پیش بینی الگوی بار میان مدت با شبکه عصبی
ABSTRACT
I. INTRODUCTION
II. SELECTION OF INPUT DATA
III. CASE STUDIES
IV. CONCLUSION
REFERENCES
ABSTRACT
The paper describes a mid-term daily peak load forecasting method using recurrent artificial neural network (RANN). Generally, the artificial neural network (ANN) algorithm is used to forecast shortterm load pattern and many ANN structures have been developed and commercialized so far. Otherwise, learning and estimation for long-term and mid-term load forecasting are hard tasks due to lack of training data and increase of accumulated errors in long period estimation. The paper proposes a mid-term load forecasting structure in order to overcome these problems by input data replacement for special days and a recurrent-type NN application. Also, the proposed RANN gives good performances on estimating sudden and nonlinear demand increase during heat waves. The results of case studies using load data of South Korea are presented to show performances and effectiveness of the proposed RANN.
INTRODUCTION
Accurate load forecasting becomes essential for an effective power system management and planning overhauls of the generators in a situation that the power consumption steeply increases and electric power reserve rate becomes insufficient. The load forecasting issues are to solve a complex nonlinear relationship related to previous load demand, social variation, and weather variation. Therefore, it still remains a challenging task to accurately forecast loads in order to supply high quality electric energy to customers in a secure and economic manner [1]. The purpose of load forecasting is generally divided into three categories: short-term, mid-term, and long-term load forecasting. Short-term load forecasting focuses on load variation from one hour to one week. Mid-term load forecasting interests in load estimation from one week to a month and long-term load forecasting can be extended to from one month to several years. The reason why the load forecasting is divided into several categories is that the estimation result from each method can be used in different operation objects. The short-term load forecasting is useful to control and schedule power generation for all generators in the system and also needed to estimate load flows and make decisions to prevent overloading on power facilities. The load forecasting for the mid-term and long-term is important to determine consequential power generation, the capacity of load consumption, and expansion of power facilities such as generators, transmission lines, and substations [2].