Abstract: In recent years, the construction of a new power system with new energy as the main body has been accelerating continuously. The contradiction between the large-scale access of distributed energy and its efficient consumption has become increasingly prominent. Virtual power plant (VPP) has become the key to solving this problem. Currently, the development of VPP in China is in the invitation stage. Although it mainly focuses on power demand response, the existing operation mode lacks full-process coordination and intelligent support, making it difficult to meet the actual dispatching requirements. Therefore, a VPP management platform for power demand response has been designed and developed, covering the full-process closed-loop management of invitation issuance, user application, demand allocation, capacity response, and load monitoring. It integrates the long short-term memory ( LSTM) model and the Ollama large model to achieve precise load prediction and natural interaction, and builds a core architecture of fullprocess control and intelligent empowerment. This platform provides an efficient collaborative carrier for power grid dispatching departments, energy aggregators, and distributed resource owners, which can improve the response speed, control accuracy, and commercial efficiency of VPP, and provide technical support for the safe and stable operation of the new power system.