In view of the problem of far too large scale neural network structure due to the existence of the complex coupling relationship among input factors and too many input factors in establishing temperature prediction model for sunlight green-house, the environmental factors those affected sunlight greenhouse temperature were selected as data sample, then the princi-pal component analysis of the data sample was conducted, the first 3 main components were extracted as the input variables of BP neural network model, thus using the Bayesian regularization algorithm to improve the BP neural network. The empirical results showed the BP neural network model was improved and simplified by the method, and its fitting curve was smooth, have better generalization ability and popularization performance.%针对日光温室温度预测模型中输入因子间存在复杂的耦合关系以及输入因子过多而导致神经网络结构规模过大等问题,选用影响日光温室温度的环境因子组成数据样本,对数据样本进行主成分分析。提取出影响日光温室内温度的前3个成分作为BP神经网络模型的输入变量,采用贝叶斯正则化算法对BP神经网络进行改进。结果表明,该方法改进的BP神经网络模型得到简化,拟合曲线光滑,具有较好的泛化能力和网络推广能力。
展开▼