论文部分内容阅读
为确定气象环境变化对散射比的影响程度,提出主成分分析(PCA)与LMBP神经网络相结合的光伏直散分离模型,利用北京地区5年逐日地面辐射资料,通过相关系数矩阵,选出清晰度指数、日照百分率、PM2.5、总云量和气温日较差作为突出气象影响因子,采用PCA法对多维影响因子作预处理,根据贡献率选出3个主成分,将其作为LMBP神经网络的输入参数,进而通过误差分析方法分别对MLR模型、PCA-MLR模型、PCA-BP模型和PCA-LMBP模型进行评估。结果表明,PCA-MLR模型和PCA-LMBP模型的散射比预测值与实测值更吻合,其中基于PCA-LMBP神经网络的直散分离模型预测精度最高、泛化性能最好,具有一定的实际应用价值。
In order to determine the influence degree of meteorological environment on the scattering ratio, a PCA model combined with LMBP neural network is proposed. By using daily ground-level radiation data of five years in Beijing and correlation coefficient matrix, a clear Degree index, sunshine percentage, PM2.5, total cloud amount and temperature day as meteorological influencing factors, the multi-dimensional influencing factors were pre-treated by PCA method. According to the contribution rate, three principal components were selected and used as LMBP nerve Then, the MLR model, PCA-MLR model, PCA-BP model and PCA-LMBP model are respectively evaluated by error analysis. The results show that the PCA-MLR model and the PCA-LMBP model scatter better than the predicted and measured values, and the PCA-LMBP neural network based model has the highest prediction accuracy and the best generalization performance, and has some practical applications value.