The main task of the building structure array is to record the failure process of civil engineering structures in detail, and to provide structural response information for many related studies such as seismic design, seismic damage assessment, and earthquake safety alarm. However, due to the constraints of economic cost, field testing technology and data processing level, it is unrealistic to deploy sensor monitoring equipment on all floors of the entire structure, so how to obtain the most complete structural information with the least number of sensors is the purpose of optimizing the layout of structural array sensors. Considering the advantages and disadvantages of the effective independent method and the modal kinetic energy method, a unit stiffness energy-driving point retention method was proposed, considering the advantages and disadvantages of the effective independent method and the modal kinetic energy method. In this method, the unit stiffness modal energy is used as the information matrix, and the principle of effective independence method is used to screen the measurement points, so as to ensure that the high-energy measurement points maintain linear independence to the greatest extent. Finally, taking a steel frame as an example, the proposed method, the effective independence method, the modal kinetic energy method and the unit stiffness method are used to lay the sensors on the model respectively, and the modal assurance criterion and the Fisher information matrix criterion are used to evaluate the layout results of the four methods. The results show that, compared with the other three methods, the proposed method has the least number of sensors when the mode vectors are linearly independent, and the proposed method can obtain the most modal information with the same number of sensors.