Reliable and accurate wind speed forecasting is the basis for the effective development of wind energy. However, wind speed is intermittent, presents nonlinear patterns, and exhibits nonstationary behavior; thus, it is generally difficult to predict it accurately and efficiently using a single linear or nonlinear model. Hence, in this study, two novel hybrid forecasting systems based on the structural characteristics of wind speed are proposed to capture the linear and nonlinear factors hidden in wind speed series. First, a decomposition algorithm is used to eliminate noise from raw data and reconstruct a more reliable wind speed time series. Then, a linear model, which employs the exponential smoothing model or autoregressive moving average model, captures the linear patterns hidden in the time series, and a nonlinear model, which applies the back propagation neural network optimized by the cuckoo search algorithm, extracts the nonlinear patterns hidden in the data. The experimental results using nine datasets show that the proposed model has better prediction accuracy than the comparison models and the root mean square error (RMSE), the mean absolute error (MAE) are respectively less than 0.2139 and 0.125, which provides a scientific basis for power grid dispatch and guarantees the stable operation of the wind power system.

可靠，准确的风速预报是有效开发风能的基础。但是，风速是断断续续的，呈现非线性模式，并且表现出不稳定的行为。因此，通常难以使​​用单个线性或非线性模型来准确而有效地进行预测。因此，在这项研究中，提出了两种基于风速结构特征的新型混合预测系统，以捕获风速序列中隐藏的线性和非线性因素。首先，使用分解算法消除原始数据中的噪声并重建更可靠的风速时间序列。然后，使用指数平滑模型或自回归移动平均模型的线性模型捕获时间序列中隐藏的线性模式，然后使用非线性模型，它应用了布谷鸟搜索算法优化的反向传播神经网络，提取了隐藏在数据中的非线性模式。利用9个数据集进行的实验结果表明，该模型的预测精度优于比较模型，均方根误差（RMSE）和均方根绝对误差（MAE）分别小于0.2139和0.125，为进一步的研究提供科学依据。电网调度，保证了风电系统的稳定运行。