Tang Yuanhong, Cao Di, Xiao Jian, Jiang Chenying, Huang Qi, Li Yunwei, Chen Zhe, Blaabjerg Frede, Hu Weihao
Power System Wide-area Measurement and Control Sichuan Provincial Key Laboratory, School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China.
MORNSUN Guangzhou Science & Technology Co., Ltd., Guangzhou 510663, China.
Fundam Res. 2023 May 25;5(3):1111-1116. doi: 10.1016/j.fmre.2023.05.004. eCollection 2025 May.
Energy losses during the conversion and supply of electric power are considered a significant issue and cannot be estimated. Improvement in the efficiency of energy conversion systems is highly restricted because of their internal nonlinearity and complexity. Thus, inspired by the successful utilization of robotic chemists, we demonstrate a pioneering concept of artificial intelligence (AI)-aided automatic online real-time optimization of a power electronics converter using a dual active bridge (DAB) converter as an example. An optimal modulation strategy was obtained through repeated automatic exploration experiments on a practical DAB converter platform. Specifically, the DAB experimental platform operated autonomously around the clock for approximately 71 h. It performed 120,000 consecutive experiments (12,000 episodes) within a six-variable experimental space driven by a deep deterministic policy gradient (DDPG) algorithm. The proposed AI-aided automatic online real-time optimization method achieved significantly improved efficiency of power conversion and supply. Consequently, zero carbon emissions may be obtained in the future.
电力转换和供应过程中的能量损失被视为一个重大问题且难以估计。由于能量转换系统内部的非线性和复杂性，其效率的提高受到很大限制。因此，受机器人化学家成功应用的启发，我们以双有源桥（DAB）变换器为例，展示了一种人工智能（AI）辅助的电力电子变换器自动在线实时优化的开创性概念。通过在实际的DAB变换器平台上进行反复的自动探索实验，获得了一种最优调制策略。具体而言，DAB实验平台连续运行约71小时。在由深度确定性策略梯度（DDPG）算法驱动的六变量实验空间内，它进行了120000次连续实验（12000个情节）。所提出的AI辅助自动在线实时优化方法显著提高了电力转换和供应的效率。因此，未来可能实现零碳排放。
