【摘要】 针对太阳辐照度时间序列的非线性特点,文章设计了一种新的基于二阶数据分解算法和蝗虫优化混合核LSSVM的太阳辐照度预测模型,并对该模型进行了验证。首先,利用集合经验模态分解(EEMD)算法对原始太阳辐照度时间序列进行分解,得到若干个频率不同的分量;然后,利用变分模态分解(VMD)算法进一步分解频率最高的分量,得到K个相对稳定的分量,其中,K由各分量与利用VMD算法分解得到的残差的相关系数确定;接着,建立基于高斯核和多项式核的混合核最小二乘支持向量机(LSSVM)预测模型,对所有分量进行预测,并利用蝗虫优化算法优化混合核函数的参数;最后,将所有分量的预测结果相加得到原始太阳辐照度时间序列的预测结果。模拟结果表明,与BP神经网络模型、ARIMA模型、LSSVM模型和基于EEMD,LSSVM的预测模型相比,基于二阶数据分解算法和蝗虫优化混合核LSSVM的太阳辐照度预测模型的预测精度更高,能有效反映太阳辐照度的变化规律。更多还原

【Abstract】 According to the nonlinear characteristics of solar irradiance time series, a new solar irradiance prediction model based on second-order data decomposition algorithm and hybrid kernel LSSVM which was optimized by grasshopper optimization algorithm was designed in this paper. First,the ensemble empirical mode decomposition(EEMD) algorithm was used to decompose the original solar irradiance time series, and several components with different frequencies were obtained. Then,the component with the highest frequency was further decomposed by the variational mode decomposition(VMD) algorithm, and K relatively stable components were obtained. K was determined by the correlation coeficient between each component and the residual decomposed by VMD algorithm. After that, the mixed kernel least squares support vector machine(LSSVM)prediction model based on Gaussian kernel and polynomial kernel was established to predict all components, and the grasshopper optimization algorithm was used to optimize the parameters of the hybrid kernel function. Finally, the prediction result of the original solar irradiance time series was obtained by adding the prediction results of all components. The simulation results showed that,compared with BP neural network model, ARIMA model, LSSVM model and prediction model based on EEMD and LSSVM, the prediction accuracy of solar irradiance prediction model based on two-stage data decomposition algorithm and hybrid kernel LSSVM optimized by grasshopper optimization algorithm was higher, which could effectively reflect the variation of solar irradiance.更多还原