A Dual-Stage Attention based RNN - Model for Short Term Load Forecasting of Individual Household

Abstract

In the modern power distribution systems, Short-Term Load Forecasting (STLF) is helpful in effective load scheduling and control. Lately, smart meters are being introduced to industrial and domestic consumers to enable two-way energy flow and robust communication between the consumer and the utility. The smart meters provide sufficient data for the training of artificial intelligence(AI) models for the accurate STLF and economic load scheduling of individual consumers. The STLF of a particular household is useful for automatic demand response management of domestic consumers. Recently, attention-based Recurrent Neural Network (RNN) models have been proposed for STLF of the individual household. In this work, a Dual Stage Attention-Based Long Short Term Memory (DSA-LSTM) model is proposed for the STLF of the individual household. The performance of DSA-LSTM is compared with the single-layer attention based time series forecasting model. The comparative analysis of the performance of DSA-LSTM with several other AI-based STLF methods is also presented in tabular form.