小波分析是一种新的信号处理技术 ,具有良好的时频局部化特征。为了克服BP网络自身算法的缺陷 ,得到更高的学习精度和更快的收敛速度 ,使用小波包分析的特征提取及神经网络的非线性映射特性 ,构造了小波神经网络 ,以此为基础开发出的软件系统具有使用的特征量少 ,建造预报系统较为简单等优点。将之应用于炼铜转炉炉渣重量及成分预报 ,该模型完全能够较准确地预报出渣量和成分。其平均拟合误差为 1 5 % ,平均预报误差为 3 1% Wavelet analysis is a new signal processing technique with good time-frequency localization. In order to overcome the shortcomings of the BP neural network algorithm, higher learning accuracy and faster convergence speed are obtained. Based on the wavelet neural network, wavelet packet neural network is developed by using wavelet packet feature extraction and neural network nonlinear mapping Of the software system has the characteristics of using less, the construction of forecasting system is relatively simple and so on. Applying this method to the slag weight and composition forecast of copper smelting converter, the model can predict the amount of slag and composition more accurately. The average fitting error is 15%, the average forecast error is 31%