基于改进B样条神经网络-PID控制器的温室温度控制技术

Temperature control technology of greenhouse based on improved B spline neural network-PID

针对温室温度控制系统所存在的大惯性、非线性等问题,提出神经网络PID控制算法,并利用知识局部存储且具有较快学习速度的B样条函数作为网络隐层神经元函数,同时,提出了β参数型-B样条曲线的重新参数化方法,通过学习算法对β参数搜索来动态调节B样条基函数,从而建立B-BP神经网络,并利用其对PID控制器的比例、积分和微分参数进行优化调整,从而为B-BP-PID控制器的参数自适应调整提供更好的保证,使温度控制系统有效跟踪系统模型并达到较高的辨识精度。仿真试验获得B-BP-PID控制器的最佳β因子为3.2,其温度控制超调量为27%,调节时间为0.8 s,而BP-PID控制器的超调量为25%,调节时间为4.8 s,RBF-PID控制器的超调量为40%,调节时间为1.2 s,新算法有效提高了温度控制过程的稳定性、精确性与鲁棒性。

In view of the problems of nonlinear and great inertia existing in the temperature control system of greenhouse,propose the neural network PID control algorithm,and the B spline function is used as the neural function of the network hidden layer,which is stored locally and with fast learning speed,at the same time,the re-parameterization method ofβparameterized B-spline curve is proposed,which realize dynamic adjustment of B spline basis function by learning algorithm,thus establish the B-BP which is used to optimize and adjust the proportion,integral and differential parameters of the PID controller,so as to provide a better guarantee for the parameter adaptive adjustment of B-BP-PID controller which make the temperature control system effectively tracking system model and achieve the high accuracy.The best factorβof B-BP-PID controller is 3.2 which obtained by the simulation experiment,its overshoot is 27%,the regulating time is 0.8 s,and the overshoot of BP-PID controller is 25%,its regulating time is 4.8 s,the overshoot of RBF-PID controller is 40%,its regulating time is 1.2 s.So the new algorithm effectively improves the stability,accuracy,robustness in the process of temperature controlling.