Short-Term Power System Load Modelling Using Artificial Neural Systems

Power systems are very dynamic in nature, and it is important for power systems to be simulated and modeled accurately for optimal performance of the system. Historically, load modeling coupled with forecasting has always been important for power system operation.  In current times with the onset of the deregulation of the energy industries, load modeling and forecasting have become even more important. Accurate Power system load modeling plays a very important role in the operation of electricity companies, such as monitoring load management policies, helping with the generator commitment problem by providing short term forecasts, and aiding in power system planning. Power system load models enable the power companies to load forecast so they can develop maintenance schedules and plan for upgrades. Power system expansion planning starts with a forecast of anticipated future load requirements. Estimates of both demand and energy requirements are crucial to effective system planning. Due to the potential for maximizing production and minimizing losses. There are many methods of performing load modeling. Traditional methods include trend extrapolation, regression method, time series analysis, etcetera and these methods are good at modeling linear data but don't always work so well with non-linear data. Optionally there are heuristic decisions made in the modeling process that can lead to inaccurate models which may lead to forecast that are either too low or too high. A forecast that is too low can easily result in lost revenue from sales to neighboring utilities or even in load curtailment. On the other hand, forecasts that are too high can result in financial problems due to excessive investment in an electric plant that is not fully utilized.

This presentation explores the use of several popular artificial neural systems for modeling the power system load. These systems allow for the estimation of possibly non-linear models without specifying a precise functional form. The use of neural networks for load forecasting is suitable since the power system load may be viewed as a non-linear function that varies with load composition, hour of the day,  day of the week, season, random weather and weekly changes in time.

 

Speaker bio: Gary L. Lebby received the BS degree in mathematics (minor in Computer Science) and the BS degree in physics in 1980, and the MS degree in physics in 1982, from the University of South Carolina, and the Ph.D. in electrical engineering from Clemson University in 1985. In 1985 he accepted the position of Assistant Professor of Electrical Engineering at North Carolina Agricultural and Technical State University. In 1992, he was promoted to Associate Professor, and was appointed Chairperson of Electrical Engineering in 1994.  In 1996, he was promoted to Professor.  In 1998, he graduated the first doctoral student from North Carolina Agricultural and Technical State University (newly formed) doctoral program. In 2009 Dr. Lebby was appointed by the Dean of the College of Engineering to plan for a graduate M.S Program in Energy Systems Engineering to meet the national demand for Energy Professionals. Dr. Lebby currently directs the Laboratory for Biologically Inspired Energy and Engineering Systems, and his areas of research include: Power Systems Modeling, Artificial Neural Systems, and Parallel Distributed Processing.

Date: 
Thursday, March 12, 2015 - 2:30pm
Location: 
E-Lab 2 Auditorium, Rm. 119
Year: 
2015
Semester: