风力发电并网的小时间尺度随机优化方法研究

The stochastic behavior of wind power has negative impact on operation cost, compensation device configuration, and voltage stability in economic dispatch and reactive power dispatch. In order to reduce the interference and further increase the wind power penetration to the grid, this project aims to propose a stochastic optimization method, which is based on a Quasi-Monte Carlo power flow simulation and a Paired-Bacteria Optimizer (PBO). The research firstly develops a small-time scale wind farm stochastic model, which is used to calculate the real/reactive power output probability density function based on the measurements of wind speed during a dispatch iteration. Weibull distributions are usually used to describe long-term wind stochastic characteristic such as months and years, which is not suitable in the applications of dispatch. Thus, this project proposes a novel stochastic model with multiple peaks in concern with the description of wind farm stochastic characteristic within the time-scale of hours. Most of conventional stochastic analysis is based on Monte Carlo simulation, which depends on power flow evaluations on a large number of samples generated using probability density function, and causes a great computation. To reduce the computational complexity, Quasi-Monte Carlo simulation is employed to generate only a small number of real/reactive output samples, which strictly follows the wind farm probability density function and covers the extreme situation, for power flow evaluation. Therefore, fuel cost, voltage stability, and wind power penetration can be comprehensively calculated accordingly. For the optimization of these targets, a PBO algorithm is introduced, which combines the conventional gradient-based algorithm and the heuristic algorithm. The optimization aims to reduce not only the mean values of the objective functions, but also their standard deviation. Thus, the solutions obtained are able to overcome the interference caused by wind power stochastic behavior with a high confidence level.

为了降低风速的随机特性对经济调度和无功优化过程中对运行成本、补偿装置配置、电压稳定性的扰动影响，提高风电渗透率，本申请提出了一种基于拟蒙特卡洛法和双粒子算法相结合的具有抗扰性的随机优化算法。研究将首先设计一种小时间尺度上的风电场随机模型，从而利用历史风速统计数据，描述时间尺度上与调度周期相一致的风电场有功出力概率。根据该模型，随机潮流分析过程创新性地通过拟蒙特卡罗方法生成一组低差异风电场出力样本，并依次进行潮流计算，获得每个样本下系统的燃料损耗、电压稳定性及风电渗透率等指标的概率。在优化过程中，经济调度和无功优化使用了传统牛顿法搜索与自启发式算法相结合的双粒子算法，能够保证优化的速度和收敛性。同时，目标函数除了包含各项指标的均值，还将引入指标的方差，以抵御风电的随机特性对优化结果的扰动，因此结果的置信度高。

为了降低风速的随机特性对经济调度和无功优化过程中对运行成本、补偿装置配置、电压稳定性的扰动影响，提高风电渗透率，本研究提出了一种基于拟蒙特卡洛法和双粒子算法相结合的具有抗扰性的随机优化算法。研究首先设计一种小时间尺度上的风电场随机模型，从而利用历史风速统计数据，描述时间尺度上与调度周期相一致的风电场有功出力概率。根据该模型，随机潮流分析过程创新性地通过拟蒙特卡罗方法生成一组低差异风电场出力样本，并依次进行潮流计算，获得每个样本下系统的燃料损耗、电压稳定性及风电渗透率等指标的概率。在优化过程中，经济调度和无功优化使用了传统牛顿法搜索与自启发式算法相结合的双粒子算法，能够保证优化的速度和收敛性。同时，目标函数除了包含各项指标的均值，还引入了指标的方差，以抵御风电的随机特性对优化结果的扰动，因此结果的置信度高。本研究的创新之处在于将首次建立小时间尺度上的风电场随机特性模型，创新性的提出了将拟蒙特卡洛算法应用于电力系统的随机优化问题，结合传统梯度法和启发式算法的优点，设计更加符合电力系统优化问题的算法。当前本项目中已经培养博士生2名，发表科研论文7篇，其中包括期刊文章3篇（SCI1，EI2）及国际会议论文4篇，专利3篇。此外，仍有2篇期刊文章正在审稿中。