摘要
在模糊神经网络中采用传统的梯度下降优化方法,其搜索速度慢,并易于陷于局部最小的缺点,提出一种自适应粒子群算法,采用由一个模糊推理机来动态地修改速度参数,模糊推理机的两个输入分别是当前速度参数,以及规范化的当前最好性能估计,输出是速度参数的增量;并将该方法用于模糊神经网络的参数的优化中,得到一种新的建模方法。最后以德士古气化炉为对象,用该方法建立炉膛温度的软测量模型,结果表明该方法该模型运算速度快,同时具有良好泛化性能,能够满足软测量建模精度的要求。
This paper offers an Adaptive PSO Algorithm.A fuzzy system is implemented to dynamically adapt the inertia weight of the PSO,here two variables are selected as inputs to the fuzzy system (the current best performance evaluation and the current inertia weight),the output variable is the change of inertia weight.At the same time,this method is applied to optimize the parameter of the FNN,so a new modeling method can be achieved. Finally,the modeling method is used to model the temperature measurement of Texaco slurry gasifie,the result shows that it can calculates at the higher speed,less fuzzy rules and better generalization capability and achieve satisfactory prediction precision.
出处
《工业控制计算机》
2006年第3期9-11,共3页
Industrial Control Computer
基金
上海市曙光计划项目(03SG26)
关键词
减法聚类
PSO
模糊模型
软测量
聚类半径
气化炉
subtractive clustering,PSO,fuzzy model,soft sensor,radius of cluster center,gasifier