基于时间序列分析的风功率超短期预测研究

发布时间：2018-07-07 13:53

  本文选题：ARIMA模型 + 经验模态分解　； 参考：《沈阳农业大学》2017年硕士论文

【摘要】：在风力发电中,对风功率的准确预测可以降低旋转备用容量,同时能够给相关部门提供科学的电网调度方案,从而大大提高清洁能源的使用率。但是由于风力发电伴随的波动性和不可控性,使风功率数据具有非线性和非平稳性,因此进行精确的风功率预测变得十分困难。时间序列中ARIMA模型可以根据风功率数据的时序性进行建模预测,但该模型在预测中随着预测的步长增加会使精确度降低。论文将时间序列中的ARIMA预测模型与经验模态分解法(Empirical Mode Decomposition,EMD)进行了组合使用,使风功率预测的精确度得到了一定的提高。在实例计算使用经验模态分解算法时,发现EMD分解法对序列分解过程中会出现模态混叠现象,为了解决该问题,本文对EMD算法进行了一步步的改进,最终建立了改进的集成经验模态分解法(Modified Ensemble Empirical Mode Decomposition,MEEMD)。为了进一步提高风功率预测的精确度,论文将改进的集成经验模态分解法、ARIMA模型以及样本熵(Sample Entropy,SE)模型的优点进行融合,建立了 MEEMD-SE-ARMA模型的混合算法。在论文中针对原始非平稳的风功率数据进行超短期预测,使用MEEMD算法进行分解,将分解后得到的多个风功率子序列使用样本熵进行重组,分为一系列复杂度差异明显的新风功率子序列,使得到的新风功率子序列接近平稳数据;利用时间序列的ARIMA模型对得到的每一个新风功率子序列进行建模预测,在建模过程中应该充分的考虑时间序列中广义自回归条件异方差模型以及拉格朗日乘子检验,并分别建立对应的ARIMA模型;将得到的各预测风功率子序列进行叠加重构,最终得到预测结果。论文分别建立了 ARIMA-GARCH 预测模型、EMD-ARIMA 模型、EEMD-ARIMA 模型以及MEEMD-SE-ARIMA模型,并将各预测模型的平均绝对百分误差进行了对比,最终证明建立的MEEMD-SE-ARMA混合算法可以有效的提高风功率超短期预测的精确度。
[Abstract]:In wind power generation, the accurate prediction of wind power can reduce the rotation reserve capacity, and provide a scientific power grid scheduling scheme to the relevant departments, thus greatly improving the utilization rate of clean energy. However, due to the volatility and uncontrollability of wind power generation, the wind power data are nonlinear and non-stationary, so that the wind power data are not nonlinear and non-stationary. It is very difficult to predict the accurate wind power. In the time series, the ARIMA model can be modeled and predicted according to the timing of the wind power data, but the model can reduce the accuracy with the increase of the prediction step. The ARIMA prediction model in the time series and the Empirical Mode Decompositio method (Empirical) are used in the time series. N, EMD) have been used in combination to improve the accuracy of wind power prediction. When the empirical mode decomposition algorithm is used in the example calculation, it is found that the EMD decomposition method will have modal aliasing in the sequence decomposition process. In order to solve the problem, this paper has improved the EMD algorithm step by step, and finally established an improved set. Modified Ensemble Empirical Mode Decomposition (MEEMD). In order to further improve the accuracy of wind power prediction, the improved integrated empirical mode decomposition, ARIMA model and the advantage of sample entropy (Sample Entropy, SE) model are integrated, and a hybrid algorithm of MEEMD-SE-ARMA model is established. In this paper, the original non stationary wind power data are predicted by ultra short term, and the MEEMD algorithm is used to decompose. The multiple wind power subsequences are reorganized with the sample entropy, which can be divided into a series of new wind power subsequences with distinct differences in complexity, making the obtained new wind power subsequence close to the stationary data and using the time sequence. The ARIMA model is used to model and predict each new wind power subsequence. In the process of modeling, we should fully consider the generalized autoregressive conditional heteroscedasticity model and the Lagrange multiplier test in the time series, and establish the corresponding ARIMA model respectively. The prediction results are obtained. The ARIMA-GARCH prediction model, EMD-ARIMA model, EEMD-ARIMA model and MEEMD-SE-ARIMA model are established respectively, and the average absolute percentage error of each prediction model is compared. Finally, it is proved that the proposed MEEMD-SE-ARMA hybrid algorithm can effectively improve the accuracy of the ultra short term prediction of wind power.
【学位授予单位】：沈阳农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614
，

本文编号：2105167