An Intelligent Prediction Method for Safety Margins of Flexible Thermal Power Units Based on PipeLine Creep Life Damage

With the increasing proportion of clean energy in power systems, coal-fired power units are gradually transitioning from steady-state operation to flexible operation. Significant load fluctuations cause deviations in the lifetime of thick-walled pressure components (represented by boiler main steam pipelines) from design values, posing non-negligible impacts on operational safety. To address this, this paper proposes an intelligent safety margin prediction method based on pipeline creep life damage. Firstly, the modified Wilshire model is employed to calculate cumulative creep damage of pipelines under different flexible operation conditions, which is integrated as an input for the prediction model. Subsequently, SHAP method are applied to evaluate all features, eliminating those with minor impacts on load and main steam pressure margins. Finally, the filtered dataset is used to train ResNet-LSTM network, establishing the safety margin prediction model. Test results demonstrate mean percentage errors of 6.11% for load and 4.20% for main steam pressure with ResNet-LSTM, validating the method's predictive accuracy.