采用热路模型参数辨识的油浸式配电变压器绕组温升预测方法

To address ihe issues of long temperature rise time and low efficiency in oil-immersed distribulion transformers, a method for predicting winding temperalure rise in such transformers using thermal circuit model parameter identification is proposed. Based on the principle of thermal-electrical analogy, a fifth-order thermal circuit model for oil-immersed distribulion transformers was established, which includes losses from the core and both high-voltage and low-voltage windings. To obtain the thermal parameters, an extended Kalman filter algorithm was introduced to achieve dynamic identification of the thermal circuit parameters by integrating the fifth-order thermal circuit model with 2 h of measured data. Numerical solutions were used lo calculate the temperature rise curves of the top oil and the high-and low-voltage windings, achieving the goal of temperature rise prediction. The results show that the identified thermal capacitances for the low-voltage winding, high-voltage winding, oil, and lank of a 400 kV • A transformer are approximately 66.1.110.1, 255.7. and 125.6 kJ • K^-1. respectively, while the ihermal resistances for the low-voltage winding-oil. high-voltage winding-oil. oil-tank, and tank-air are approximately 8.76, 6.61, 0.56, and 10.96 mK • W^-1, respectively, with fluctuations in each parameter being less than 10%. The ihermal resistance between the high-voltage winding and oil significantly decreases with increasing transformer capacily, dropping from 6.61 mK • W^-1 at 400 kV • A to 4.00 mK • W^-1 at 800 kV • A. Under rated loss power, the maximum relative error in predicting the steady-state temperalure rise of the top oil for 400, 630, and 800 kV • A transformers is 4.86%, with a maximum absolute error of 2.6 K. Prediction results under different loss power levels indicate that the prediction accuracy under 1.0 limes the total loss power is below 2 K. while under 0.5 limes ihe total loss power, nonlinear parameter drift at low temperalure rise leads lo an increased error of up lo 7.3 K. The proposed method can be used for rapid temperature rise detection and assessment of oil-immersed distribution transformers.