The Efficient Creation Of Datasets For Data Driven Power System Application

Abstract

The development of data driven power system applications requires large datasets with high quality data. However, the creation of such datasets can be time consuming and expensive. This paper proposes an efficient approach for creating datasets that can be used for data driven power system applications. The proposed approach is based on the concept of data augmentation. Data augmentation involves generating new data samples from the existing data samples by applying various transformations. The idea is to use a small set of original data samples and generate a large number of augmented data samples. The augmented data samples can then be used to create a dataset that is representative of the actual data distribution. The approach involves three main steps: Data selection, data augmentation and dataset creation. In the data selection step, a small set of representative data samples is selected from the available data. In the data augmentation step, various transformations include rotation, scaling, translation, and noise addition. In the dataset creation step, the augmented data samples are combined with the original data samples to create a final dataset. The proposed approach was tested on two datasets, IEEE 14 bus and IEEE 118 bus datasets. The results showed that proposed approach was able to create dataset that were representative of the actual data distribution. The dataset created using the proposed approach were used to train and evaluate a data driven power system application for voltage stability assessment. The results showed that the application achieved high accuracy in preaching the voltage stability of the power system. In conclusion, the proposed approach provides in efficient and cost effective method for creating datasets for data approach is based on the concept of data augmentation, which involves generating new data samples from the existing data samples by applying various transformations. The approach was tested on two datasets and showed promising results. The approach can be extended to other domains where the creations of large dataset are required.