Simulation Code and Dataset for: Computational Framework for Multi-Objective Optimization of Activated Biochar Properties Using Machine Learning and Evolutionary Algorithms

This repository contains the Python simulation script and 13 output CSV data files supporting the manuscript titled "Computational Framework for Multi-Objective Optimization of Activated Biochar Properties Using Machine Learning and Evolutionary Algorithms," submitted to Scientific Reports (Springer Nature). The simulation framework generates a synthetic dataset of 800 parameter combinations representing activated biochar production across seven process input variables (pyrolysis temperature 400–900 °C, residence time 0.5–3.0 h, heating rate 5–50 °C min⁻¹, activation agent-to-char mass ratio 0.5–3.0, feedstock type, activation agent, and EU deployment region). Response surface functions are calibrated to literature-reported property bounds derived from 45 peer-reviewed experimental studies published between 2015 and 2025. A multi-output Random Forest surrogate model (200 estimators, maximum depth 25) maps the seven-dimensional input vector to six biochar properties simultaneously: specific surface area (m² g⁻¹), CO₂ adsorption capacity (mmol g⁻¹), specific capacitance (F g⁻¹), energy storage capacity (Wh kg⁻¹), carbon stability index (dimensionless), and process energy demand (MJ kg⁻¹). Differential evolution optimization (population size 50, 100 generations) identifies Pareto-optimal process configurations. The framework additionally performs a spatially explicit EU-scale carbon mitigation assessment across five European agricultural regions. All stochastic operations use random seed 42 for deterministic reproducibility. The complete pipeline is implemented in Python 3.14 using NumPy 2.1, pandas 2.2, scikit-learn 1.5, SciPy 1.14, and Matplotlib 3.9. Included files:1. biochar_optimization_framework.py — main simulation, training, and optimization script2. biochar_simulation_dataset.csv — full 800-sample synthetic dataset3. model_performance.csv — R² test scores per target property4. feature_importance.csv — permutation-based feature importance rankings5. optimal_parameters.csv — optimal process parameter configuration6. optimal_properties.csv — predicted biochar properties at optimal conditions7. pareto_properties.csv — Pareto-front non-dominated solution properties8. eu_mitigation.csv — EU regional net carbon mitigation estimates (MtCO₂e yr⁻¹)9. activation_summary.csv — mean SSA and CO₂ adsorption grouped by activation agent10. feedstock_summary.csv — mean carbon stability index grouped by feedstock type11. validation_predictions_test_set.csv — observed versus predicted values on 160-sample test set12. carbon_stability_correlations.csv — H/C ratio, O/C ratio, carbon stability index, temperature, and activation agent13. temperature_sensitivity_analysis.csv — binned temperature analysis (400–550, 550–700, 700–850, >850 °C) with mean and SD statistics