The abstract of this paper covers a key topic, that is, the research on two-way transmission of source and load in energy Internet based on the analysis of source and load characteristics of power grid. This paper emphasizes the importance of power system, which is one of the key infrastructures to support modern society, but it is also facing more and more complex challenges. One of them is the rapid growth of renewable energy, and its volatility and uncertainty pose a challenge to the management and operation of power system. In the research, this paper discusses the different characteristics and modes of power load curve, including intraday load curve, seasonal change and the influence of renewable energy. Through the analysis of source-load characteristics of power grid, this paper emphasizes the importance of a deep understanding of power system for better planning and management of power supply. Comparing the energy storage capacity of different load forecasting models, the maximum energy storage capacity of this model can reach 3556 MW, while SVM (Support Vector Machine) is 1000 MW and RNN (Recurrent Neural Network) is 1700 MW. It can be seen that this model can achieve higher utilization rate of renewable energy and reliability of power system. To sum up, the summary of this study highlights the importance and challenges in the field of power system, and emphasizes the potential value of source-load characteristic analysis and energy Internet technology. These findings are of great significance for realizing clean, sustainable and efficient power supply, and provide useful insights for power system planning and development.