Data-Driven Approach to Attemperator Steam Temperature Prediction in Biomass Power Plant

Thermal power plants utilize high temperature and high pressure steam to generate electricity. The steam temperature and pressure influence power generation rate, facilities, and significant subcomponents of power plants. Therefore, controlling steam temperature is critical. In this paper, we conducted modeling of temperature prediction model for steam temperature of attemperator, and compared three prediction models. The target plant in this study is a biomass power plant that utilizes fluidized-bed boiler. The target system is an attemperator which assists in controlling the temperature of superheated steam. The target variables in the models consist of the output temperature of the attemperator and the difference of temperature between inlet and outlet temperature. The least squares method, locally weighted regression, and a neural network (NN) are employed for learning algorithm of the prediction model. The k-fold cross-validation was used to optimize the prediction models. The experimental results obtained by all of three methods show low root-mean-squared error (RMSE) values. The NN model achieves the lowest RMSE and the largest correlation coefficient value among the three prediction methods.