人工神经网络技术发展及在大气科学领域的应用

Progress in Researches on Artificial Neural Network Theory and Application in Atmospheric Science

20世纪 8 0年代以来 ,人工神经网络技术在理论和应用研究方面有了快速的发展 ,并在众多学科领域取得了大量的应用研究成果。文章首先概述了国外有关人工神经网络理论研究发展的主要过程及我国开展神经网络理论和应用研究的状况。重点介绍了国内外大气学科中 ,关于神经网络方法在中、短期天气预报 ,短期气候预测 ,农业气象 ,空气污染预报 ,卫星云图识别以及气象观测资料处理等许多方面的应用研究工作。并且也进一步介绍了有关神经网络气象预报建模研究的关键技术 -过拟合 ,网络结构问题的研究工作以及模糊系统理论与神经网络相结合的模糊神经网络 ,遗传算法与神经网络相结合的气象预报建模的最近研究工作。

Since the 1980s, theoretical and applied artificial neural network researches have developed rapidly and gained great achievements in many fields. The main developmental courses of theoretical and applied researches in artificial neural network both at home and abroad in China are described. Emphases are put on the application of artificial neural network in many fields, such as mid-term and short-term weather forecasting, short-term climate prediction, agrometeorology, air pollution forecasting, recognition of satellite cloud images, meteorological observation data processing, and so on. Descriptions are further made on the recent researches on the key technology of modeling researches in weather forecasting using artificial neural network: over-fitting, network structure, fuzzy neural network (the combination of fuzzy system with neural network), and the weather forecast modeling of combining Genetic Algorithms with neural network.