dPV: An End-to-End Differentiable Solar-Cell Simulator

TL;DR

dPV is a novel, fully differentiable simulator for photovoltaic cells that enables efficient optimization and parameter discovery by computing PCE and its derivatives using automatic differentiation in Python.

Contribution

The paper presents dPV, the first end-to-end differentiable PV simulator based on drift-diffusion and Beer-Lambert laws, facilitating integrated optimization and machine learning applications.

Findings

dPV accurately computes PCE derivatives with respect to design parameters.

The simulator enables efficient optimization compared to traditional methods.

dPV can be integrated with neural networks for data-driven PV design discovery.

Abstract

We introduce dPV, an end-to-end differentiable photovoltaic (PV) cell simulator based on the drift-diffusion model and Beer-Lambert law for optical absorption. dPV is programmed in Python using JAX, an automatic differentiation (AD) library for scientific computing. Using AD coupled with the implicit function theorem, dPV computes the power conversion efficiency (PCE) of an input PV design as well as the derivative of the PCE with respect to any input parameters, all within comparable time of solving the forward problem. We show an example of perovskite solar-cell optimization and multi-parameter discovery, and compare results with random search and finite differences. The simulator can be integrated with optimization algorithms and neural networks, opening up possibilities for data-efficient optimization and parameter discovery.