Abstract Accurate wind power prediction (WPP) plays an important role in the secure operation and dispatch of power systems. This paper proposes an adaptive time-resolution method to improve the accuracy of ultra-short-term wind power prediction (USTWPP). Firstly, the hidden prediction error (HPE) with its fluctuation magnitude and rate indicators is defined to reveal the fluctuation characteristics of real-time wind power. Then, the adjustment time of time-resolution can be dynamically determined by evaluating the fluctuation magnitudes, and the adjustment rules are formulated by mining the regularities of fluctuation rate of historical wind power data and establishing the interval grouping optimization model. Finally, by coupling the adjustment time and rules into the prediction model of back propagation neural network (BPNN), the rolling prediction with adaptive adjustment of time-resolution is achieved. Extensive tests have not only demonstrated the validity of the proposed method, but also confirmed its capability to cope with the USTWPP especially under situations of extremely violent fluctuations of wind power.

中文翻译：

---

一种超短期风电功率预测的自适应时间分辨率方法

摘要 准确的风电功率预测（WPP）对电力系统的安全运行和调度具有重要作用。本文提出了一种自适应时间分辨率方法来提高超短期风电功率预测（USTWPP）的准确性。首先，定义隐藏预测误差（HPE）及其波动幅度和速率指标，以揭示实时风电的波动特征。然后，通过评估波动幅度动态确定时间分辨率的调整时间，并通过挖掘历史风电数据波动率的规律，建立区间分组优化模型，制定调整规则。最后，通过将调整时间和规则耦合到反向传播神经网络（BPNN）的预测模型中，实现了时间分辨率自适应调整的滚动预测。大量的测试不仅证明了所提出方法的有效性，而且证实了其应对 USTWPP 的能力，尤其是在风功率波动极其剧烈的情况下。