文章摘要
史良胜,查元源,胡小龙,杨琦.智慧灌区的架构、理论和方法之初探[J].水利学报,2020,51(10):1212-1222
智慧灌区的架构、理论和方法之初探
A preliminary exploration of framework,theory and method for intelligent irrigation district
投稿时间:2020-03-17  
DOI:10.13243/j.cnki.slxb.20200169
中文关键词: 智慧灌区  人工智能  大数据  认知智能  感知智能
英文关键词: intelligent irrigation district  artificial intelligence  big data  perception intelligence  cognition intelligence
基金项目:国家自然科学基金项目(51861125202)
作者单位
史良胜 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 430072 
查元源 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 430072 
胡小龙 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 430072 
杨琦 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 430072 
摘要点击次数: 1545
全文下载次数: 819
中文摘要:
      我国灌区面临着多元化、精准化和智能化的管理需求。结合人工智能技术的发展趋势和水利学科本身特点,尝试提出智慧灌区的定义和基本功能。从灌区的感知智能、认知智能和管理决策三个方面讨论了智慧灌区研究的难点,初步论述了智慧灌区架构、理论和方法。在灌区感知智能方面,提出非接触式与接触式观测相结合、移动式和固定式观测相结合的观测体系,开发适用于灌区不同尺度观测的机器视觉技术、灌区特征解译技术和天地空数据智能交换技术;建立大数据环境下的信息理论,支撑灌区多源观测体系的研发和海量数据的解译和分析。在灌区认知智能方面,提出物理方程和机器学习相结合的灌区建模思路,建立“数据-物理”混合认知智能的理论基础,开发非完整先验物理机制下灌区水分、盐分、养分、污染物迁移转化以及作物生长和生态系统演化模型,开发灌区物理机制挖掘方法和动态建模方法。在灌区决策方面,提出可满足众多决策者需求以及可处理多种目标的智能优化决策系统,实现灌区尺度上的水量、水质和生态的最优化管理。
英文摘要:
      Irrigation district in China is confronted with the requirement of diverse, precise and intelligent management. Integrating the development trend of artificial intelligence and water resources engineering, the definition and basic functions of intelligent irrigation district (ⅡD) is tentatively proposed in this paper, identifying the major challenges of ⅡD study from the aspects of perception intelligence, cognition intelligence and management decision making. The framework, theories and approaches for ⅡD are preliminarily explored. With respect to perception intelligence, it should be a combination of non-contact and contact measurement, and a mixture of mobile and fixed measurement. Machine vision, data interpretation and "space-sky-ground"data exchange technology need development for multi-scale observation of irrigation district. Information theory under big-data environment is required to support the multi-source observation system and support the interpretation and analysis of massive data. With respect to cognition intelligence, a modelling framework is proposed by integrating physical equation and machine learning. The theoretical basis of hybridizing data and physical mechanisms should be built. Competent models for water, salt, nutrient,pollutant transport and transformation,crop growth,and ecosystem evolution should be developed even with incomplete priori physical mechanisms. Machine learning physical mechanism from data and adaptive modelling technique should be explored. The intelligent decision-making system is expected to satisfy the requirements from a large number of decision makers and handle multiple objective functions,so that optimal strategies can be made to manage water quantity,water quality and ecosystem.
查看全文   查看/发表评论  下载PDF阅读器
关闭