Short Term Load Forecasting In Power System Using Neural Networks

Abstract

Electricity is one of the most important factors in the modern times. As the consumption of electricity increases, generation also shall rise. The development of nation is obvious with consumption of electricity. The functional categories of power system are generation, transmission, distribution and utilization. The performance of the power system relies on efficient usage of its operation and maintenance units. The major expenditures on operation and maintenance will be on generation and distribution sectors. Certainly, economic operations in distribution sector will be on high demand and requires wide planning, design, construction and operation. The important function in power system called planning is aiming to reduce the cost of sub-transmission systems, substations, feeders, laterals and other components as well as the cost associated with losses. newlineLoad forecasting is of utmost significance to utilities, engineers, the power sector and the country as a whole. Load forecasting covers a lot of intricate technical concerns and economic analysis. newlineAmong various categories of Load forecasting, Short-Term Load Forecasting (STLF) plays a crucial role in various grid operations that involve dispatch and reliability analysis. Further, it helps to avoid over estimation and under estimation of the energy demand and thus contribute substantially in the reliability of grid. Weather factors like temperature and humidity are vital in improving the efficiency of STLF. The important performance metrics of forecasting model efficiency is scaled using root mean squared error (RMSE) and mean absolute percentage error (MAPE). newlineAlthough several conventional methods and Artificial Intelligence (AI) strategies have been put forth in the literature, Neural Network (NN) based solutions are preferred because of their capacity to manage substantial numbers of non-linear electrical loads. newlineAfter realising the importance of STLF, the following models are discussed and presented in this research work. The neural networks models used are