引入免疫粒子群优化算法的木材干燥模糊神经网络控制系统设计

Wood Drying Temperature and Humidity Method by Immune PSO Algorithm

针对木材干燥窑温湿度控制采用的模糊神经网络比较依赖于网络初始权值,且网络的训练时间较长、容易陷入非要求的局部极值,采用粒子群优化算法(PSO)的全局寻优性能,设计一种引入免疫PSO算法的木材干燥模糊神经网络控制系统。为避免PSO算法的早熟和进一步导入待求解问题的先验知识与经验,加快算法的全局收敛能力,引入免疫算法的接种疫苗、免疫选择、良种迁移3种免疫算子。仿真结果表明:温度和湿度,能更加快速、平滑地到达设定值(温度需要70 s左右,湿度需要75 s左右)。实例验证结果表明:温度曲线均方误差仅为0.020 7,拟合优度高达0.979 7;湿度曲线均方误差均在0.3以下,拟合优度均在0.96以上。说明免疫PSO算法具有较高的收敛速度和识别率,对不确定非线性系统具有良好的控制效果。

To improve the temperature and humidity control precision of wood drying process, fuzzy neural network control system for wood drying was designed by immune PSO algorithm. According to the overused network with initial weights, the long-time network training and the non-required local extremum in the previous fuzzy neural network of controlling temperature and humidity on lumber kiln, the global particle swarm optimization （PSO） algorithm was adopted. However, in order to avoid earliness of PSO and lead in prior knowledge and experience of unsolved problems, as well as accelerating global convergence of algorithm, three improved immune operator were added, including vaccination, immune selection and fine breed migration. Simulation results show that the temperature and humidity can be more quickly and smoothly reaches the set value （the temperature takes 70 s, and the humidity takes about 75 s）. The temperature curve mean square error is 0.020 7, the goodness of fit is 0.979 7, humidity curve mean square errors are below 0.3, and the goodness of fit are above 0.96. This method has higher convergence rate and recognition rate with better control effect on uncertain nonlinear systems.