Main Steam Temperature Prediction Modeling Based on Autoencoder and GRU

There is an important parameter in the normal operation of thermal power plant units, that is, the main steam temperature, too high or too low will have a great impact on production. For safety and efficiency considerations, the main steam temperature needs to be controlled within a stable and appropriate range. In this way, the safety of equipment and people is guaranteed, but also ensures that resources will not be wasted. The prediction of the main steam temperature can achieve this aim. In view of the numerous influencing factors and non-linear characteristics of the main steam temperature, this paper proposes a main steam temperature prediction model based on autoencoder and GRU. First, build an autoencoder model, use historical data for training, enable it to accurately extract the essential information of the data. Then multiple influence parameters are dimensional reduced by the trained autoencoder, and use the reduced data as the input of the neural network built using GRU for training. The better model is finally compared with RNN and LSTM to verify the effectiveness and accuracy of the method.