AUTHOR=Fan Dongchuan , Wang Ruizhe , Qi Haonan , Deng Xiaoyun , Chen Yongdong , Liu Tingjian , Liu Youbo 

TITLE=Edge intelligence enabled optimal scheduling with distributed price-responsive load for regenerative electric boilers

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.976294

DOI=10.3389/fenrg.2022.976294

ISSN=2296-598X

ABSTRACT=Heat supply occupies a substantial amount of terminal energy usage. However, along with price raise of primary energy, an urgent need exists to reduce the average cost of energy consumption during the power purchasing of thermal services. Electric heating, an electricity-fed heating production and delivery technology, has been argued as a promising method to improve heating efficiency due to the ease of electricity scheduling. However, the traditional centralized operating methods on electricity purchasing rely on explicit physical modeling of every detail and accurate prediction of desired look-ahead information. That is rarely practical in real implementation. To enable model-free decision of power purchasing aiming at cost-saving in a real-time price environment, this paper proposes a deep reinforcement learning (DRL) based scheduling framework for electric heating in the field of electricity purchasing, heat storage, and supply management with the existence of responsive users. Firstly, the structure of distributed heating system fed by regenerative electric boilers (REB) which facilitates shiftable heat-load control is introduced. Besides, a terminal heat demand response model based on thermal sensation vote (TSV) is proposed, characterizing the consumption flexibility of responsive users. Secondly, due to the thermal system inertia, the sequential decision problem of electric heating load scheduling is transformed into a specific Markov decision process (MDP). Finally, the edge intelligence (EI) deployed on demand side employs twin delayed deterministic policy gradient (TD-3) algorithm to address the action space continuity of electric heating devices. The combination of DRL strategy and EI computing power enables optimal real-time scheduling. Unlike the traditional method, the trained intelligent agent makes adaptive control strategies according to the currently observed state-space avoiding the prediction uncertainty. The simulation results validate that the intelligent agent responds positively to changes in electricity price and weather conditions, reducing electricity consumption costs while maintaining user comfort. The adaptability and generalization of the proposed approach to different conditions are also demonstrated.