基于优化神经网络的温室厚皮甜瓜病害预测

在温室环境中,厚皮甜瓜较易感染一些病害,而传统的病害预测模型收敛速度慢,易在局部局限在极小值,为准确预测温室厚皮甜瓜病害,在BP神经网络的基础上进行优化,引入了遗传算法,在全局最优解的附近进行局部搜索,以遗传算法的全局搜索能力克服了传统神经网络的局部极小值问题与收敛速度缺陷。经以Matlab对试验数据进行仿真分析,证实引入遗传优化算法进行温室厚皮甜瓜病害预测误差显著减小,取得了较理想的拟合结果。

Muskmelon Disease Forecasting Based on Optimized BP Neural Network

In the environment of greenhouse,muskmelon is often infected by diseases.The traditional neural network algorithm converges slowly,easy to limited to the minimum in the local convergence.In this paper,genetic algorithm was led into BP network to overcome the defects by its global search ability.The Matlab simulation analysis of test data confirmed that the introduction of genetic optimization algorithm significantly reduced the prediction errors of greenhouse muskmelon disease,and obtained better fitting results.