Accurate evaluation on peak shaving capacity of combined-heat-and-power thermal power units based on physical information neural network

The high proportion of new energy requires the power system to have sufficient flexibility and peak shaving capacity. The combined-heat-and-power thermal power unit is one of the main flexibility resources. Accurately evaluating the peak shaving capability from thermal power units is of great significance for power system scheduling. However, there are some problems in the previous evaluation models. Fully mechanistic models may exhibit poor accuracy due to component degradation and other reasons, while fully data-driven models may exhibit poor generalization performance due to limited training conditions. To address these issues, this paper first uses physical information neural network to fuse thermal power mechanisms and operational data, and constructs a high-precision and strong-generalization power output evaluation model. Then, error indicators and generalization indicators are used to validate the effectiveness and superiority of the model. In the test set, the mean absolute error, root mean square error and Pearson correlation coefficient of the model are 1.633 MW, 1.971 MW and 0.9643, respectively. Based on this model, the peak shaving capacity range of the target thermal power plant under different heating demands is calculated. Finally, the experiment analyzed the real-time peak shaving capability and peak shaving margin.