一种风电功率混沌时间序列概率区间简易预测模型

A simple model for probabilistic interval forecasts of wind power chaotic time series

本文基于极限学习机构建了一种简易模型以直接输出风电功率概率区间.同时,为优化模型训练过程中输出区间的性能,本文基于对数据集区间带偏差信息的分析构建了一种新的优化准则,并采用量子细菌觅食优化算法以获取问题的最优解,提高模型泛化能力.对比分析两个风电场在不同置信水平和不同优化准则下的概率预测结果,仿真表明本文模型具有更高的可靠性和更窄的区间带宽,可为风电并网安全稳定运行提供决策支持.

Integration of wind power into grids requires accurate forecasting, however, traditional wind power point forecast errors are unavoidable and they cannot be eliminated due to the highly volatile and uncertain in the chaotic time series of wind power. Unlike point prediction, which conveys no information about the prediction accuracy, probabilistic interwl forecasts can provide a range, within which the target will lie with a certain probability, for estimating the potential impacts and risks facing the system operation. Most existing prediction interval （PI） construction methods are often placed after a deterministic forecasting model with or without prior assumptions, this paper propose a novel lower-upper bound estimation approach using extreme learning machine to directly construct PIs for wind power series. Based on the analysis of the interval forecasting error information in training dataset, a new problem formulation is developed in this method to get better PIs. In addition, in order to obtain the global optimal solution of the above model, a quantum bacterial foraging optimization algorithm is proposed by introducing the theory of quantum mechanics into bacteria foraging behavior. The testing results from two real wind farms with different confidence probability and optimization criterion demonstrate the excellent quality of PIs in terms of both reliability and sharpness, which provide a support for the steady operation of power system with wind power integration.