Literature review for large-scale load forecasting with large volume of smart-meter data

Large-scale short-term load forecasting involves predicting energy consumption across geographic areas or large sets of users. This practice is crucial in power systems, particularly for energy suppliers. The widespread installation of smart metering technology has facilitated the collection of extensive data on user load profiles. By incorporating such granular data, large-scale load forecasting becomes more accurate and reliable, capturing the variability and trends across different consumer segments. However, transforming smart-meter data into effective load forecasting models faces significant challenges. For instance, smart-meter data present issues related to its high volume, variety, and fine-grained temporal resolution. Consequently, different techniques can be considered to mitigate these issues before applying forecasting models to the data. In this paper, we conduct a two-step literature review to provide insights into the data, forecasting approaches, and model evaluation used in large-scale, short-term load forecasting from smart-meter data. We propose classifying the different forecasting approaches into three strategies: integrated, residential-based, and cluster-based. We additionally draw insights on the effect of data volume on forecasting models, performance comparison between forecasting approaches, and the tendency of model complexity in this research domain.