长短期记忆模型在小流域洪水预报上的应用研究
Application of the Long Short-Term Memory Networks for Flood Forecast
DOI: 10.12677/JWRR.2019.81003, PDF,  被引量    国家自然科学基金支持
作者: 郭 炅, 张艳军, 王俊勃, 吴金津, 董文逊, 王素描：武汉大学水资源与水电工程科学国家重点实验室，湖北 武汉；袁正颖：长江水利委员会水文局，湖北 武汉
关键词: 长短期记忆；洪水预报；Long Short-Term Memory (LSTM)； Flood Forecast
摘要: 在山区小流域，降水资料稀缺，且难以反应其降水的空间异质性，使得仅依靠降水资料进行洪水预报十分困难。为了提高山区小流域洪水预报精度，本文以官山河流域为例，选择可同时输入降水和径流资料进行水文模拟和预报的长短期记忆模型(LSTM)，对洪水过程进行模拟。同时构建了新安江模型模拟，进行对比研究。研究结果表明，若使用1975~1987年逐日数据对模型进行率定和检验，传统水文模型检验期的纳什效率系数为0.55，而对应的LSTM检验期的纳什效率系数为0.73，长短期记忆模型(LSTM)能够较大地提高降水资料缺少地区的水文模拟和预报效果。
Abstract: Flood forecasting is difficult in mountain watershed because precipitation data is scarce and hard to reflect spatial heterogeneity. To improve the accuracy of flood forecasting in mountain watershed, long short-term memory model (LSTM) and Xin’anjiang model are used to simulate flood in Guanshan river watershed. The results show that the Nash efficiency coefficient of verification period in the tra-ditional hydrological model is 0.55, while that in the LSTM is 0.7 with daily data from 1975 to 1987. LSTM can greatly improve the hydrological simulation and forecast effect in the areas lacking precipi-tation data.
文章引用：郭炅, 张艳军, 王俊勃, 袁正颖, 吴金津, 董文逊, 王素描. 长短期记忆模型在小流域洪水预报上的应用研究[J]. 水资源研究, 2019, 8(1): 24-32. https://doi.org/10.12677/JWRR.2019.81003

参考文献

[1] 包红军, 王莉莉, 沈学顺, 等. 气象水文耦合的洪水预报研究进展[J]. 气象, 2016, 42(9): 1045-1057. BAO Hongjun, WANG Lili, SHEN Xueshun, et al. Advances of flood forecasting of hydro-meteorological forecast technology. Meteorological Monthly, 2016, 42(9): 1045-1057. (in Chinese)
[2] 吴超羽, 张文. 水文预报的人工神经网络方法[J]. 中山大学学报(自然科学版), 1994(1): 79-90. WU Chaoyu, ZHANG Wen. Application of artificial neural net: A new approach for hydrological forecasting. Acta Scientiarum Naturalium Universitatis Sunyatseni, 1994(1): 79-90. (in Chinese)
[3] 包红军, 赵琳娜. 基于集合预报的淮河流域洪水预报研究[J]. 水利学报, 2012, 43(2): 216-224. BAO Hongjun, ZHAO Linna. Flood forecast of Huaihe River based on TIGGE ensemble predictions. Journal of Hydraulic Engineering, 2012, 43(2): 216-224. (in Chinese)
[4] 俞亭超, 张土乔, 毛根海, 等. 预测城市用水量的人工神经网络模型研究[J]. 浙江大学学报(工学版), 2004, 38(9): 1156-1161. YU Tingchao, ZHANG Tuqiao, MAO Genhai, et al. Study of artificial neural network model for forecasting urban water demand. Journal of Zhejiang University (Engineering Science), 2004, 38(9): 1156-1161. (in Chinese)
[5] CHIANG, Y. M., CHANG, L. C. and CHANG, F. Comparison of static-feed forward and dynamic-feedback neural networks for rainfall-runoff modeling. Journal of Hydrology, 2004, 290(3-4): 297-311. [Google Scholar] [CrossRef] 
[6] TAKUMI, A., SATOSHI, Y. and KOHEI, S. Distributed CFG-based symbolic execution for assembly programs: Consumer electronics. IEEE Consumer Electronics, Osaka, 2015: 76-80.
[7] 江衍铭, 张建全, 明焱. 集合神经网络的洪水预报[J]. 浙江大学学报(工学版), 2016, 50(8): 1471-1478. CHIANG Yenming, ZHANG Jianquan and MING Yan. Flood forecasting by ensemble neural networks. Journal of Zhejiang University (Engineering Science), 2016, 50(8): 1471-1478. (in Chinese)
[8] 于家斌, 尚方方, 王小艺, 等. 基于遗传算法改进的一阶滞后滤波和长短期记忆网络的蓝藻水华预测方法[J]. 计算机应用, 2018, 38(7): 2119-2123, 2135. YU Jiabin, SHANG Fangfang, WANG Xiaoyi, et al. Cyanobacterial bloom forecast method based on genetic algorithm-first order lag filter and long short-term memory network. Journal of Computer Applications, 2018, 38(7): 2119-2123, 2135. (in Chinese)
[9] 冯钧, 潘飞. 一种LSTM-BP多模型组合水文预报方法[J]. 计算机与现代化, 2018, 275(7): 82-85. FENG Jun, PAN Fei. A hydrologic forecast method based on LSTM-BP. Computer and Modernization, 2018, 275(7): 82-85. (in Chinese)
[10] 王旭东, 严珂, 陆慧娟, 等. LSTM的单变量短期家庭电力需求预测[J]. 中国计量大学学报, 2018, 29(2): 142-148. WANG Xudong, YAN Ke, LU Huijuan, et al. Short-term household electricity demand forecast based on LSTM single variable. Journal of China University of Metrology, 2018, 29(2): 142-148. (in Chinese)
[11] JAVIER, O. F., DANIEL, R. Deep convolutional and LSTM recurrent neural networks for multimodal wearable activity recognition. Sensors, 2016, 16(1): 115. [Google Scholar] [CrossRef] [PubMed]
[12] 权波, 杨博辰, 胡可奇, 等. 基于LSTM的船舶航迹预测模型[J]. 计算机科学, 2018, 45(S2): 126-131. QUAN Bo, YANG Bochen, HU Keqi, et al. Prediction model of ship trajectory based on LSTM. Computer Science, 2018, 45(S2): 126-131. (in Chinese)
[13] 杨凯, 阮晓波, 游圆. 新安江模型在马尪溪流域径流模拟的应用[J]. 浙江水利水电学院学报, 2017, 29(5): 10-13. YANG Kai, RUAN Xiaobo and YOU Yuan. Application of Xin’anjiang model in Mawang River Basin runoff simulation. Journal of Zhejiang University of Water Resources and Electric Power, 2017, 29(5): 10-13. (in Chinese)
[14] 刘佩瑶, 郝振纯, 王国庆, 等. 新安江模型和改进BP神经网络模型在闽江水文预报中的应用[J]. 水资源与水工程学报, 2017, 28(1): 40-44. LIU Peiyao, HAO Zhenchun, WANG Guoqing, et al. Application of Xin’anjiang model and the improved BP neural network model in hydrological forecasting of the Min River. Journal of Water Resources and Water Engineering, 2017, 28(1): 40-44. (in Chinese)
[15] 劳应文, 吴志斌. U形板桩在官山河道治理中的应用[J]. 价值工程, 2015, 34(16): 139-140. LAO Yingwen, WU Zhibin. Application of U-shaped plate piles in Guanshan River regulation. Value Engineering, 2015, 34(16): 139-140. (in Chinese)
[16] 谢俊. 径流时间序列的长记忆特性分析[D]. 武汉: 华中科技大学, 2012. XIE Jun. Research on long memory characteristic of runoff time series. Wuhan: Huazhong University of Science and Technology, 2012. (in Chinese)