Impact of Presence of Calender Feature in the Performance of Day-Ahead Electric Load Power Forecasting with New Framework Realized in Artificial Neural Network Tehinique using Matlab Programming

Abstract

In Smart Grid era, Electric Load Power Load Forecasting (ELPF) is an important process in any power system network to make it a customer friendly distribution network. Majority of literatures claims that an Artificial Neural Network (ANN) technique produces a better accuracy compared to any other technique. Hence ANN based model is chosen with a new architecture for this research work. This is a multiple input multiple output (MIMO) Feed Forward neural network with back propagation algorithm. The input features like climate, season, week end, week day, national and festival holiday factors are taken into due consideration in the development of the architecture. This research work is mainly focused on the impact of calendar variables in day-ahead hourly ELPF for the year 2017 by using data of the previous year 2016. The evaluation criteria are presented with determination of various statistical errors. The performance of each model is analyzed and the results are presented in this paper. These models developed based on ANN technique are implemented using Matlab programming. The proposed models are proven for their simplicity, less measurement requirement and easiness for implementation.