60049 - Multi-Parameter Prediction for Steam Turbine Based on Real-Time Data Using Deep Learning Approaches 

Nowadays, the scale of renewable energy is continuously increasing because of the increasing energy demand and environmental protection requirements. In order to tackle the accompanying issue of intermittency, the flexibility of traditional power generation system must be improved. Accurate parameter forecasting is of great importance to the steam turbine control and predictive maintenance which can help the improvement of system. However, traditional physics models and statistical models can no longer meet the needs of precision and flexibility when thermal power plants frequently undertake more and more peak and frequency modulation tasks.

In this study, deep-learning models including recurrent neural network (RNN) and convolutional neural network (CNN) for multi-parameter prediction are proposed, and are applied to predict real-time parameters of steam turbine based on data from a power plant. Firstly, the prediction results of RNN and CNN models are compared by the overall performance. The two models show good performance on forecasting of six state parameters while RNN performs better. Moreover, the detailed performance on a certain day show that the relative error of two models are both less than 2%. Finally, the influence of model designs including loss function, training size and input time-steps on the performance of RNN model are also explored. The effects of the above parameters on the prediction performance, training and prediction time of the models are studied. The results can provide a reference for model deployment in the power plant. It is convinced that the proposed method has a high potential for dynamic process prediction in actual industrial scenarios through the above research.