Abstract

Electrical load forecasting is crucial to achieving better efficiency, reliability, and power quality in modern power systems. Applying short‐term load forecasting, a balance can be preserved between supply and demand; the cost of electricity production will also be decreased. Several methods are proposed for short‐term load forecasting in smart grids in recent years and each of them has its own advantages and weaknesses. Among these methods, popularity is increasing in machine learning techniques. This study is to review three common artificial intelligence load forecasting methods, including long short‐term memory, group method of data handling, and adaptive neuro‐fuzzy inference system that have been used to forecast load in a smart grid consisting of a photovoltaic, wind turbine, battery energy storage system, and electric vehicle charging stations. The performance of these methods is evaluated given accuracy and the system's hardware requirements. The noisy condition of the system is also investigated when these methods are used for load forecasting. The results show that the long short‐term memory model is more accurate than the group method of data handling and adaptive neuro‐fuzzy inference system models. However, this method requires much better hardware requirements.