基于双隐层BP网络的沈阳地区雷暴潜势预报模型

A Thunderstorm Potential Forecast Model of Shenyang Based on the Double Hidden Layer BP Artificial Neural Network

利用沈阳市2007-2011年6、7、8月探空因子与闪电定位资料,选取适合当地雷暴潜势预报的9个因子,构建双隐层BP人工神经网络预报模型,对当地雷暴的潜势发展进行预报.结果表明:基于双隐层BP人工神经网络的预报模型临界成功指数CSI=57.46%、探测概率POD=79.38%、虚假报警率FAR=32.46%和总指数TS=74.89%,预报效果较令人满意;通过12h和6h潜势预报结果对比发现,6hBP神经网络模型预报效果好,更适用于沈阳地区雷暴潜势预报.

In an experiment reported in this paper,nine factors suitable for thunderstorm potential forecast of Shenyang were selected based on the sounding factors and the lightning location data from 2007 to 2011in June,July and August in this city,and a double hidden layer BP artificial neural network prediction model was built to forecast the local thunderstorm potential development.The experiment showed that the double hidden layer BP artificial neural network prediction model gave satisfactory results,with a critical index of success of CSI=57.46%,a detection probability of POD=79.38%,a false alarm rate of FAR=32.46% and a total index of TS=74.89%.A comparison of the potential forecast results between 12hand6 hdemonstrated that the forecast effect of the 6hBP neural network model had better effect and was,therefore,more suitable for the thunderstorm potential forecast in Shenyang.