基于改进HHT的分钟级超短期风速预测

Minute-scale Ultra-short-term Wind Speed Prediction Based on Improved HHT

针对分钟级的风速具有更强随机性和非线性的问题,建立了一种基于改进希尔伯特黄变换(HHT)的分钟级超短期风速预测模型。首先,利用HHT对原始风速数据进行分解;然后,采用希尔伯特变换(HT)对分解所得各本征模态函数(IMF)进行谱分析,针对其频谱特性分别建立不同的神经网络模型进行预测;最后,分别运用权重浮动区间模型和数学解析模型求取各分量权重系数,按权重叠加各分量预测结果得到2组风速预测值,再通过优化组合得到最终预测结果。采用某风电场实际运行数据,对所建模型的可靠性进行了验证。结果表明:该方法预测效果优于原HHT组合预测模型,可有效提高风速预测精度。

To solve the problem that minute-scale wind speed was difficult to predict due to its stronger randomicity and nonlinearity, a minute-scale ultra-short-term wind speed prediction model was established based on improved Hilbert-Huang transform(HHT). HHT was applied to decompose the original wind velocity data. The spectrum of each intrinsic mode function(IMF) was analyzed through Hilbert transform(HT). According to the spectrum characteristics of each component, different neural network models were built to predict wind speed. Weighted coefficients of each component were calculated through the weight floating interval model and the mathematical analytic model. Two groups of wind speed prediction values were obtained by summarizing each component results according to the weighted values. Final prediction results were obtained by optimizing such combination. The availability of the established model was verified by the actual operation data in a wind farm. Results show that compared with the original HHT combined prediction model, the method is better and can effectively improve the accuracy of wind speed prediction.