<p>This paper presents a data-driven economic model predictive control (EMPC) framework for nonlinear systems. Leveraging Koopman operator theory and the extended dynamic mode decomposition method, a lifted linear model in the high-dimensional function space of the nonlinear dynamics is first identified from the collected dataset. Then, an EMPC strategy used to optimize process economics is designed in the lifted space, which employs the Koopman linear model as the predictor. To guarantee closed-loop stability, an artificial constraint is constructed by solving a convex quadratic programming problem. The recursive feasibility and closed-loop stability of the proposed approach are rigorously analyzed. Benefiting from the linear structure of the Koopman model, the online computational burden of the EMPC is substantially reduced. The effectiveness of the proposed method is demonstrated through simulations on a nonlinear chemical reactor.</p>