Short-term multinodal load forecasting in distribution systems using general regression neural networks

Abstract

Multinodal load forecasting deals with the loads of several interest nodes in an electrical network system, which is also known as bus load forecasting. To perform this demand, it is necessary a technique that is precise, trustable and has a short-time processing. This paper proposes two methodologies based on general regression neural networks for short-term multinodal load forecasting. The first individually forecast the local loads and the second forecast the global load and individually forecast the load participation factors to estimate the local loads. To design the forecasters it wasn't necessary the previous study of the local loads. Tests were made using a New Zealand distribution subsystem and the results obtained are compatible with the ones founded in the specialized literature. © 2011 IEEE.