Solving Nonlinear Energy Supply and Demand System Using Physics-Informed Neural Networks

Computation, Год журнала: 2025, Номер 13(1), С. 13 - 13

Опубликована: Янв. 8, 2025

Nonlinear differential equations and systems play a crucial role in modeling where time-dependent factors exhibit nonlinear characteristics. Due to their nature, solving such often presents significant difficulties challenges. In this study, we propose method utilizing Physics-Informed Neural Networks (PINNs) solve the energy supply–demand (ESD) system. We design neural network with four outputs, each output approximates function that corresponds one of unknown functions system describing four-dimensional ESD problem. The model is then trained, parameters are identified optimized achieve more accurate solution. solutions obtained from for problem equivalent when compare evaluate them against Runge–Kutta numerical order 5(4) (RK45). However, networks considered modern promising approach, as it effectively exploits superior computational power advanced computer systems, especially complex problems. Another advantage model, after being can across continuous domain. other words, not only trained approximate solution but also represent dynamic relationships between system’s components. approach requires time due need training. Furthermore, evaluated based on experimental results, ensuring stability convergence speed poses challenge. key influencing include manner which architecture designed, selection hyperparameters appropriate optimization functions. This critical highly task, requiring experimentation fine-tuning, demand substantial expertise time.

Язык: Английский

Application of PCA-GBRT/RF model for predicting heat flux on boiler water-cooled wall

Applied Thermal Engineering, Год журнала: 2025, Номер unknown, С. 126580 - 126580

Опубликована: Апрель 1, 2025

Язык: Английский