论文部分内容阅读
针对BP算法易陷入局部极小、收敛速度慢等缺点,遗传算法是全局优化算法和具有很强的全局搜索能力,遗传算法优化BP神经网络初始连接权值和阈值形成混合算法。以安徽宣城市为例,将汛期降水量作为预测对象,前期74项大气环流特征量、500 hPa、100 hPa月平均高度场、月平均海平面气压场和月平均海温场资料中选取预测因子,建立汛期降水短期气候预测模型。结果表明,该方法计算稳定,预报误差小,具有实用价值。
Aiming at the shortcomings of BP algorithm, such as local minima and slow convergence speed, genetic algorithm is a global optimization algorithm and has strong global search ability. The genetic algorithm optimizes the initial connection weight and the threshold formation algorithm of BP neural network. Taking Xuancheng City, Anhui Province as an example, the forecast factors of precipitation in the flood season are 74 forecasted atmospheric circulation features, 500 hPa, 100 hPa monthly mean height field, monthly average sea level pressure field and monthly average SST field data , To establish a short-term climate forecast model for precipitation in flood season. The results show that the method is stable in calculation and has small prediction error and practical value.