Abstract: The scheduling of cooking stoves during the peak period is the key activity to "shorten the waiting time, improve customer satisfaction and achieve revenue growth" for the catering enterprises with take-out business. Based on the characteristics, which are “multiple stoves, variety of dishes in one order with a same pick-up time, same dishes can be combined”, a double-layer scheduling model is constructed. In the model, the lower layer is a multi-stove dish-package scheduling model based on parallel machine scheduling, and the upper is an order selection model based on knapsack problem. In order to solve the model in a reasonable time and extend the model to the dynamic scheduling case, a specific genetic algorithm is proposed. Numerical experiment is designed to evaluate the convergence and efficiency of the algorithm, meanwhile the solution is compared with 3 common strategies, which are the most urgent next, the longest process next and the most popular dish next. The results show that this model can achieve the highest order-finishing rate and the shortest overtime.