Operation cost prediction of the WFGD system based on physics-informed neural network

With the increasing stringency of environmental protection requirements, emission standards for desulfurization systems in power plants have become more rigorous. However, reducing SO2 emissions will inevitably increase operational costs. The historical operating data of the wet flue gas desulphurization (WFGD) system cannot cover all operating conditions, resulting in purely data-driven models that do not fully reflect the actual cost-generating mechanisms, making it challenging to predict operating costs accurately. This paper proposes an operational cost prediction model based on the physics-informed neural network (PINN) algorithm to achieve economically efficient operation of desulfurization systems. A PINN-based prediction model is established by identifying the physical monotonic relationships between the operational costs of the desulfurization system and three key operational parameters. The effectiveness of the PINN model is verified by the operation data of the desulfurization system, compared with three state-of-the-art ML models. The results indicate that the PINN model provides the most accurate predictions and has superior interpretability.