最小最大概率回归机在短期风电功率预测中的应用

Application of the Minimax Probability Machine Regression in Short-term Wind Power Forecasting

为了对短期风电功率的预测进行研究,提出了一种基于最大最小概率回归机(MPMR)的预测方法。MPMR方法是将最小最大概率分类机(MPMC)向回归问题的应用推广。该方法仅须假定产生预测模型的数据分布均值与协方差矩阵已知,即能够最大化模型的预测输出位于其真实值边界内的最小概率。验证试验表明,MPMR方法能更好地跟踪风电功率的变化,有效地提高风电功率的预测精度,具有很好的应用前景。

In order to research the forecasting of short - term power of wind generation, the forecasting method based on minimax probability machine regression （ MPMR ） is proposed,in which the minimax probability machine classification （MPMC） is extended to be used in promotion of regression. Only by presuming the mean and covariance matrix of data distribution that produces the forecasting model is known, the minimum probability for obtaining forecasting output of the model within the boundary of true value can be maximized. The verifying experiments indicate that the MPMR method can well track the variation of the power of wind generation and effectively enhance the forecasting accuracy, so it possesses good applicable potential.