Air conditioner cluster aggregation external characteristic modeling method based on online deep learning

The invention relates to an air conditioner cluster aggregation external characteristic modeling method based on online deep learning and an air conditioner cluster aggregation external characteristic model architecture based on a long short-term memory neural network. The method comprises the following steps: pre-training the built external characteristic model architecture by utilizing a limited historical sample set; and on-line training is carried out on the pre-trained external characteristic model based on the real-time measurement sample and the on-line deep learning error. The air conditioner cluster aggregation external characteristic modeling method has the advantages that the strong nonlinear coupling relation between the air conditioner cluster aggregation reserve capacity and the historical outdoor temperature, the compensation price, the grid-connected point voltage and the aggregation response power is described; meanwhile, the provided aggregated external characteristic modeling method does no