A Distributed Short-Term Load Forecasting Method Based on Local Weather Information

Abstract

A centralized forecasting method is difficult to accurately follow load variation and weather diversity throughout the region in a bulk power system that covers a large geographical area. A distributed load forecasting method based on local weather information is proposed in this paper. First, the bulk power system is partitioned into some subnets based on local weather information. Second, separate forecasting models are established for subnets. These models are selected from load forecasting model base, which includes neural network, autoregressive integrated moving average model, autoregressive and moving average, gray model, and so on. Cosine distance is used to evaluate the similarity between vectors of influencing factors, so that representative samples can be selected from large data sets as training set for local load forecasting models. Finally, a system load forecasting model is developed to aggregate local load forecasts. Case study shows the advantages of the proposed method in a bulk power system.