SCAPS Database Generator: An Automated Machine Learning Dataset Construction Tool for Solar Cell Simulation

The scaps_db_generator introduced in this article is a Python toolkit designed to address the inefficiency of manual parameter scanning in SCAPS-1D solar cell simulations. It can automatically perform batch simulations and generate structured datasets suitable for machine learning training, sensitivity analysis, and device physics research, helping researchers improve their work efficiency.

SCAPS-1D is a widely used simulation software in photovoltaic device research, but manually running hundreds or even thousands of simulations for tasks like parameter scanning and ML dataset construction is impractical. scaps_db_generator was created to address this pain point and automate the simulation process.

Parameterized Space Scanning

Supports configurable range scanning of three key parameters: interface defect density (cm⁻²), bulk defect density (cm⁻³), and absorber layer thickness (µm).

Multi-core Parallel Computing

Automatically detects the number of CPU cores, splits the parameter space into sub-intervals for parallel processing, and aggregates results into a single CSV. A theoretical speedup of nearly 8x is expected on an 8-core processor.

Automated Workflow

Covers an end-to-end automated workflow including script generation, simulation execution, result parsing, data aggregation, and temporary file cleanup—no manual intervention required.

Modular Architecture

The project adopts a clear modular structure, including directories like baseline (templates), csv (output), scripts (temporary scripts), and core files such as config.py (configuration) and scaps_simulation.py (simulation logic).

Usage Steps

1. Configure the .env file to set environment variables like the SCAPS path;
2. Adjust parameter scanning ranges in config.py;
3. Run modes: single-round simulation (db_generator.py), batch parallel (db_batch_generator.py), visualization (plot_iv_curves.py, etc.).

Output Data

The generated CSV file has each row representing a simulation, including current values corresponding to voltage steps and performance metrics (Voc, Jsc, FF, PCE), compatible with mainstream data analysis and ML frameworks.

Application Scenarios

• ML research: Performance prediction, inverse design, surrogate models;
• Sensitivity analysis: Parameter importance ranking, interaction effect analysis;
• Device physics: Defect engineering, thickness optimization, interface engineering.

Technical Highlights

High degree of automation, intelligent parallel efficiency, modular design, configuration-driven (no code modification needed), standardized output, built-in visualization functions.

System Requirements

• OS: Windows (SCAPS-1D is a Windows application);
• Python: 3.8+;
• Dependencies: numpy, pandas, matplotlib, scikit-learn, scipy, python-dotenv;
• Pre-install SCAPS-1D and configure its path.

scaps_db_generator transforms tedious manual simulations into an efficient batch processing workflow, significantly improving the efficiency of photovoltaic device research. It is recommended for researchers engaged in photovoltaic simulation, ML-assisted material design, or computational physics to try this tool to accelerate their research progress.