Recurrent Neural Based Electricity Load Forecasting of G-20 Members

TL;DR

This paper presents a recurrent neural network approach, specifically LSTM, for forecasting electricity load among G-20 countries, achieving a mean absolute test error of 16.2193 TWh, to improve supply-demand balance.

Contribution

It introduces a novel application of RNN with sliding window for load forecasting of G-20 members, demonstrating improved accuracy over traditional methods.

Findings

Achieved MAE of 16.2193 TWh with LSTM.

Effective use of sliding window approach for data generation.

Enhanced load prediction accuracy for G-20 countries.

Abstract

Forecasting the actual amount of electricity with respect to the need/demand of the load is always been a challenging task for each power plants based generating stations. Due to uncertain demand of electricity at receiving end of station causes several challenges such as: reduction in performance parameters of generating and receiving end stations, minimization in revenue, increases the jeopardize for the utility to predict the future energy need for a company etc. With this issues, the precise forecasting of load at the receiving end station is very consequential parameter to establish the impeccable balance between supply and demand chain. In this paper, the load forecasting of G-20 members have been performed utilizing the Recurrent Neural Network coupled with sliding window approach for data generation. During the experimentation we have achieved Mean Absolute Test Error of 16.2193 TWh using LSTM.