∂PV: An end-to-end differentiable solar-cell simulator

We introduce ∂PV, an end-to-end differentiable photovoltaic (PV) cell simulator based on the drift-diffusion model and Beer–Lambert law for optical absorption. ∂PV is programmed in Python using JAX, an automatic differentiation (AD) library for scientific computing. Using AD coupled with the implicit function theorem, ∂PV 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.

我们提出了∂PV，一款基于漂移扩散模型与光吸收比尔-朗伯定律（Beer–Lambert law）的端到端可微光伏（PV）电池模拟器。∂PV依托用于科学计算的自动微分（automatic differentiation, AD）库JAX，以Python语言实现。结合自动微分与隐函数定理，∂PV可在与求解正向问题相当的时间内，计算输入光伏设计的功率转换效率（power conversion efficiency, PCE），以及该效率相对于任意输入参数的导数。我们展示了钙钛矿太阳能电池优化与多参数发现的示例，并将结果与随机搜索、有限差分法进行对比。该模拟器可与优化算法及神经网络集成，为数据高效优化与参数发现提供了全新可能。