摘要
针对传统径向基神经网络(RBF)在大坝安全监测应用中易陷入局部最优及预测精度不高的问题,引入粒子群算法(PSO),对输入的大坝安全监测数据进行初步的聚类处理,找出初步聚类中心后令其为PSO的初值,根据运算法则更新初值以寻求适合训练数据的最优基函数中心。以小湾大坝为例,应用Matlab仿真模拟计算了大坝变形量,结果表明PSO-RBF与传统RBF的拟合效果都很好,PSO-RBF预测准确度更高。
The standard RBF neural network is easy to jump into local optimum and has low forecasting accuracy for dam safety monitoring. The particle swarm algorithm is introduced to cluster the dam monitoring data and find the cluste ring centers, which can be defined as initial values of PSO. Then the initial values are updated to find the optimal basis function center of training data based on the operational rule of PSO. Taking Xiaowan dam for an example, the dam deformation monitoring is simulated with Matlab software. The results show that both of the PSO-RBF and standard RBF have good fitting effect while PSO-RBF has higher prediction accuracy.
出处
《水电能源科学》
北大核心
2012年第8期77-79,共3页
Water Resources and Power
基金
国家自然科学基金资助项目(50909041
51079046
51079086
51139001)
河海大学水文水资源与水利工程科学国家重点实验室专项基金资助项目(2009586012
2009586912
2010585212)
关键词
大坝安全监测
聚类算法
径向基函数神经网络
粒子群算法
小湾大坝
dam safety monitoring
clustering algorithm
radial basis function neural network
particle swarm optimi-zation
Xiaowan dam
