Probabilistic Aggregated Load Forecasting with Fine-grained Smart Meter Data

Probabilistic load forecasting (PLF) has been extensively studied to characterize the uncertainties of future loads. Traditional PLF is implemented based on the historical load data itself and other relevant factors. However, the prevalence of smart meters provides more fine-grained consumption information. This paper proposes a novel probabilistic aggregated load forecasting algorithm that makes full use of fine-grained smart meter data. It first applies clustering-based methods for point aggregated load forecasting. By varying clustering algorithms, multiple point forecasts can be obtained. On this basis, different quantile regression models are implemented to combine these point forecasts in order to form the final probabilistic forecasts. Case studies on a real-world dataset demonstrate the superiority of our proposed method.