Loading control and thermal protection for power transformers using neural networks

The real need for advanced power system automation is associated with (i) the growing demand for reliable and flexible power supply systems, and (ii) the desire for optimised network operation in both normal and emergency conditions. In order to enhance system capability a new control and protection system, based on identifying Artificial Neural Networks (ANNs) is proposed for on-line predictive computation of dynamic thermal rating of power transformators, which is up-to-dated considering real heat exchange conditions, lifetime of the insulation, network conditions. The scontrol and protection system, on the base of winding hot-spot, load current and environmental conditions provides the winding hot-spot temperature trend as the transformer load current and environmental conditions vary. The system design and ANN learning process have been optimised and validated in relation to a real transformer. Because of its characteristics the system can be used by plant operators to fix acceptable transformer loads and overload limits during contingencies. The greater accuracy of the ANN thermal model compared to the traditional thermal models shown in the load guides undoubtedly allows to increase the exploitation margins of the power transformers.