文献类型：期刊文献

英文题名：Deep Reinforcement Learning-Based Energy-Conscious Scheduling Under Time-of-Use Electricity Price

作者：Wang, Caixia[1,2,3];Fang, Wenxuan[4];Dai, Xin[1,2,3];He, Renchu[5];Du, Wei[1,2,3];Tang, Yang[1,2,3]

机构：[1]East China Univ Sci & Technol, State Key Lab Ind Control Technol, Minist Educ, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[3]Huzhou Inst Ind Control Technol, Huzhou 313099, Peoples R China;[4]Westlake Univ, Trustworthy & Gen Artificial Intelligence Lab, Hangzhou 310030, Zhejiang, Peoples R China;[5]China Univ Petr, Coll Artificial Intelligence, Sch Dept Automat, Beijing 102249, Peoples R China

年份：2025

卷号：22

起止页码：11490

外文期刊名：IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING

收录：;EI(收录号：20250617838460);WOS:【SCI-EXPANDED(收录号:WOS:001464989100008)】；

基金：This work was supported in part by the National Natural Science Foundation of China under Grant 62293500, Grant U2441245, Grant 62293502, and Grant 62173144; in part by Shanghai Rising-Star Program under Grant 22QA1402400; in part by the Program of Introducing Talents of Discipline to Universities (the 111 Project) under Grant B17017; in part by the Fundamental Research Funds for the Central Universities under Grant 222202517006; and in part by the State Key Laboratory of Industrial Control Technology, China under Grant ICT2024A22.

语种：英文

外文关键词：Electricity; Costs; Production; Job shop scheduling; Automation; Energy consumption; Training; Conferences; Power demand; Decision making; Deep reinforcement learning; energy-conscious scheduling problem; time of use electricity price; optimization algorithm

摘要：In the context of carbon peaking and carbon neutrality, green production scheduling that considers energy has attracted increasing attention. Reinforcement learning (RL) has emerged as a topic for developing efficient algorithms for solving complex combinatorial optimization problems, such as real-world shop scheduling problems. In this paper, we investigate the energy-conscious scheduling problem (ECSP) under time-of-use (TOU) electricity price with the goal of minimizing both the waiting time and extra electricity cost. A mathematical model is formulated, and an efficient deep RL (DRL)-based optimization method is proposed to solve the problem effectively. We design a novel ECSP network (ECSPNet) tailored to handle various ECSP scales based on the characteristics of the problem. Moreover, the Tchebycheff decomposition method is used to solve the multi-objective optimization problems, complemented by the application of the policy gradient method from reinforcement learning to train the ECSPNet without size limitations. Experiments verify that the proposed ECSPNet outperforms state-of-the-art methods and is computationally efficient, even on instances of larger scales unseen in training. Real-world case studies reveal that the proposed method can reduce annual total electricity cost by approximately 30% while effectively maximizing production efficiency.