论文部分内容阅读
针对赤潮发生的突发性及非线性等特点,提出了基于LMBP神经网络(Levenberg–Marquardt Back-Propergation Neural Network Algorithm)的赤潮预测模型。通过多组对比实验构建了最优的LMBP神经网络模型,并将该模型与标准BP网络模型进行了对比实验,对比结果证明了LMBP模型在收敛速度和拟合精度上的优越性。进而对该赤潮预测模型进行了测试,其结果充分表明:以叶绿素a,透明度为输入的该LMBP神经网络模型对烟台四十里湾海域的赤潮预测是有效的。目前,该模型已经进行了应用实验,实时预测效果良好。
In view of the sudden and non-linear characteristics of red tide occurrence, a red tide prediction model based on LMBP neural network (Levenberg-Marquardt Back-Propergation Neural Network Algorithm) is proposed. The optimal LMBP neural network model is constructed by comparing several groups of experiments, and the model is compared with the standard BP network model. The comparison results prove the superiority of the LMBP model in convergence rate and fitting accuracy. The result shows that the LMBP model based on chlorophyll a and transparency is effective for the prediction of red tide in Yishilai sandy waters. At present, the model has been applied experiment, real-time prediction effect is good.