Hierarchical optimal scheduling of electricity-hydrogen integrated energy systems via collective neurodynamic optimization

In this paper, a bilevel optimal scheduling model for electricity-hydrogen integrated energy systems (IESs) is developed, which simultaneously incorporates the trading of both electricity and hydrogen energy. Firstly, the upper level model is tantamount to a mixed-integer programming (MIP) optimization problem and the lower level model is modeled as a distributed convex optimization problem. Then, two neurodynamic algorithms are proposed to manage the hierarchical scheduling of electricity-hydrogen IESs collaboratively. For the upper level MIP optimization problem, a two-timescale duplex neurodynamic (TTDN) algorithm is adopted, which employs two recurrent neural networks operating collaboratively on different timescales. Compared to the particle swarm optimization (PSO) algorithm, multi-objective evolutionary algorithm based on decomposition (MOEA/D) and the non-dominated sorting genetic algorithm III (NSGA-III), the TTDN algorithm achieves better local optimal solutions. For the lower level distributed convex optimization problem, based on the sliding mode control technique, a distributed neurodynamic algorithm with finite-time convergence properties is designed, which is rigorously proved based on the Lyapunov stability theory. Moreover, the computational speed of the collective neurodynamic approach is further accelerated by using the neural network to train a deep-learning-based surrogate model (DLBSM). Finally, experimental analyses demonstrate that the collective neurodynamic approach enables a more efficient development of better electricity-hydrogen trading strategies.