Recurrent Neural Network

Overview
Cite Work

ABSTRACT

The quest for effective forecasting models to meet the ever growing electricity demand around the globe has heightened and taken a much challenging dimension that cuts across rural and urban communities in under-developed, developing and developed economies. Electric Load Forecasting which is the accurate prediction of both the magnitude and geographical distributions of electrical energy demand over the different periods of planning horizon has become a vital process in the planning and operation of electric utility, electricity supply, system operations and other market activities because of the significance of electricity in modern day economy (Oamek and English, 1984; Hippert et al. 2001, Alfares and Mohammad 2002, Rafal 2006). An often quoted estimate suggests that an increase of 1% in forecasting error could imply a ten million Pounds (or equivalent) increase in operating costs per year (Hippert et al, 2001, Alfares and Mohammad, 2002). Thus, it is a key area of research in electrical energy with the aim of reducing the errors.

Furthermore, load forecasting is considered a difficult task for three main reasons. Firstly, the system load consists of several thousands of individual components (consumption agents) which are independent of each other and whose parameters vary. Consequently, secondly, the load time series exhibit variability on daily, weekly and annual time scales due to the nature of the consumption agents. Thirdly, many exogenous variables like weather conditions, social events, economic factors, demographic features etc, which may be difficult to capture, affect consumption pattern.

Read Download

Recurrent Neural Network

Related Works