基于可解释机器学习的固体填充床储热装置 快速设计研究

In order to address the issues of time-consuming performance calculations, lack of design methods, and difficulty in meeting various heat storage requirements for packed-bed heat storage devices, a design method for solid packed-bed heat storage devices is proposed which is versatile and can rapidly provide accurate design results. Firstly, a dataset of heat storage efficiency for solid packed-bed heat storage devices is established, and an artificial neural network model is trained to accurately predict the device's heat storage efficiency. Secondly, the SHapley Additive exPlanation method is used to explain the predictions of the artificial neural network model, quantifying the effects of material properties, device structure dimensions and operating parameters on the device's heat storage efficiency. Finally, a correlation formula for the heat storage efficiency of solid packed-bed heat storage devices is established. Based on the heat storage efficiency correlation formula, a design process for packed-bed heat storage devices suitable for various heat storage scenarios is proposed and analyzed using the heat storage device in the Andasol 1 solar thermal power station in Spain as a case study. The results demonstrate that the relative deviation of the heat storage efficiency correlation calculation results from the numerical simulation results is within 10%, suitable for rapid calculation and accurate prediction of heat storage efficiency for packed-bed heat storage devices. Compared to traditional numerical simulation methods, the computational efficiency is increased by five orders of magnitude. In addition, the design solution is economically feasible, emphasizing the convenience and practicality of the thermal storage efficiency correlation formula in engineering practice.