中文摘要：

提出了基于小波神经网络的太阳辐照强度预测方法。利用皮尔逊相关系数分析法和曲线估计筛选出影响太阳辐照强度的重要因素;采用小波理论和神经网络理论相结合的小波神经网络分别建立春、夏、秋、冬4个预测模型;采用最小均方误差能量函数法自动优化网络结构,把历史太阳辐照强度、经度、纬度、海拔高度、天气类型、日照时数、最高温度、最低温度、相对湿度、大气压强作为模型的最优输入;采用L-M训练方法对太阳辐照强度进行了min级预测。通过对4个季节特殊天气类型的太阳辐照强度预测,并与BP神经网络进行对比,验证了该方法的可行性和准确性。

英文摘要：

A method of solar irradiation intensity forecast is put forward based on wavelet neural network.First of all, the Pearson correlation coefficient analysis method and curve estimation are used to filter out the important factors influencing the intensity of solar irradiation; Then, by using wavelet neural network theory which combines neural network theory with wavelet theory, four forecasting models for spring, summer, autumn and winter are established.The method of minimum mean square error energy function is used to optimize the network structure automatically.The history data such as solar irradiation intensity, longitude, latitude, altitude, weather, sunshine time,maximum temperature, minimum temperature, relative humidity and atmospheric pressure are regarded as optimal inputs for this model; L-M training methods are used to fulfill the solar irradiation intensity forecast in minutes scale.Comparing the predictions of four seasons special weather types made by proposed method with that by the BP neural network, the predicted results show the feasibility and accuracy of the proposed method.