Dataset for Research on the Design of Catalysts for Electrochemical NO Reduction to Ammonia Based on Stacked Ensemble Learning

Electrocatalytic reduction of nitric oxide for ammonia synthesis (NORR), as a core technology in the field of green energy conversion, relies on the development of highly efficient NORR electrocatalysts to enhance ammonia yield and Faradaic efficiency. However, conventional experimental methods for screening high-activity NORR catalysts often entail high resource consumption and time costs. This study employs a combination of machine learning and SHAP feature analysis to construct a stacked ensemble model based on a single algorithm, systematically analyzing key features influencing NORR performance using an experimental database. Evaluation of eight model algorithms revealed that the Stacked-SVR model achieved an R2 of 0.9223 and an RMSE of 0.0608 for predicting on the test set, while the Stacked-RF model achieved an R2 of 0.9042 and an RMSE of 0.0900 for predicting on the test set. The stacked ensemble model integrates the strengths of individual algorithms and demonstrates strong NORR prediction performance while avoiding overfitting. SHAP feature analysis results indicate that the Cu content in the catalyst composition has the most significant impact on catalytic performance; the combination of wet chemical reduction synthesis, carbon fiber (CP) conductive substrate, and HCl electrolyte is more conducive to enhancing catalytic activity. Additionally, moderately lowering the working potential, controlling electrolyte volume at low to medium levels, reducing catalyst loading, and increasing electrolyte concentration can synergistically promote simultaneous improvement in ammonia yield and Faradaic efficiency.Figure 1 Machine learning workflow diagramFigure 2 (a) Heatmap of Pearson correlation coefficient matrix for ammonia production; (b) Heatmap of Pearson correlation coefficient matrix for ammonia Faraday efficiencyFigure 3 Distribution of characteristic parameters after normalization: (a) Ammonia production database; (b) Ammonia Faraday efficiency database.Figure 4 Validation results of different machine learning models for ammonia production predicting: (a) RF; (b) XGBoost; (c) SVR; (d) Stacked-RF; (e) Stacked-XGBoost; (f) Stacked-SVR; (g) CNN; (h)ANNFigure 5 (a) R2and (b) RMSE of different machine learning models for Ammonia production predicting on training and test setsFigure 6 Validation results of different machine learning models for Ammonia Faraday efficiency predicting: (a) RF; (b) XGBoost; (c) SVR; (d) Stacked-RF; (e) Stacked-XGBoost; (f) Stacked-SVR; (g) CNN; (h) ANNFigure 7 (a) R2 and (b) RMSE of different machine learning models for Ammonia Faraday efficiency predicting on training and test setsFigure 8 SHAP-based feature importance analysis plots: (a) Stacked-SVR prediction model for ammonia production ; (b) Stacked-RF prediction model for  Ammonia Faraday efficiency Figure 9 Local interpretation plot of the ammonia production prediction model based on SHAP values: (a) Metal type; (b) Preparation method; (c) Conductive substrate; (d) Electrolyte typeFigure 10 Local interpretation plot of the Ammonia Faraday efficiency prediction model based on SHAP values: (a) Metal type; (b) Preparation method; (c) Conductive substrate; (d) Electrolyte type