Data-Driven Modeling of Photovoltaic Power Generation System Based on MLP and LSTM Neural Network

The photovoltaic (PV) array and inverter play a pivotal role in PV power generation systems. However, these systems often face challenges such as high computational complexity and difficulties in determining internal parameters. To address these challenges, this study proposes a data-driven (DDR) modeling approach for PV power generation systems. This approach leverages neural networks, shifting the focus from the internal parameters of key components to their external characteristics under various operating conditions. Specifically, this study implements a Multilayer Perceptron (MLP) neural network for PV array modeling and a Long Short-Term Memory (LSTM) neural network for inverter modeling. By employing two distinct networks, the DDR model aligns closely with the physical response characteristics of the system. Finally, the neural networks are integrated into the commercial simulation software PSCAD for verification. The simulation results confirm that the proposed DDR model exhibits high accuracy and robust generalization capabilities.