The key technology for grid integration of wind power : direct probabilistic interval forecasts of wind power

Zhao Xu, C. Wan

Research output: Journal article publicationJournal articleAcademic research

Abstract

The wind power is an important renewable energy,but it has features of high volatility and uncertainty,therefore a large scale integration of wind power into power system will impose significant challenges in system operation. Accurate wind power prediction is one of the key technologies to reduce the risk of its grid integration. Because of the nonstationarities and nonlinearities of wind power series,traditional point prediction methods cannot provide satisfactory prediction results. In contrast,probabilistic interval based wind power forecasting techniques can simultaneously quantify the prediction error and the associated probability,thereby can more effectively support power system operation to cope with various uncertainties and risks. This paper firstly summarizes the latest developments in wind power forecasting techniques,then proposes an Extreme Learning Machine( ELM) and evolutionary computation based method to directly generate wind power prediction intervals. Compared to the existing methods,the advantage of the proposed method is able to directly generate prediction intervals through one optimization process,thus to largely simplify the model construction and avoid prediction errors analysis. The proposed method has been tested with practical wind farm data in Denmark,and the results demonstrate that it can efficiently and accurately provide probabilistic prediction intervals of wind power.||风电是一种重要的可再生能源,但风电具有的高波动性和随机性使其大规模并网运行面临各种困难和挑战。准确的风电功率预测是减少风电并网风险的关键技术之一。由于风电功率时间序列的高度波动性,传统的基于点预测方法无法提供可靠的风电功率预测结果。基于概率区间的风力发电预测技术能够同时量化预测误差和相关概率,从而降低由于预测误差带来的各种风险,可以更有效地支持电力系统应对各种不确定性和风险。首先总结风电功率预测技术的最新发展,然后提出了一个基于超级学习机和进化计算的方法直接生成风电预测区间。相较于已有的方法,所提出的算法优点在于能够直接通过一次性优化过程产生预测区间,从而在保证高有效性的前提下简化了模型和计算量,避免了传统方法中包含的误差数据分析等高计算量的步骤。通过丹麦实际风电场数据对所提出的方进行了各种测试,结果表明该方法能够高效和准确地提供风电功率概率预测区间。
Original languageEnglish
Pages (from-to)1-9
Number of pages9
Journal南方电网技术 (Southern power system technology)
Volume7
Issue number5
Publication statusPublished - 2013

Keywords

  • Wind power
  • Probabilistic interval forecast
  • Extreme learning machine
  • Evolutionary computation

Cite this