Short-term Industrial Demand Response Capability Forecasting using Hybrid EMD-AGTO-LSTM Model

Accurate forecasting of potential demand response could provide benefits to grid operators and large customers. Though there are studies on generation forecasting and load forecasting, "demand response" forecasting has not been studied. This research work proposed a hybrid forecasting strategy using empirical mode decomposition (EMD), artificial gorilla troops optimizer (AGTO), and a long short-term memory network (LSTM) for day-ahead forecasting of demand response. Two and half years of half-hourly demand response data have been analyzed to validate the performance of the proposed strategy. The Python 3.10 with TensorFlow is used to perform the simulation. The efficiency of the proposed strategy is evaluated by contrasting it with hybrid EMD+PSO-FNN (EMD+particle swarm optimization-feedforward neural network) and hybrid CNN-LSTM (convolutional neural network-LSTM) algorithms. The simulation findings indicate that the proposed strategy gives 14% and 21% accurate results compared to EMD+PSO-FNN and CNN-LSTM, respectively, as measured by the average root means square error (RMSE).