Please use this identifier to cite or link to this item:
Title: A real-time short-term load forecasting system using functional link network
Authors: Dash, P.K.
Satpathy, P.
Liew, A.C. 
Rahman, S.
Issue Date: 1997
Citation: Dash, P.K.,Satpathy, P.,Liew, A.C.,Rahman, S. (1997). A real-time short-term load forecasting system using functional link network. IEEE Power Engineering Review 17 (5) : 59-. ScholarBank@NUS Repository.
Abstract: The short-term load forecast (one to twenty four hours) is of importance in the daily operations of a power utility. It is required for unit commitment, energy transfer scheduling and load dispatch. With the emergence of load management strategies, the short-term load forecast has played a broader role in utility operations. The development of an accurate, fast and robust short-term load forecasting methodology is of importance to both the electric utility and its customers. Many algorithms have been proposed in the last few decades for performing accurate load forecasts. The most commonly used techniques include statistically based techniques, expert system approaches and artificial neural network algorithms (ANN). The time series and regression techniques are the two major classes of conventional statistical algorithms, and have been applied successfully in this field for many years. The expert systems based algorithm for short-term load forecasting use a symbolic computational approach to automating intelligence. This approach takes advantage of the expert knowledge of the operator which is, however, neither easy to elicit nor articulate. A major advantage of using ANN over expert systems is its nondependency on an expert. Furthermore ANN also performs nonlinear regression among load and weather patterns and can also be used to model the time series method or as a combination of both. Generally, time series approaches assume that the load can be decomposed into two components. One is weather dependent and the other is weather independent. Each component is modeled separately and the sum of these two gives the total load forecast. The behavior of weather independent load is mostly represented by Fourier series or trend profiles in terms of time functions. The weather sensitive portion of the load is arbitrarily extracted and modeled by a predetermined functional relationship with weather variables. An adaptive neural network approach has been recently proposed by Peng et al which incorporates the familiar Box and Jenkins time series model. Instead of off-line simulation, Adalines are used to update the model parameters and simulation results indicate that one-week ahead hourly forecasts can be generated with mean absolute percentage error of less than 3.4 per cent. This forecasting model does not consider the weather dependency of the load, particularly the temperature and uses only past load. The objective of the present approach is to study the Functional-Link Net architecture to identify a time-series load model incorporating the nonlinearity due to temperature variations. The functional-link-network has an input vector comprising the fourier series functions and nonlinear components comprising the temperature functions and their enhancements. The model parameters are identified during training and once the convergence is achieved, the forecasting model is ready for prediction. This new approach is totally adaptive and generalized and does not depend upon the season and day type. The forecasting accuracy of functional-link net compared to the adaline for one week ahead forecasting is less than 2.5 percent instead of 3.4 percent for the later. The approach is highly flexible and Sundays and holidays can be easily included. The approach presented in this paper is amenable for real-time implementation, as hourly or daily adaptation of model parameters can be done.
Source Title: IEEE Power Engineering Review
ISSN: 02721724
Appears in Collections:Staff Publications

Show full item record
Files in This Item:
There are no files associated with this item.

Page view(s)

checked on Jan 26, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.