论文部分内容阅读
极限学习机是近年来提出的一种前向单隐层神经网络训练算法,具有训练速度快、不会陷入局部最优等优点,但其性能会受到随机选取的输入权值和阈值的影响.针对这一问题,提出一种基于多目标优化的改进极限学习机,将训练误差和输出层权值的均方最小化同时作为优化目标,采用带精英策略的快速非支配排序遗传算法对极限学习机的输入层到隐层的权值和阈值进行优化.将该算法应用于摇摆工况下自然循环系统不规则复合型流量脉动的多步滚动预测,分析了训练误差和输出层权值对不同步长预测效果的影响.仿真结果表明,优化极限学习机预测误差可以用较小的网络规模获得很好的泛化能力.为流动不稳定性的实时预测提供了一种准确度较高的途径,其预测结果可以作为核动力系统操作员的参考.
Extreme learning machine is a forward single hidden layer neural network training algorithm proposed in recent years, which has the advantages of fast training speed and not falling into the local optimum, but its performance will be affected by randomly selected input weights and thresholds. To solve this problem, an improved limit learning machine based on multi-objective optimization is proposed, which minimizes the mean square of training errors and output layer weights simultaneously. At the same time, it uses the fast non-dominated ranking genetic algorithm with elitism strategy to limit learning machine To the hidden layer weights and thresholds.The algorithm is applied to multi-step rolling prediction of irregular complex flow-pulsation of natural circulation system under rocking condition, and the error of training and output layer weight to unsynchronized Long prediction effect.The simulation results show that the prediction error of the ELM can be well generalized with a smaller network size.It provides a more accurate approach for the real-time prediction of the flow instability, Its predictions can serve as a reference for nuclear power system operators.