BP神经网络在浅水湖泊风浪波高预测中的应用  

Application of BP Neural Network in Prediction of Wind Wave Height in Shallow Lakes

作　　者：高昂 吴时强[1,2] 吴修锋 戴江玉[1,2] 王芳芳 张维乐[1,2] 徐佳怡 GAO Ang;WU Shi-qiang;WU Xiu-feng;DAI Jiang-yu;WANG Fang-fang;ZHANG Wei-le;XU Jia-yi(State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering,Nanjing Hydraulic Research Institute,Nanjing 210029,China;Key Laboratory of Taihu Basin Water Management,Ministry of Water Resources,Nanjing 210029,China)

机构地区：[1]南京水利科学研究院水文水资源与水利工程科学国家重点实验室,江苏南京210029 [2]水利部太湖流域水治理重点实验室,江苏南京210029

出　　处：《水电能源科学》2022年第8期41-44,168,共5页Water Resources and Power

基　　金：国家重点研发计划(2018YFC0407200);国家自然科学基金项目(51909169);江苏省自然科学基金项目(SBK2019042181);中央级公益性科研院所基本科研业务费专项资金项目(Y120010,Y121006)。

摘　　要：风浪是影响浅水湖泊水环境与水生态的重要动力因子,波高是量化风浪强度的重要指标,准确预测风浪波高对浅水湖泊治理具有重要意义。基于太湖实测数据,建立了考虑风速、吹程、水深、风历时等因素影响的波高BP神经网络模型,分析了不同因素对模型预测能力的影响。结果表明,SMB方法在预测浅水湖泊风浪波高时适配性受限,吹程和风速较小时会高估风浪实际波高。BP神经网络模型在考虑风速、吹程与水深影响后,预测与实测波高相关系数可达0.86,当进一步考虑风历时影响后,相关系数最高为0.90。吹程与波高的相关性大于水深与波高的相关性,风浪对气流的最佳迟滞时间为20~30 min。研究论证了BP神经网络在浅水湖泊风浪波高预测方面的适用性。Wind wave is an important dynamic factor affecting water environment and water ecology in shallow lakes,and wave height is an important index to quantify wind wave strength.How to accurately predict wind wave height is of great significance to management of shallow lakes.Based on the measured data of Taihu Lake,the BP neural network model of wave height was established considering the impact of wind speed,blowing distance,water depth and wind duration,and the impact of different factors on the prediction ability of the model was analyzed.The results show that the SMB method has limited suitability in prediction of wind wave height in shallow lakes,and the wind wave height will be significantly overestimated when the blowing distance and wind speed are small.After considering the impact of wind speed,blowing distance and water depth,the correlation coefficient between predicted wave height and measured wave height can reach about 0.86 in BP neural network model.When further considering the impact of wind duration,the highest correlation coefficient can reach 0.90.The correlation between blowing distance and wave height is greater than that between water depth and wave height,and the best hysteresis time of wind wave on airflow is 20-30 minutes.The applicability of BP neural network in the prediction of wind wave height in shallow lakes is demonstrated.

关 键 词：浅水湖泊 风浪 神经网络 波高预测 迟滞效应 

分 类 号：TV124[水利工程—水文学及水资源] P338.9[天文地球—水文科学]