Traditional power system operation quality control ignores the role of high-voltage distribution network topology on power flow transfer. In this paper, the high-voltage distribution network is included in the proposed algorithm of urban transmission network operation quality control, and the operation quality of the transmission network will be improved through the reconstruction of the topology of the high-voltage distribution network. A large amount of topology data of high-voltage distribution network is generated by Markov chain Monte Carlo sampling, and the operational quality of each topology is calculated, including line loss, bus voltage, line load ratio and section load ratio. A deep neural network is used to fit the nonlinear relationship between the topology of a high-voltage distribution network and the above state parameters, and the deep neural network-based surrogate model of urban transmission network operation quality estimation is generated. The data driven surrogate model can realize fast and efficient transmission network state estimation. Then, the model is embedded in the optimization calculation of the Non-Dominated Sorting Genetic Algorithm (NSGA-II), and the topology of the high-voltage distribution network is iterated to find the topology reconstruction strategy which can improve the operational quality of the urban transmission network. The algorithm is verified in a city power grid, which improves the operational quality of the city power grid significantly. This work is supported by the National Natural Science Foundation of China (No. 51977133) and the Science and Technology Project of State Grid Sichuan Electric Power Company (No. 52199718001A).