基于掩模经验模态分解的风速组合预测模型

在风电预测中，传统的经验模态分解法将风速信号分解为若干具有不同特征尺度的数据分量时，其所得分量可能存在模态混叠现象，影响风速预测的精度。为此，提出一种基于掩模经验模态分解法和遗传神经网络的风速预测组合模型。首先，通过掩膜信号法（masking signal，MS）对经验模态分解法进行改进，将风速信号分解为频率相对固定、更为平稳的分量。之后，利用遗传神经网络算法分别对这些分量进行预测，将各分量预测结果叠加后得到最终风速预测值。通过C++语言编程进行算法实现，采用实际风场数据进行仿真，其结果表明，所提方法计算时间较短，预测精度较高，特别适用于在线超短期（10 min）和短期（1 h）的风速预测，具有实际的工程应用价值。

Wind Speed Forecasting Modelling by Combination of Masking Signal Based Empirical Mode Decomposition and GA-BP Neural Network

For wind power forecasting, the traditional empirical mode decomposition method usually decompose the wind speed signal into several components with different frequencies. However, the mode-mixing phenomenon may exist and affect the accuracy of forecasting. To solve the problem, a new combination model based on the advanced empirical mode decomposition(EMD) and GA-BP neural network was proposed. Firstly, the traditional empirical mode is improved by the masking signal method(MS), by which the original data can be decomposed into more stationary signals with different frequencies. Secondly, each signal is taken as the input data to establish GA-BP neural network forecasting model. Finally, the forecasting results can be obtained by adding up the predicted data of each signal. The proposed method was programmed by C++ and tested by using the data from an actual wind farm. The simulation experiments show that the proposed method can improve the forecasting accuracy and its running time is also short, which is suitable for ultra short term(10 min) and short term (1hour) wind speed forecasting on line. Consequently, it has a certain practical significance.