Dynamic Loading of Substation Distribution Transformers: Detecting Unreliable Thermal Models and Improving the Accuracy of Predictions.

摘要

The existing deregulated market structure for electricity necessitates that utilities make the generation, transmission and distribution of electricity cost-effective. This encourages investment in technological upgrades to utilize the equipment optimally, thus reducing the operation and maintenance costs, while ensuring an extended operational life. This goal can be achieved for a transformer with the help of dynamic loading.;Dynamic loading of a transformer implies optimally loading it given available load, cooling and ambient conditions. This can be of significance in maintaining the reliability of the electric supply. Dynamic loading allows the utility to load a transformer above its nameplate rating for a specified duration of time, such that its service life is not unduly reduced. Overheating is more often than not the cause behind premature insulation breakdowns and insulation breakdowns often lead to overhaul or replacement of transformers. Hottest-spot temperature (HST) and top-oil temperature (TOT) are reliable indicators of the insulation temperature. The objective of this project is to use thermal models to estimate the transformer's maximum dynamic loading capacity without violating the HST and TOT thermal limits set by the operator. In order to ensure the optimal loading, the temperature predictions of the thermal models need to be accurate. A number of transformer thermal models are available in the literature. In present practice, the IEEE Clause 7 model is used by the industry to make these predictions. However, a linear regression based thermal model has been observed to be more accurate than the IEEE model. These two models have been studied in this work.;This document presents the research conducted to discriminate between reliable and unreliable models with the help of certain metrics. This was done by first eyeballing the prediction performance and then evaluating a number of mathematical metrics. Efforts were made to recognize the cause behind an unreliable model. Also research was conducted to improve the accuracy of the performance of the existing models.;A new application, described in this document, has been developed to automate the process of building thermal models for multiple transformers. These thermal models can then be used for transformer dynamic loading.