finally,the improved beetle swarm optimization(IBSO)is applied to BP neural network weights and thresholds optimization to predict ship pitching motion attitude.The accuracy of algorithm is increased by 85.7%,74.6%and 77.2%respectively compared with the prediction results of BP network,BSO-BP network and PSO-BP network.The improved Beetle Swarm Optimization has a strong ability to find the best.