張 潮，詹全忠

　　(水利部信息中心 北京，100053)

　　摘 要：近年來，對水資源評價的時效性與準確性要求不斷提高，數據量隨之不斷增長，對數據處理技術要求越來越高。為解決水資源動態評價大量數據的匯集、存儲、計算等需求，深入分析水資源動態評價大數據的數據和計算需求，提出數據采集—存儲計算—數據應用的分層大數據計算框架，并著重研究水資源動態評價數據的多源實時匯集技術，實現多源數據統一采集處理。針對數據分析的計算算法模型，研究CPU和GPU混合計算架構，實現機器和深度學習計算支撐。目前，研究成果已經在水利部完成部署，運行良好，可為水資源動態評價工作提供計算支撐。

　　關鍵詞：水資源評價;動態評價;大數據匯集;技術框架; 大數據平臺;混合計算

　　Research and practice of big data calculation framework for dynamic evaluation of water resources

　　ZHANG Chao, ZHAN Quanzhong

　　(Information Center of Ministry of Water Resources, Beijing100053, China)

　　Abstract: In recent years, the requirements for timeliness and accuracy of water resources evaluation have been continuously improved. The data size is increasing and the requirements for data processing technology are getting higher. In order to meet the demand of water resources dynamic evaluation collection, storage and calculation, this paper deeply analyzes the data and calculation requirements of big data in water resources dynamic evaluation, proposes a hierarchical big data calculation framework of data collection-storage computing-data application, and focuses on the research of multi-source real-time collection technology of water resources dynamic data evaluation to realize unified collection and processing of multi-source data. For the computing algorithm model of data analysis, the hybrid computing architecture of CPU and GPU is studied to realize the computing learning and deep computing support. At present, the research results of this paper have been deployed in the Ministry of Water Resources and the systems are running well, which can provide computational support for the dynamic evaluation of water resources.

　　Key words: water resources evaluation; dynamic evaluation; big data collection; technical framework; big data platform; hybrid computing