摘要
针对热工过程难以建立精确数学模型的特点,将广义动态模糊神经网络应用于热工复杂系统的辨识。该算法以模糊完备性作为高斯函数宽度的确定准则,避免初始化选择的随机性。同时,该算法不仅能对模糊规则而且能对输入变量的重要性做出评价,从而使得每个输入变量和模糊规则都可以根据误差减少率进行修正。通过对某电厂一次风量和平均床温实测数据的仿真实验结果表明,该方法具有较高的辨识精度和效率。
It is difficult to establish accurate mathematical model for thermal processes. For this characteristic, a generalized dynamic fuzzy neural network was used in identification of thermal complex systems. The breadth of Gaussian function was determined in this algorithm based on fuzzy -completeness, which avoided random initialization choice. At the same time, this algorithm can not only make the evaluation on the hnportance of the fuzzy rules, but also that of input variables, so that each input variables and fuzzy rules can be amended in accordance according to the error reduced rate. Through the simulation on the first wind data and the average bed temperature data, the experimental results showed that the proposed approach has higher accuracy and efficiency in thermal system identification.
出处
《电力科学与工程》
2009年第7期38-41,共4页
Electric Power Science and Engineering
关键词
模糊规则
广义动态模糊神经网络
热工辨识
仿真实验
fuzzy rules
generalized dynamic fuzzy neural network
thermal system identification
simulation experiment
