Short-Term Load Forecasting using Hybrid Quantized Elman Neural Model

A novel Elman neural model with hybrid quantized architecture is proposed in this paper for short-term power load forecasting. For the new networks structure, the quantum map layer is employed to address the pattern mismatch between the context layer and the quantized input layer, the laws of quantum physics are employed in the states of qubit neurons and their interactions with the other classic neurons. For the new learning algorithm, the quantized extended-gradient method is used to obtain the extra information of load sequences. The numerical experiments are carried out to verify the theoretical results and clearly show that the hybrid quantized Elman neural model has good forecasting ability in term of accuracy.