Designing the input vector to ANN-based models for short-term load forecast in ...

• Author(s): Santos, P. J. ; Martins, A. G. ; Pires, A. J.
• Date: 2007
• Persistent ID: http://hdl.handle.net/10316/4057
• Origin: Estudo Geral - Universidade de Coimbra
• Subject(s): Electric energy systems; Distribution networks; Load forecast; Artificial neural networks
• Document Type: Article
• Language: English

Description

The present trend to electricity market restructuring increases the need for reliable short-term load forecast (STLF) algorithms, in order to assist electric utilities in activities such as planning, operating and controlling electric energy systems. Methodologies such as artificial neural networks (ANN) have been widely used in the next hour load forecast horizon with satisfactory results. However, this type of approach has had some shortcomings. Usually, the input vector (IV) is defined in a arbitrary way, mainly based on experience, on engineering judgment criteria and on concern about the ANN dimension, always taking into consideration the apparent correlations within the available endogenous and exogenous data. In this paper, a proposal is made of an approach to define the IV composition, with the main focus on reducing the influence of trial-and-error and common sense judgments, which usually are not based on sufficient evidence of comparative advantages over previous alternatives. The proposal includes the assessment of the strictly necessary instances of the endogenous variable, both from the point of view of the contiguous values prior to the forecast to be made, and of the past values representing the trend of consumption at homologous time intervals of the past. It also assesses the influence of exogenous variables, again limiting their presence at the IV to the indispensable minimum. A comparison is made with two alternative IV structures previously proposed in the literature, also applied to the distribution sector. The paper is supported by a real case study at the distribution sector. http://www.sciencedirect.com/science/article/B6V2T-4MCW9X5-1/1/00590212b5295357d45465c710d645ae