Winding hot-spot temperature prediction for oil-immersed power transformer based on Kalman filter algorithm

作者：Su Xiaoping^*; Chen Weigen; Xi Hongjuan; Zhou Qu; Pan Chong; Qian Guochao

来源：High Voltage Engineering, 2012, 38(8): 1909-1916.

DOI：10.3969/j.issn.1003-6520.2012.08.014

• Grating sensor
• Hotspot temperature
• Kalman filter algorithms
• Kalman filter model
• Mean absolute percentage error
• Measurement equations
• Oil-immersed power transformers
• Predicting models
• Real-time estimation
• Root mean square errors
• State equations
• Statistical information
• Stochastic difference systems
• Thermal-dynamic effects
• Top oil temperature

We proposed a differential equation for predicting the hot-spot temperature (HST) after taking the thermal dynamic effect of load and the effects of top-oil temperature variations into account. Meanwhile, we employed the kalman filter method to establish a real-time estimation model for hot-spot temperature by means of state equations and measurement equations. This kalman filter model was applied to both a transformer set-up and a real oil-immersed power transformer in service. The experimental temperatures which were measured by fiber bragging grating sensors were used to verify the model's filter and interpolation performance through simulating the monitored data. Meanwhile, based on limited experimental data and the statistical information of the measured error, the proposed model was extrapolated so as to evaluate the hot spot temperature timely, and the HSTs were compared with those predicted by the classical IEEE-Annex G model. Results show that, the HSTs obtained by the proposed model are closer to the online monitored values, and the yielded mean absolute percentage error and root mean square error are both superior to those obtained by the IEEE model. The potential applicability and generality in prediction of HSTs indicate that proposed model provides a useful tool for transformer operators to monitor hot-spot temperature.

• 出版日期2012