基于改进气象聚类分型的短期风电功率概率预测方法

Short-term Wind Power Probability Forecasting Method Based on Improved Meteorological Clustering and Classification

精准的风电功率预测是对以新能源为主体的新型电力系统安全高效运行的重要支撑。针对现有预测方法未充分考虑不同气象条件下风电出力特性差异的问题,提出基于改进秃鹰搜索-最大期望算法的高斯混合模型(IBES-EM-GMM)聚类与气象分型的风电功率概率预测方法。首先,基于Levy飞行与自适应t分布变异策略改进秃鹰搜索算法,提出IBES-EM-GMM聚类模型,以增强其全局搜索能力。以此为基础,利用IBES-EM-GMM聚类模型对历史气象-功率数据集进行聚类划分,并采用混合深度神经网络与Cornish-Fisher级数分别训练不同气象模式的数据集以得到其概率预测结果。最后,选取中国冀北地区风电场实际数据进行算例仿真。结果表明,相较于无聚类和高斯混合模型聚类方法,所提IBES-EM-GMM聚类模型使聚类效果、风电功率点预测与概率预测精度均得到了显著提升。

Accurate wind power forecasting is a critical support for the safe and efficient operation of the renewable-energy-based power system.Aiming at the problem that the existing forecasting methods do not fully consider the difference in wind power output characteristics under different meteorological conditions,a wind power probability forecasting method based on Gaussian mixture model(GMM)clustering with improved bald eagle search(IBES)-expectation maximum(EM)algorithm(IBES-EMGMM)and meteorological classification is proposed.First,based on Levy flight and adaptive t-distribution mutation strategy,the bald eagle search algorithm is improved,and GMM clustering model based on the IBES-EM algorithm is proposed to enhance the global search ability.Based on this,the IBES-EM-GMM clustering model is applied to cluster the historical weather-power data set,and the hybrid deep neural network along with Cornish-Fisher series is used to train the data sets with different meteorological patterns to obtain their probability forecasting results.Finally,the actual data of wind farms in Jibei,China are selected as an example for simulation.The results show that compared with the methods without clustering and with GMM clustering,the proposed IBES-EM-GMM clustering model leads to a significant improvement in clustering effect and the accuracy of point and probability forecasting for wind power.