Power transformer temperature field distribution inversion method with physical model support and uncertain factor influence reduction

The invention provides a power transformer temperature field distribution inversion method with physical model support and uncertain factor influence reduction, core physical characteristics are reserved by using a transformer reduced-order model, and a transformer internal temperature field inversion model is constructed based on an artificial neural network. A temperature field compensation inversion model between the temperature measured values and the predicted deviation values of the external and internal measuring points of the transformer is constructed based on support vector regression, and the influence of potential factors is eliminated. According to the invention, the two models are combined to obtain a high-precision and high-robustness inversion result. 本发明提出一种具有物理模型支持和减少不确定性因素影响的电力变压器温度场分布反演方法，运用变压器降阶模型保留核心物理特征，基于人工神经网络构建变压器内部温度场反演模型。提出基于支持向量回归构建变压器外部、内部测点温度实测值与预测偏差值间的温度场补偿反演模型，消除潜在因素影响。本发明联合两个模型获得高精度、高鲁棒性反演结果。