摘要
为提高库区岸坡变形的非线性预测精度,提出利用极限学习机构建库区岸坡的非线性预测模型。首先,利用逐步试算法优化极限学习机的激励函数和隐层神经元数;其次,采用Rosenstein算法评价边坡变形序列的混沌特性,利用空间重构来实现极限学习机的混沌优化,进而构建混沌优化ELM模型。分析表明,不同实例的最优网络参数具有差异,通过逐步试算法能很好地确定最优参数;库岸边坡的变形序列均具有混沌特性,通过混沌理论的空间重构优化,能有效提高预测精度,且预测结果较传统神经网络具有较大的优越性。
In order to improve the nonlinear prediction accuracy of bank slope deformation in reservoir area,a nonlinear prediction model of bank slope is proposed based on extreme learning machine.Firstly,the stepwise trial ealculation method is used to optimize the excitatibn function and the number of hidden layer neurons in extreme learning machine.Secondly,the Rosenstein algorithm is used to evaluate the chaotic characteristics of slope deformation sequence,and the space reconstruetion is used to realize the chaos optimization of extreme learning machine and then the chaos optimization ELM model is constructed. The ease analysis shows that,(a)the optimal network parameters of different examples are different,and the optimal parameters can be well determined by stepwise trial calculation;and (b)the deformation sequences of bank slope are all chaotic,and the prediction accuracy can be effectively improved and the prediction results are more effective than traditional neural network by the spatial reconstruction of chaos theory.
作者
张志会
ZHANG Zhihui(State Nuclear Electric Power Planning Design &Research Institute Co.,Ltd.,Beijing 100095,China)
出处
《水力发电》
北大核心
2018年第12期39-42,104,共5页
Water Power
关键词
库岸边坡
变形预测
极限学习机
逐步试算法
混沌理论
bank slope
deformation prediction
extreme learning machine
stepwise trial algorithm
chaos theory
