基于径向基神经网络的可用输电能力概率分析

Probabilistic analysis for available transfer capability based on radial basis function neural network

可用输电能力(ATC)是电力系统经济安全运行的一项重要指标。由于大量风电并网和用户用电行为的多样化,ATC的计算必须要考虑其带来的不确定源。而在对不确定源相关性的处理时,Nataf变换中标准正态分布域相关系数的求解尤为复杂,传统的基于辛普森数值积分和二分法的相关系数转换法耗时极其严重。在概率计算中,采用蒙特卡洛法基于最优潮流的ATC计算模型实现ATC的概率分析耗时也极其严重。基于以上困难,本文采用径向基神经网络实现Nataf变换中相关系数的转换过程,同时采用径向基神经网络近似ATC的确定性计算模型,从概率建模和单次ATC的计算量两个方面对ATC的概率分析进行提速。本文基于两个相连接的IEEE 9节点算例,以传统相关系数转换法作为相关系数精度参考,以基于最优化ATC模型的蒙特卡洛法作为概率结果精度参考,测试了本文所提方法的性能。

Available transfer capability(ATC)is an important indicator for the economic and security operation of power systems.Due to the vast integration of wind power and the variety of power consumption behavior,the uncertainty of them must be taken into consideration during the calculation of ATC.For the correlation of the uncertainty sources,it is a very time-consuming step to use Simpson’s numerical integral and bisection approach to calculate the correlation coefficients in standard normal distribution scope during the Nataf transformation.Furthermore,there exists huge calculation burden to achieve the Monte Carlo simulation method based probabilistic analysis for ATC based on the optimal power flow ATC calculation model.To address these issues,radial basis function neural network(RBFNN)is adopted for the calculation of correlation coefficients in standard normal distribution scope,and RBFNN is further used for constructing the deterministic ATC model for accelerating the probabilistic analysis.In this paper,a power system including two connected IEEE 9-bus systems is adopted for testing the performance of the proposed method,where the traditional correlation coefficient calculation method and Monte Carlo simulation method are treated as the reference.