基于集合经验模式分解和遗传-高斯过程回归的短期风速概率预测

Short-term Wind Speed Probabilistic Forecasting Based on EEMD and Coupling GA-GPR

短期风速概率预测对实现大规模风电并网具有重要意义。当前风速预测方法大多为点预测,无法描述风能的随机性。提出了一种基于集合经验模式分解(EEMD)和遗传-高斯过程回归(GAGPR)的组合概率预测方法,首先对筛选和归一化后的风速时间序列进行集合经验模式分解,然后对各分量分别建立高斯过程回归模型,并引入遗传算法代替共轭梯度法,改进协方差函数的超参数寻优过程。最后叠加子序列预测结果得到风速概率预测结果,并与分位点回归法进行比较。仿真结果表明,该方法能够有效提高概率预测准确度,并为类似工程提供借鉴。

Short-term wind speed probabilistic forecasting is quite significant for grid integration of large wind energy. By now the wind speed forecasting methods are mostly point predictions, whose results cannot describe the randomness of wind energy. A hybrid probabilistic forecasting method based on ensemble empirical mode decomposition( EEMD) and genetic algorithm-Gaussian process regression( GA-GPR) is proposed. Firstly,the EEMD is used to decompose the selected and normalized wind speed time series. Then,the GPR models of each component are established, in which the conjugate gradient algorithm is replaced by GA to optimize the hyper-parameters of covariance functions. Finally,the wind speed probabilistic forecasting results are obtained via superimposing the results of each component,which are compared with the quantile regression algorithm. The simulation results show that the proposed model can enhance the prediction precision,which can be served as a reference for similar engineering projects.