The GAC-BiLSTM-AM Method and Its Application to the Prediction of Cracking Outlet Temperature in Ethylene Cracking Furnaces

Aiming at the problem that the traditional prediction model is limited by the information processing capacity and leads to insufficient generalization ability, a GAC-BiLSTM-AM prediction method combined with the mistake correction mechanism is proposed. The method constructs a base learner by organically combining dual-channel convolutional neural network, bidirectional long short-term memory, attention mechanism and fully connected network to deeply mine the intrinsic laws and key information of data. Meanwhile, during the iteration process of the base learner, the proposed mistake correction mechanism is used to dynamically adjust the sample weights and assist the optimizer in refining the model parameters. Furthermore, the linear regression model and the Gold Rush Optimizer are further utilized to achieve result mapping and structural parameter optimization. Experimental results demonstrate that the GAC-BiLSTM-AM method exhibits excellent generalization performance across multiple benchmark prediction tasks.