基于神经网络的孵化过程组合预测方法  

Method of Combination Prediction for Incubator Based on Network

作　　者：周国雄[1] 廖迎新 沈学杰[1] 

机构地区：[1]中南林业科技大学计算机与信息工程学院,长沙410004

出　　处：《系统仿真学报》2014年第4期892-896,共5页Journal of System Simulation

基　　金：湖南省自然科学基金项目(10JJ3066)

摘　　要：针对禽蛋孵化过程是一个具有高度非线性、大滞后且强耦合性的农业生产过程,常规的控制方法难以达到较好的控制效果,提出一种禽蛋孵化过程的组合预测方法。该方法分别采用神经网络模型和过程记忆神经网络对禽蛋孵化过程温、湿度进行预测,然后采用方差–协方差优选组合预测法对两种单一模型的预测结果进行加权集成,以获得较为准确的预测精度,实现禽蛋孵化过程温度和湿度的有效预测。仿真运行结果表明对两种单一预测模型的预测结果进行加权组合后得到的组合预测模型的预测精度明显要优于单一预测模型,能较好的预测孵化过程温、湿度,从而较好的保证禽蛋孵化过程的稳定控制。Due to the incubation process which is highly nonlinear, large delay and strongly coupling and can’t be controlled effectively by common method, a combination forecasting method for incubator was proposed. Firstly, the process neural networks and associative memory neural networks were respectively used on prediction for incubator, then combining the two sub-models with weighted integration by using the variance-covariance combination forecasting method, the value of the temperature and humidity of incubator could be more accurate and effective, which could make the temperature and humidity of incubator control effectively. Simulation shows the accuracy of combination forecasting model prediction is significantly better than the single prediction model, which can better predict the temperature and humidity of incubation process and ensure the stability of the control of the hatching process.

关 键 词：过程神经网络 联想记忆神经网络 孵化 组合预测 

分 类 号：TP229[自动化与计算机技术—检测技术与自动化装置]