考虑新能源发电预测误差及其联合分布特性的电力系统随机优化理论研究

Due to the random character of renewable and sustainable energy, the accuracy of renewable and sustainable energy power generation forecast is not high. Prediction error of renewable and sustainable energy is large if we still regard the renewable and sustainable energy generation as deterministic variables when optimizing power system operation. With the rapid development of photovoltaic power generation in 2011, wind power and photovoltaic power will be connected to power system together in the future. When different renewable and sustainable energy is connected to power grid, distributions of their prediction errors influence each other and the problem of optimal operation become a multi-constrained and nonlinear programming problem with more than one random variable. This project concentrates on the distribution of prediction error. Our study includes prediction methods and data statistic for wind power and photovoltaic power to determine prediction results. After getting the distribution of prediction error, the confidence interval of prediction error can be calculated. The joint probability model for prediction errors of renewable and sustainable energy power is established by calculating the sum of different random variables. After renewable and sustainable energy power generation is connected to the grid, the influence of prediction errors on power flow and harmonic are also studied by unified modelling methods in this project. The optimization models based on the expectation programming and the chance constrained programming are established according to the stochastic optimization method. The influence of prediction error is considered in power system optimal operation including unit commitment and economic dispatch through mathematical conversion. This project provides strong support for power system optimal operation from theory and technology after different renewable and sustainable energy power is connected to power grid.

由于新能源的随机性，使得新能源发电功率预测的准确性不高。如果沿用传统处理方法，将新能源出力视为确定性变量来安排系统优化运行，会存在较大误差。伴随2011年光伏发电的迅猛发展，未来电力系统将同时消纳风力发电和光伏发电。多类新能源并网必将带来不同新能源预测误差分布的相互影响，其优化运行问题将是多随机变量、多约束的复杂非线性规划问题。本项目以预测误差分布为主线对此进行研究，主要内容包括：研究风力发电和光伏发电的预测方法机理和实际数据统计，确定预测误差的分布形式和置信区间；建立多种新能源发电预测误差的联合概率模型，即多维随机变量线性和的概率模型；分析新能源预测误差对随机潮流和谐波潮流的影响，并探讨统一建模方法；采用随机规划理论中期望值和机会约束模型，将预测误差分布通过数学变换纳入到机组组合、经济调度等系统优化基础问题中。本项目将为多种新能源发电并网下的电力优化运行提供有力的理论支持和技术支撑。

随着高比例新能源发电的快速发展，电力系统运行与规划正由确定性理论转向不确定性理论。本项目在不确定性环境下进行电力系统优化决策，分为两个部分：采用新型预测理论来刻画新能源发电预测的误差分布和生成典型场景；结合概率分布和优化理论，研究电力系统运行（侧重机组组合、经济调度和谐波分析）和规划问题。主要研究内容及研究成果包括：.(1) 提出基于k近邻算法和核密度估计的新能源出力概率预测方法，可用于刻画新能源发电预测的误差分布。围绕新能源出力统计特征，提出利用对数函数改善偏斜特征、利用边界核函数消除边界密度泄漏问题、在距离函数中引入反映特征变量重要程度的权重系数。算例结果显示，与传统方法相比，所提方法在覆盖率和区间宽度方面表现更好。.(2) 为了将概率预测应用到电力系统运行与规划中，利用多元正态分布协方差矩阵描述不同时刻之间相关性，提出了考虑时间相关性的新能源发电运行场景生成方法，开发了基于概率距离的运行场景削减算法，为随机优化提供了可靠的基础输入数据。.(3) 基于新能源发电运行场景，建立了考虑风电随机波动的机会约束机组组合模型，提出了机会约束的双线性重构方法。针对两阶段问题的可分解结构，提出了基于双线性Benders分解的快速求解算法。实际计算结果表明所提方法求解速度快，并能有效调度调节性能强的机组，以应对风电随机波动可能导致的发电能力不足。.(4) 讨论了风电出力的尖峰特性，并分析了尖峰出力特性对风电并网规划的影响，提出了基于多运行场景和期望值模型的考虑风电尖峰出力特性的大规模风电并网规划方法，并设计Benders分解方法进行求解。算例结果表明，所提方法能够显著降低新建线路投资并有效限制了风电效益损失。.(5) 从用户侧角度探讨了负荷的随机波动特性，重点是采用了简化等效模型进行含多个电动汽车充电站的概率谐波潮流计算，提出了一种充电站工作状态模拟的方法，采用蒙特卡罗法计算了各工况下电网谐波的分布，为进一步考虑源-荷随机性下系统谐波潮流奠定基础。