The accuracy of the load model has great effects on power system analysis andcontrol. However, load modeling has been a well-known difficult problem and unsolvedso far, which is mainly due to the fact that the load is always changing, both in itsamount and its constitutes. Since the load model can only be built on the recordedmeasurements, the generalization capability of the model has great effects on its validity.There are problems coming from the practice of using the model in power systemanalysis. This paper proposes a novel method to analysis the Statistical characteristics ofthe load by applying the Support Vector Machine in Statistical Learning Theory. Afeature load data-space is set up first by applying the Support Vector Machine Tool. Thenthe TVA load model is built based on this feature space. Although the load model thusbuilt is based only on a very small load data subspace, it has a strong generalizationcapability.This analysis shows that the load space is low dimension and how to use themeasured data. Case studies on all load data recorded in north-east grid have shown theeffectiveness of the proposed method.
|