【摘要】 溶解氧是水产养殖中水质的重要指标,针对鱼塘溶解氧量的预测进行研究,可为科学管理鱼塘提供重要的理论依据。鱼塘溶解氧量受多种因素影响,参数之间作用机理复杂,目前常用的BP神经网络算法能有效地解决非线性映射问题,但是对于高维输入变量时,会使网络训练复杂化,容易陷入局部最优化问题,难以得到全局优化解。为了解决上述问题,提出一种将主成分分析方法和BP神经网络相结合的溶氧值预测模型。利用主成分分析方法对输入样本进行降维,消除因子间相互影响,进而建立BP神经网络模型进行预测,并与基本BP神经网络模型进行比较。仿真结果表明,采用基于主成分分析方法的BP神经网络预测模型预测稳定性较好、精度更高,对鱼塘溶氧值预测具有一定的实际参考价值。更多还原

【Abstract】 It is important to study the dissolved oxygen precision in aquaculture.The relationship between dissolved oxygen and influence factors are complicated,and BP neural network is widely used for dissolved prediction.In order to overcome the shortcoming of BP neural network to solve the problem of high dimension input variables,such as increasing the training time,easy to fall into local optimization problems,we studied a new neural network predict model of dissolved oxygen by combining the principal component analysis and BP neural network.First,we used the principal component analysis to reduce the dimension of input samples and eliminate the mutual influence.Then we established the predict model and compared it with the basic BP neural network.The result shows that the predict model is more accurate and of better stability than the basic BP.更多还原