Study on wall temperature prediction model of supercritical W-shape flame boiler based on algorithms integration
Addressing the water wall cracking in some thermal power plant under deep peak shaving conditions requires the monitoring of the temperature gradient between water wall tubes.This paper builds a predictive model based on PCA-SSA-LSTM algorithms integration to achieve effective prediction of outlet wall temperature near susceptible cracking tubes.First,to enhance the predictive performance of the model,17 input parameters are selected by wall temperature calculation theory,and PCA is employed for dimensionality reduction.Then,the SSA algorithm is adopted to optimize the hyperparameters of the PCA-LSTM prediction model,resulting in the optimal PCA-SSA-LSTM wall temperature prediction model.Comparison with other prediction models reveals a certain increase in the coefficient of determination (R2) for the model,along with varying degrees of reduction in root mean square error (RMSE) and mean absolute percentage error (MAPE).Application demonstrates the algorithms integration model effectively enhances the accuracy of predicting outlet wall temperatures for multiple tubes,serving as a supplement for the power plant' s DCS system to issue early alerts of critical wall temperatures,thereby reducing the risk of water wall cracking caused by large temperature differences between tubes.
algorithms integrationmodelingsupercritical w-flame boilerwater-cooled wall tubeswall temperature prediction
国家科技期刊平台
NSTL
万方数据
