Based on deep learning, a fully connected neural network-decision tree algorithm is designed for intrusion detection of full data link in power grid collaborative business scenarios. This paper analyzes the causes of the problems in the power grid collaborative business scenario, and introduces the common classification algorithms. It combines the fully connected neural network in deep learning with the decision tree algorithm in machine learning, and obtains the designed fully connected neural network-decision tree intrusion detection algorithm model. Through the test on the relevant data, the detection accuracy of the model in network intrusion can reach 0.99, the precision can reach 0.984, the recall rate is 0.97, and the F1 score is 0.977. The results show that compared with similar algorithms, the proposed algorithm has outstanding advantages, provides more accurate technical support for the full data link in the grid collaborative business scenario, and plays an important role in the optimization of power grid planning.