基于灵敏度与相关性的综合负荷模型参数优化辨识策略

Optimized Identification Strategy for Composite Load Model Parameters Based on Sensitivity and Correlation Analysis

综合负荷建模及其参数辨识是电力系统分析研究的重点和难点之一。为了研究综合负荷模型参数对模型响应的影响以及参数之间的动态关联,提出一种基于参数灵敏度与相关性分析的综合负荷模型优化辨识策略。首先,对配电网集结等效的综合负荷模型进行解析灵敏度分析以及Hessian矩阵特征值表征的灵敏度分析;其次,通过Pearson相关系数判断参数一阶灵敏度之间的相关性,得到对输出响应作用相似的参数;然后,在遗传算法与Levenberg-Marquardt算法相结合的混合算法基础上,提出固定灵敏度小的参数,按比例简化辨识线性相关参数的优化辨识策略;最后,通过实测曲线及完整负荷模型的仿真,验证了该辨识策略的有效性。

Load modeling and parameter identification are important and difficult in power system analysis. To study the correlations between parameters and the effects of parameters on model response, an optimized identification strategy based on parameter sensitivity and correlation analysis is presented. Firstly, parameter sensitivities are analyzed based on analytic sensitivity and eigenvalues of Hessian matrix. Secondly, through the Pearson correlation coefficient, the correlations among 1st order sensitivity are identified to select the parameters that have the similar effects on model responses. Then, fix the parameters with small sensitivities and simplify the identification of the linearly dependent parameters proportionately, using the hybrid algorithm combined genetic algorithm with levenberg-marquardt algorithm. Finally, efficiency of the optimized strategy is verified by comparing the measure data with the simulations of the integrated load model.