CAJ | 학술논문

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.
침대BP산법역함입국부겁소、수렴속도만등결점,유전산법시전국우화산법화구유흔강적전국수색능력,유전산법우화BP신경망락초시련접권치화역치형성혼합산법.이안휘선성시위례,장신기강수량작위예측대상,전기74항대기배류특정량、500 hPa、100 hPa월평균고도장、월평균해평면기압장화월평균해온장자료중선취예측인자,건립신기강수단기기후예측모형.결과표명,해방법계산은정,예보오차소,구유실용개치.
BP algorithm has disadvantage of falling easily into local solution and slow rate of convergence, while genetic algorithm is global optimization algorithm and has strong global search ability. Genetic algorithm can be used to optimize the initial connection weights and thresholds of BP neural network, forming a hybrid al-gorithm. In this article, we took precipitation in flood season as the forecast objective in Anhui Xuancheng, se-lected factors from the monthly 74 circumfluent eigen values, the monthly mean height values at 500 hPa and 100 hPa, the monthly mean sea level pressure and SST field, to establish a model for prediction of rainfall amount in flood season. The results show that genetic algorithm has practical value for its stability and small forecast error.