An Optimization Study of a Mechanism-Assisted Hybrid Neural Network Prediction Model

The primary airflow of a boiler is a crucial parameter that influences the efficiency of combustion and the potential for energy savings. Conventional prediction techniques are ill-equipped to handle intricate nonlinear and time-varying relationships, making precise forecasting a formidable challenge. This paper proposes a hybrid model combining CNN, LSTM and Attention mechanism, with the objective of improving the accuracy of predictions regarding primary airflow. The model employs a three-stage process. Initially, CNN is used to extract local features. Subsequently, LSTM is employed to model temporal dependencies. Finally, an attention mechanism is used to focus on key features. In order to verify the effectiveness of the proposed method, experiments were conducted using historical data from a 600 MW thermal power unit in a power station. A variety of comparative experiments were carried out, which demonstrated that the proposed model exhibited superior prediction performance compared to traditional LSTM and CNN-LSTM models in the test set, with an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 0.95 and an RMSE of 3.1. Moreover, the model demonstrates remarkable performance in fast-varying and oscillating regions, offering a valuable tool for optimizing boiler operations and advancing industrial intelligence.