基于流线分析的10kV油浸式变压器绕组热点温度反演模型建立及验证研究

Establishment and Verification of 10kV Oil Immersed Transformer Winding Hot Spot Temperature Inversion Model Based on Streamline Analysis

准确获取运行变压器绕组热点温度一直是电力行业的难点问题之一。源自变压器绕组热源区域的热流是导致变压器外场域温度升高的直接推动力,该文基于变压器温度流体场耦合数值计算,分析变压器内部热流扩散规律,通过提取流经绕组热点区域与外壳散热区域的典型热流流线,选取变压器外部可测并与绕组热点温度具有强关联关系的特征测温点,并采用网格搜索法优化的支持向量回归机构建特征测温点温度与绕组热点温度间的多维非线性关系,进而建立变压器绕组热点温度反演检测模型。为拓宽反演模型的普适性,采用正交设计法构建反演模型所需的训练样本和测试样本,采用变压器多工况温升试验验证变压器温度流体场仿真结果的准确性,并验证绕组热点温度反演模型的准确性,绕组热点温度反演平均绝对误差为1.82%,最大温差小于3℃。

Accurately obtaining the hot spot temperature of operating transformer winding has always been one of the difficult problems in the power industry.The heat flow from the heat source area of transformer winding is the direct driving force leading to the temperature rise in the field outside the transformer,based on the coupled numerical calculation of transformer thermal fluid field,the heat flow diffusion law inside the transformer is analyzed in this paper.Besides,by extracting the typical heat flow lines flowing through the winding hot spot area and the shell heat dissipation area,the characteristic temperature measurement points that can be measured outside the transformer and have a strong correlation with the winding hot spot temperature are selected.The multi-dimensional nonlinear relationship between the temperatures of characteristic temperature measurement points and winding hot spot temperature is established by using the support vector regression machine optimized by grid search method,and then the inversion detection model of transformer winding hot spot temperature is established.In order to broaden the universality of the inversion model,the training samples and test samples required by the inversion model are constructed by using the orthogonal design method.In this paper,the accuracy of the simulation results of transformer temperature fluid field is verified by the multi working condition temperature rise test of transformer,and the accuracy of the winding hot spot temperature inversion model is verified.The average absolute percentage error(MAPE)of winding hot spot temperature inversion is 1.82%,and the maximum temperature difference is less than 3℃.