低保守性自适应鲁棒优化及其在含大规模风电电网调度中的应用

Due to the significant uncertainty and variability of wind energy, its rapid growth brings great challenges for the secure and economic operation of power systems. It is critical to enhance the ability of system scheduling under uncertainty for handling the wind power integration problem. The existing stochastic optimization approach is computationally intensive and depends on the accurate probability distribution of the uncertainty. Although the computable robust optimization approach is independent of the probability distribution, it always faces the challenge on its over conservatism. This project aims to propose a less conservative adaptive robust optimization method and applies it to the unit commitment and economic dispatch (UC-ED) problem. Its less conservativeness depends on well-designed uncertainty sets and robust models, and its adaptability refers to the dynamic response to uncertain wind energy. To construct proper uncertainty sets with probability guarantee, we present a data-driven method and apply it to characterizing the correlation of wind uncertainty. Based on the constructed uncertainty sets, a probability based less conservative robust model is proposed and used to solve the UC-ED problem. The project will further design policy based multi-level solution method. The proposed models and methods will be validated by data from the northeast power grid. The research results will provide a set of secure and efficient decision support for power systems scheduling, and enhance the ability of power systems to integrate wind energy.

具有强不确定性和波动性的风电大规模并网，对电网安全经济运行带来了新挑战。消纳大规模风电的一个重要途径在于提升电网应对不确定性的能力。现有随机调度方法计算复杂，并依赖于风电功率的精确概率分布函数；而可计算性强、不依赖于精确分布函数的鲁棒方法往往过于保守。为此，项目提出一种低保守性自适应鲁棒优化方法，并应用于电网机组组合和经济调度问题。低保守性来自于不确定性集的精炼和新型鲁棒模型的构建；自适应性在于调度决策对不确定性风电功率的动态响应。针对不确定性的表示问题，提出数据驱动的不确定性集构造方法，给出其概率保证，并应用于不确定性风电功率的表示。利用具有概率保证的不确定性集，提出融合概率信息的低保守性自适应鲁棒模型，并应用于机组组合—经济调度问题，进而研究基于策略的分层算法。最后在东北电网调度决策中进行应用验证。项目研究成果将为电网调度提供安全可行、实用有效的决策支持，提升电网消纳大规模风电的能力。

大规模风电的消纳已成为影响我国新能源发展和电网安全的关键因素，而不确定性环境下的电网调度能力的提升是减少弃风量、大规模消纳风电的重要途径。本项目旨在通过对低保守性自适应鲁棒优化理论和方法的研究，为不确定性风电调度提供安全可靠、可行有效的技术方法。项目从不确定性参数描述、不确定性风电调度优化建模与分析和随机鲁棒优化算法设计三个方面展开了研究。具体研究成果包括提出一种数据驱动的椭球不确定性集构造方法。该方法可以在不同置信水平下给出一系列具有概率保证的不确定性集合；研究了同时包含不确定性风功率水平和随机负荷的两阶段虚拟发电厂调度优化问题，给出了最优调度策略。研究表明风功率预测精度与系统运行总成本成反比例关系，然而当预测精度达到一定水平后，预测精度提高带来的边际收益递减。针对基于矩信息的随机鲁棒优化模型，提出了一种基于参数搜索的精确、高效算法-Parametric Search算法。该算法在期望意义下，具有多项式时间的计算复杂度，并可将目前相关算法计算效率提升几个数量级。课题组在Production and Operations Management、INFORMS Journal on Computing、European Journal of Operational Research等国内外学术期刊发表SCI论文10篇，EI期刊论文1篇，EI会议论文3篇。