基于奇异谱分析-模糊信息粒化和极限学习机的风速多步区间预测

Wind Speed Multi-Step Interval Prediction Based on Singular Spectrum Analysis-Fuzzy Information Granulation and Extreme Learning Machine

不同于风速点预测,风速区间预测能描述风速的随机性。因此,提出一种基于奇异谱分析-模糊信息粒化和极限学习机组成的风速多步区间预测模型。该方法采用奇异谱分析提取原始数据的趋势成分、振荡成分和噪声成分,并对所有分量进行重构,然后利用模糊信息粒化对重构后的噪声成分进行有效挖掘,提取每个窗口最小值、平均值和最大值。对各分量采用极限学习机分别建立预测模型,为了提高预测精度、缩小区间范围,采用改进布谷鸟算法对预测模型的参数进行优化。最后将所有分量的预测结果进行叠加,实现风速区间预测。以风电场实际数据为算例,结果表明所提方法具有较高的预测精度和可靠的多步区间预测,且运行效率高,能有效跟踪风速变化。

Unlike wind speed prediction, wind speed interval prediction can describe randomness of wind speed. A novel model for wind speed interval prediction was proposed by combination of singular spectrum analysis-fuzzy information granulation（SSA-FIG） and extreme learning machine（ELM）. SSA was used to extract trend, oscillating and noise components of original data, and reconstruct all the components. Fuzzy information granulation of the reconstructed noise components was performed. The minimum, average and maximum values of each window is extracted according to need, and ELM algorithm is adopted to build forecasting model for each component. Improved cuckoo search algorithm（ICS） is introduced to optimize model parameters for further improving prediction accuracy and reducing interval range. The overall interval prediction with a certain confidence level is obtained by superimposing the forecasted results of three components. Results for a practical case show that, the proposed method can get higher forecasting accuracy, more reliable multi-step interval forecast and higher efficiency, and is able to track wind speed variation.