Prediction of hydrogen production in proton exchange membrane water electrolysis via neural networks

• PEMWE is promising for hydrogen production. • Optimizing hydrogen production rate in PEMWE systems is investigated. • ANN is proposed to accurately predicts hydrogen production. • ANN significantly outperforms traditional models with an R 2 value of 0.9903. Advancements in water electrolysis technologies are crucial for green hydrogen production. Proton exchange membrane water electrolysis (PEMWE) is characterized by its efficiency and environmental benefits. The prediction and optimization of hydrogen production rates (HPRs) in PEMWE systems is difficult and still challenging because of the complexity of the system as well as the operational parameters. The integration of artificial intelligence (AI) and machine learning (ML) appears to be effective in optimization within the energy sector. Hence, this work employs the artificial neural network (ANN) to develop a model that accurately predicts HPR in PEMWE setups. A novel approach is introduced by employing the Levenberg–Marquardt backpropagation (LMBP) algorithm for training the ANN. This model is designed to predict HPR based on critical operational parameters, including anode and cathode areas (mm 2 ), cell voltage (V) and current (A), water flow rate (mL/min), power (W), and temperature (K). The optimized ANN configuration features an architecture with 7 input nodes, two hidden layers of 64 neurons each, and a single output node. The performance of the ANN model was evaluated against conventional regression models using key metrics: mean squared error (MSE), coefficient of determination (R 2 ), and mean absolute error (MAE). The findings of this study reveal that the developed ANN model significantly outperforms traditional models, achieving an R 2 value of 0.9989 and an MAE of 0.012. In comparison, random forest (R 2 = 0.9795), linear regression (R 2 = 0.9697), and support vector machines (R 2 = − 0.4812) show lower predictive accuracy, underscoring the ANN model's superior performance. This work demonstrates the efficiency of the LMBP in enhancing hydrogen production forecasts and sets a foundation for future improvements in PEMWE efficiency. By enabling precise control and optimization of operational parameters, this study contributes to the broader goal of advancing green hydrogen production as a viable and scalable alternative to fossil fuels, offering both immediate and long-term benefits to sustainable energy initiatives.