Abstract: In the context of a new power system, the high penetration of new energy units continuously weakens the regulation capacity on the generation side. Meanwhile, the adjustable resources on the demand side possess substantial potential for adjustment. There is a wide variety of load types on the demand side, each exhibiting complex electrical characteristics. Employing a data-driven approach can effectively address the challenges of precise modeling and scheduling management on the demand side. Therefore, an energy storage and multi-user demand response optimization strategy is proposed based on data-driven and customer directrix load. Firstly, an assessment is conducted on the adjustment potential of various adjustable resources on the demand side participating in demand response. Secondly, considering the response characteristics of the classified loads, the grid company provides differentiated rewards to different users participating in demand response. Thirdly, the demand response optimization model is transformed into a Markov Decision Process, shaping the load curve throughout the entire time period to closely align with the target directrix, thereby minimizing the overall operational cost of the grid. Finally, the experimental results indicate that the customer directrix load demand response not only significantly decreases the overall operational cost of the park's grid but also effectively accommodates new energy locally. Compared to different types of loads responding to the directrix separately, the effect of aggregating adjustable industrial, commercial, and residential loads with energy storage in a park microgrid within a region yields superior results in responding the directrix.