Energy Reports (Dec 2023)
Current status, challenges, and prospects of data-driven urban energy modeling: A review of machine learning methods
Abstract
Urban energy modeling is essential in planning electricity generation and efficiently managing electric power systems. Various urban energy models were developed for several energy-driven applications, including emission reduction, retrofit analysis, and forecasting. Electricity load forecasts help to estimate the load demand and effectively aid in power system operation and balancing. The accuracy of load forecasts at high temporal and spatial resolution can impact system planning and operation. Therefore, it is essential to know the factors that affect the accuracy of these forecasts and how they can be improved regarding the current state of the art. This article reviews the recent literature on data-driven electricity load forecasts in three steps. First, different phases of the review process are explained to select and analyze recent literature on machine learning-based short-term load forecasts. Then various aspects of load forecasting techniques have been reviewed, addressing their advantages, disadvantages, temporal resolution, and performance. Finally, the review covers the current challenges in load forecasting and describes the reasons for performance degradation and lower accuracy. Based on the reviewed literature, it was found that temperature, user load profiles, and proper management of input data highly affect load forecast accuracy. In addition, shortcomings of existing performance evaluation metrics make the applicability of those techniques questionable. Finally, we conclude the review by highlighting the necessary actions to improve load forecast accuracy that are relatively unexplored and can be used as a reference for future research on accurate load forecasts.
