The traditional distribution automation system, demand side management system, and distributed generation access and control system, respectively, solve the problems of regional distribution network power supply, customer power consumption, and new energy utilization to varying degrees. Intelligent Park evolved from the concept of intelligent park of State Grid Corporation, Modern enterprises, or residential areas that can make full use of modern communication, computer, automation, and other technologies to implement the power supply demand of the park. For the modern intelligent park with user-side temperature control load and demand response load access, an intelligent park energy management and optimal scheduling method based on deep reinforcement learning (DRL) algorithm is proposed. Through the interaction between the agent and the energy environment of the intelligent park, the control strategy is adaptively learned. This method can realize the continuous action control of energy management in intelligent park and can realize the optimal scheduling decision of intelligent park in various scenarios. Firstly, based on the characteristics and types of intelligent park aggregation unit, the environment model of intelligent park energy management system interacting with agent is established. Secondly, the basic principle of deep deterministic policy gradient deep algorithm is introduced, and on this basis, the key links of deep reinforcement learning algorithm, such as action space, state space, reward mechanism, neural network structure, and learning process, are designed. Finally, the effectiveness of the proposed algorithm is verified by an example of a city intelligent park.