Ice storage technology, being a pivotal technology for demand response, can be deployed to facilitate the scheduling of grid loads by implementing cold storage mode. In this study, we propose a day-ahead optimized scheduling (DOS) model grounded in load forecasting. During the construction of the forecasting model, the support vector machine (SVM) is employed to determine the optimal parameters. To avoid falling into local optima, multiple algorithms are introduced to optimize the parameters, thereby forming combined forecasting models, and the accuracy of these combined prediction models is verified through case studies. Based on historical operational data, we establish both the energy consumption model of the ice storage air-conditioning system (ISACS) and the DOS model. Subsequently, three typical days were chosen for the DOS analysis aimed at minimizing operational costs. A comparison is made between energy consumption patterns before and after the implementation of DOS. The DOS model effectively reduces the overall operating costs of ISACS while optimizing its power consumption patterns. The operating costs for three typical days decreased by 2.9%, 6.9%, and 3.6%, respectively, with the implementation of load forecasting-based DOS.