基于拟合剔除的优化小波神经网络太阳辐射量预测

Prediction of solar radiation based onoptimized wavelet neural network with fitting and culling

准确预测太阳辐射量,对太阳能相关产业具有重要意义,针对太阳辐射的波动性和间歇性,提出一种基于曲线拟合和拉依达准则的数据处理和优化的小波神经网络的太阳辐射量的预测方法。通过历史太阳辐射数据和气象数据对太阳辐射量进行直接预测。对测量值求拟合曲线,利用拉依达准则对数据的拟合值和测量值的偏差做粗大误差的判断,修正后的数据作为小波神经网络的输入,避免输入极端数据造成预测信息畸形的问题。增加测试数据对小波神经网络做隐含层节点数寻优的计算,克服小波神经网络无法确定隐含层节点数的缺点。通过建立不同预测模型进行对比,验证了所提算法和模型的正确性。

Accurate prediction of solar radiation is of great significance to solar energy related industries,aiming at the volatility and intermittent of solar radiation,a solar radiation prediction method based on the optimized wavelet neural network with curve fitting and Pauta criterion is put forward in this paper.The solar radiation amount is predicted through the historical solar radiation data and meteorological data.The fitting curve of the measured value is obtained,and the fitting value and the measured value get rid of the great error according to the Pauta criterion.The correctional data is used as the input of the wavelet neural network,to avoid problem of measuring information malformation because of input with the extreme data.The test data is added to do the optimization calculation of hidden layer node number for wavelet neural network to overcome the shortcoming that the wavelet neural network cannot determine the number of hidden layer nodes.The correctness of the proposed algorithm and model is verified by comparing different prediction models.