| Civil engineering is always surrounded by the natural environment, which is affected by various factors, such as temperature variations. Temperature effects, as a nonlinear factor, is difficult to be analyzed quantitatively. Meanwhile, it will influence the results of modal testing and cause the errors in the measured dynamic response data, which set up obstacles to the evaluation of real structural damage situation. Furthermore, damage identification method based on optimization algorithm is easy to be trapped in local optimal and lower computing efficiency when the method has been used to identify damage location and extent. Aiming to the above problems, in this paper, a damage identification method, which is based on support vector machine (SVM) and enhanced moth-flame optimization (EMFO), is proposed to solve structural sparse damage identification problem considering temperature variations. Firstly, SVM is used to quantify structural temperature variations and eliminate the temperature effects. Then, sparse regularization method is introduced to determine structural sparse damage condition. Thirdly, the temperature variations and damage situation obtained in the previous step are adopted to perform the initialization of EMFO, which can narrow search space, improve efficiency and enhance the accuracy of damage identification. Finally, two examples, a numerical simply supported beam considering temperature variations and random noise effects, and a practical engineering of I-40 Bridge, a large steel-concrete composite bridge, are utilized to verify the effectiveness of the proposed method.