Multi-Term Electrical Load Forecasting of Smart Cities Using a New Hybrid Highly Accurate Neural Network-Based Predictive Model

Abstract

This paper presents FARHAN, a novel hybrid model designed to address the challenges of electrical load forecasting in smart grids. FARHAN combines descending neuron attention, long/short-term memory (LSTM), and Markov-simulated neural networks to optimize accuracy and analysis time for short-, mid-, and long-term smart grid planning decisions. FARHAN processes electricity load data efficiently by utilizing two LSTM blocks (LSTM.B1 & LSTM.B2) with attention layers, a 90% gain averager, and a Markov chain analyzer. The comparative analysis demonstrates FARHAN's superiority over traditional LSTM models and other methodologies, exhibiting remarkable Mean Absolute Percentage Errors (MAPEs) of 0.019162%, 0.0386%, and 0.039% for 14 years, annual, and monthly estimations, respectively. Root Mean Square Percentage Errors (RMSPEs) of 2.5%, 5.2%, and 1.2% and an overall R2 of 1 validate its exceptional accuracy. FARHAN's innovative approach establishes it as a robust and intelligent tool for enhancing electrical load forecasting in smart grids and energy systems, promising significant advancements in the field.