摘要
在液压伺服系统动态流量软测量模型核心算法的研究中,综合考虑GA(Genetic Algorithm)和BFGS(Broyden-Fletcher-Goldfarb-Shanno)算法在最优化问题上的优势和不足,提出BFGS-GA.在GA中引入BFGS算子,对每一代中若干个精英个体以一定概率对其进行BFGS线性迭代运算.采用BFGS-GA对实数编码的染色体进行优化,得到最优的RBF(Radial Basis Function)网络结构和参数.实验结果表明该算法比传统GA优化网络的速度提高16.09%,预测精度提高2.99%,能更好地满足动态流量软测量的要求.
During studying the core algorithms of the dynamic flow soft measurement model in hydraulic servo system, a new algorithm named BFGS-GA was proposed. It took the merits and the shortcomings of the GA and the BFGS algorithm in optimization into account. A BFGS operator was introduced into the genetic algorithm. BFGS operator carded on the BFGS iterative computation by certain probability for certain elitists of every generation. The chromosomes were coded with real number and optimized by the BFGS-GA. The experimentation results show that comparing to the traditional genetic algorithm, the BFGS-GA could save the training time about 16.09% and increase the forecast precision about 2.99%. So it can satisfy the request of dynamic flow soft measurement well.
出处
《小型微型计算机系统》
CSCD
北大核心
2009年第9期1831-1833,共3页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(50675189)资助
河北省自然科学基金项目(F2006000267)资助
关键词
软测量
动态流量
RBF神经网络
GA
BFGS算法
soft measurement
dynamic flow
radial basis function neural network
genetic algorithm
BFGS algorithm
