Key Laboratory of Ministry of Education on Smart Power Grids (Tianjin University), Tianjin 300072, China

The medium- and long-term scheduling of traditional microgrids is difficult to take into account the energy cycle process of energy storage during the daily energy balance, which may not only lead to the inability of the daily energy balancing scheme to support the charging and discharging strategy of energy storage during the initial period of the day, but also cannot adapt to the high loss of the electricity-hydrogen conversion process. Therefore, a two-stage stochastic optimal scheduling method combining adaptive period division and variable-resolution is proposed. Firstly, for the “near small and far big” problem of uncertainty, a model of source-load output characteristics based on the modified martingale model is established. Secondly, a two-stage variable-resolution stochastic optimization architecture for the medium- and long-term scheduling of micro-energy networks with hydrogen is constructed. At stage one, an adaptive time division method based on deep neural network is proposed. At stage two, with the goal of minimizing the system operation costs and combining with time-segment chance constraints, the stochastic optimal scheduling models with coarse and fine resolutions are established, respectively. The latter arranges the hourly equipment output plans based on the state of charge of hydrogen storage equipment decided by the former, and a solution scheme based on sampling method is proposed. Finally, the effectiveness of the proposed model and method is verified through numerical simulations.