Inverter-based distributed generation (IBDG) sources are continually being connected to the grid. Such IBDG sources act as current sources that are unbalanced in either one or two phases. When connected to the grid, unbalanced IBDG sources inject the unbalanced current during an electric fault. Moreover, unbalanced load currents can also change the short-circuit current. Therefore, short-circuit analyses of power systems with either clustered or distributed IBDG sources should not ignore the contribution from unbalanced IBDG sources and load currents. Thus, the objective of this study is to present a method that calculates the short-circuit current flowing from unbalanced IBDG sources. For this purpose, zero- and negative-sequence networks are referred to the positive-sequence network and new bus impedance matrices are defined for each sequence network. This study also proposes a hybrid method that combines Thevenin's equivalent circuit and the three-phase bus impedance matrix for distribution systems with untransposed lines. The method examines the effect of unbalanced loads on the short-circuit current. The proposed methods are verified by using well-known IEEE test feeders.

This editorial overviews the contents of the Special Issue “Machine Learning for Energy Systems 2021” and review the trends in machine learning (ML) techniques for energy system (ES) optimization

Accurate power load forecasting is critical to maintaining the stability and efficiency of power systems. However, due to the complex and fluctuating nature of power load patterns, physical calculations are often inefficient and time‐consuming. In addition, traditional methods, known as statistical learning methods, require not only mathematical background and understanding but also statistical background and understanding. To overcome these difficulties, the authors proposed a simpler way to predict load by using artificial intelligence. This study investigated the performance of forecasting techniques, including three single‐layer and seven hybrid multilayer deep learning models that combine them. This study also analyzed the effect of hyperparameters on the learning results by varying the epoch and activation functions. To evaluate and analyze the performance of the deep learning model, this study used load data from the power system in Jeju Island, Korea. In addition, this study included weather factors that may affect the load to improve the prediction performance. The prediction process is performed on the Python platform, and the model that showed the highest accuracy was LSTM‐CNN, a hybrid combination of LSTM and CNN models. Considering both the maximum and minimum error, the error value was low at 0.231%. By providing detailed insights into the entire research process, including data collection, preprocessing, scaling, prediction, and analysis, this study provided valuable guidance for future research in this area.