Frequency Regulation and Emission Economy Optimization of nonlinear Hydrogen-Electric System Based on Deep Koopman-EMPC Method

This paper presents a hydrogen-electric system (HES) consisting of an electrolyzer and a fuel cell, and a Koopman based economic model predictive control (EMPC) method is proposed for this nonlinear system. Hydrogen energy can enhance the flexibility of power grid stability regulation and hydrogen energy storage system (HESS) has strong nonlinear dynamics because its model is based on electrochemical principles. In order to obtain a high precision approximate linear model, the Koopman operator is designed to be implemented by deep neural networks (DNNs). Then, an approximate high-dimensional linear model of HES was obtained through offline training. To take economic optimization into account during the frequency regulation process, the EMPC method using the Koopman approximate linear model is adopted. The EMPC method directly takes economic optimization costs (including operating costs and emissions) into account in the design of the objective function, and the weighted sum of the frequency control objective and the economic objective constitutes the EMPC objective function. The effectiveness and economic viability of the proposed method are demonstrated through simulation results.