改进蚁群算法及其在变压器绝缘介质响应电路参数辨识中的应用

Improved Ant Colony Algorithm and Its Application in Parameter Identification for Dielectric Response Equivalent Circuit of Transformer

回复电压测量法是一种有效诊断变压器绝缘状态的无损检测方法,建立油纸变压器介质响应等值电路模型是分析绝缘老化与特征量之间关系的有效途径。为此,在建立等值电路数学模型的基础上,将等值电路参数辨识问题转化为数学优化问题,并采用改进蚁群算法进行了等值电路模型参数辨识。最后通过4种测试函数验证了:蚁群算法中将全局最优解结合当次迭代最优解进行全局信息素更新,能够在保证快速收敛的同时,扩展算法搜索范围,而动态信息素局部更新也增强了算法摆脱局部最优的能力。模型计算与实际测量的极化谱对比也说明了改进算法求解的等值模型能准确地反映油纸绝缘变压器的介质响应特性。

Return voltage measurement is an effective non-destructive testing method for diagnosis of transformer in- sulation state. Establishing transformer dielectric response equivalent circuit model is an effective way to analyze the relationship between insulation aging and characteristic quantity. Therefore, on the basis of establishing equivalent circuit mathematical model, equivalent circuit parameters identification was transformed into the mathematical opti- mization, then the improved ant colony algorithm was adopted to identify model parameters. Four test functions proved that using both global optimal solution and each iteration optimal solution in global pheromone updating could improve convergence speed, expand search scopet and using the dynamic local pheromone updating could also strengthen the ability of jumping out of local optima! value. The contrast of polarization spectra between model cal- culation and practical measurement shows that the solved equivalent model can accurately reflect the medium re- sponse characteristics of oil-paper transformer insulation.