针对在线支持向量回归(online support vector regression,online SVR)算法进行复杂时间序列精确预测时效率较低的问题,提出一种改进减量训练策略的快速预测方法,通过对非支持向量样本的采样选择,采取加速减量训练实现对在线训练数据集...针对在线支持向量回归(online support vector regression,online SVR)算法进行复杂时间序列精确预测时效率较低的问题,提出一种改进减量训练策略的快速预测方法,通过对非支持向量样本的采样选择,采取加速减量训练实现对在线训练数据集规模的缩减,从而达到快速在线训练和预测的目的.将该算法应用于黑龙江移动通信话务量数据的预测中,实验结果表明,在保持OnlineSVR预测精度的条件下,算法执行效率得到大幅提高。展开更多
Energy optimization is one of the key problems for ship roll reduction systems in the last decade. According to the nonlinear characteristics of ship motion, the four degrees of freedom nonlinear model of Fin/Rudder r...Energy optimization is one of the key problems for ship roll reduction systems in the last decade. According to the nonlinear characteristics of ship motion, the four degrees of freedom nonlinear model of Fin/Rudder roll stabilization can be established. This paper analyzes energy consumption caused by overcoming the resistance and the yaw, which is added to the fin/rudder roll stabilization system as new performance index. In order to achieve the purpose of the roll reduction, ship course keeping and energy optimization, the self-tuning PID controller based on the multi-objective genetic algorithm （MOGA） method is used to optimize performance index. In addition, random weight coefficient is adopted to build a multi-objective genetic algorithm optimization model. The objective function is improved so that the objective function can be normalized to a constant level. Simulation results showed that the control method based on MOGA, compared with the traditional control method, not only improves the efficiency of roll stabilization and yaw control precision, but also optimizes the energy of the system. The proposed methodology can get a better performance at different sea states.展开更多
文摘针对在线支持向量回归(online support vector regression,online SVR)算法进行复杂时间序列精确预测时效率较低的问题,提出一种改进减量训练策略的快速预测方法,通过对非支持向量样本的采样选择,采取加速减量训练实现对在线训练数据集规模的缩减,从而达到快速在线训练和预测的目的.将该算法应用于黑龙江移动通信话务量数据的预测中,实验结果表明,在保持OnlineSVR预测精度的条件下,算法执行效率得到大幅提高。
基金Foundation item： Supported by the National Natural Science Foundation of China （Grant No. 61174047） and the Fundamental Research Funds for the Central Universities （HEUCF041406）.
文摘Energy optimization is one of the key problems for ship roll reduction systems in the last decade. According to the nonlinear characteristics of ship motion, the four degrees of freedom nonlinear model of Fin/Rudder roll stabilization can be established. This paper analyzes energy consumption caused by overcoming the resistance and the yaw, which is added to the fin/rudder roll stabilization system as new performance index. In order to achieve the purpose of the roll reduction, ship course keeping and energy optimization, the self-tuning PID controller based on the multi-objective genetic algorithm （MOGA） method is used to optimize performance index. In addition, random weight coefficient is adopted to build a multi-objective genetic algorithm optimization model. The objective function is improved so that the objective function can be normalized to a constant level. Simulation results showed that the control method based on MOGA, compared with the traditional control method, not only improves the efficiency of roll stabilization and yaw control precision, but also optimizes the energy of the system. The proposed methodology can get a better performance at different sea states.
