LAUSR

Load data is an essential input for power system analysis and management. However, traditional load analysis methods do not fit well with emerging deep learning models that may require image matrix as input, such as Convolutional Neural Network (CNN) and Generative Adversarial Nets (GAN). This article proposes a novel analysis method “Load Photo” to create the required image matrix for various loads in power systems by using the HSV color space. A load photo is a 2-D pixel graphical representation of load, in which the x-axis represents sampling time points (e.g., hours) for every day from Sunday to Saturday, the y-axis represents weeks from the first week till the last week, and the pixel color demonstrates the normalized load. A load photo stores all essential load-related parameters such as start time, end time, and sampling period. A load photo can be used to characterize the energy consumption pattern of any load type, just like taking a photo for the load that clearly demonstrates the load’s behavior. Furthermore, a load photo could work before feature selection as an input to deep learning models. Accordingly, the latest advancement in deep learning models related to image processing can be easily applied to the analysis and data mining of load data.