基于模糊识别和遗传神经网络的渤海湾叶绿素a预测研究  

Prediction of chl-a in Bohai Bay by genetic neural network and fuzzy recognition

作　　者：向先全[1] 陶建华[2] 

机构地区：[1]天津大学环境科学与工程学院,天津300072 [2]天津大学机械工程学院,天津300072

出　　处：《海洋环境科学》2011年第2期239-242,278,共5页Marine Environmental Science

基　　金：国家自然科学基金项目(10872144)

摘　　要：由于海洋生态系统的高度复杂性和非线性,利用新兴的水信息学技术,包括模糊模式识别、遗传算法、人工神经网络等构建渤海湾叶绿素a预测模型。以渤海湾实测水质数据为依据,利用遗传算法的全局搜索能力优化BP神经网络的初始权值和阈值,从而避免陷入局部最优解,构建GANN叶绿素a预测模型。模型预测的均方根误差为3.81μg/L,仿真效果好于传统的BP网络模型。为进一步提高预测精度,用模糊模式识别方法遴选出与测试样本较匹配的训练样本,输入到GANN模型中进行训练。模型预测的均方根误差为1.45μg/L,预测效果有较大的提高。研究表明,所提出的改进模型应用在渤海湾叶绿素a的预测中是合理、可行的,具有较高的精度。Owing to the high complexity and the non-linearity of marine ecosystems,a new emerging technology,i.e.hydro-informatics such as artificial neural networks,fuzzy recognition and genetic algorithm have been employed to establish the prediction model on chl-a.Based on the water quality monitoring data in Bohai Bay,the prediction model of GANN chl-a was build through the initial weight values and the threshold value of BP neural network optimized by genetic algorithm with its global search ability.The root mean square error of GANN model was 3.81 g/L,which was better than that of BP model without optimization by GA.In order to further improve the prediction accuracy,fuzzy recognition was applied to select the training samples matching with the test samples as input of GANN model.The root mean square error was 1.45 g/L,showing the great improvement.The results showed that the proposed model was feasible,reasonable and highly precise for the prediction of chl-a in Bohai Bay.

关 键 词：神经网络 遗传算法 模糊模式识别 渤海湾 叶绿素A 

分 类 号：X171[环境科学与工程—环境科学] TP181[自动化与计算机技术—控制理论与控制工程]