Stable and efficient hybrid controller of solar thermal membrane reactor based on machine learning and multi-objective optimization

Ensuring the efficient and stable operation of solar thermal devices under variable solar energy conditions is important and challenging. This study develops a hybrid feedforward feedback (FF-FB) control method for solar thermal-driven membrane reactor (STMR) to achieve multiple targets of stable conversion, low preheating input, and efficient hydrogen production and separation. The feedforward control part of the hybrid control method consists of multi-objective optimization based on machine learning model, and the feedback control part utilizes optimized PI control. The effects of control methods are tested under step and real continuous solar radiation. The results show that the FF-FB control combines the predictability of feedforward and the stability of feedback, giving FF-FB control the best overall performance. Also, under high frequency and high amplitude fluctuations of continuous solar radiation, the methane conversion in FF-FB control shows only an average deviation of 2.1 × 10−4 with a maximum of 1.16 × 10−3. The FF-FB control relatively improves the hydrogen yield by 33 % and hydrogen recovery by 43 % compared to the feedback control, while reducing the input preheat ratio by 30 %. The robustness of the hybrid FF-FB control effect is still maintained under multiple realistic error scenarios, and the average errors are all within 3 × 10−4.