Abstract: The change of power system operation mode can cause the change of the power equipment parameters, operation mode of the calculation with the power grid is gradually increasing with the increase of complexity. In order to solve the heavy workload and repetitive tasks, this paper describes how to take the advantage of TensorFlow architecture and machine learning algorithm to study thegrid operation mode calculation. The linear regression algorithm, logical regression two-classification algorithm and depth neural network algorithm are analyzed comparatively. Index of performance evaluation can be provided combining with the actual requirements.