Aiming at the protection of user sensitive information in smart grid terminals，and to counter the privacy inference attack on user sensitive information，this paper proposes an efficient similarity matching based data privacy-protection scheme for edge terminals in smart grid. First of all，this paper proves that the user’s sensitive information can be inferred based on the time series data prediction model. In order to resist the privacy inference，the FastDTW algorithm is used to identify the time series data related to sensitive information，and to achieve the purpose of privacy protection by disturbing the relevant data and reducing the inference accuracy. The experiment results show that the privacy protection method proposed in this paper can make the attacker unable to accurately obtain the actual power consumption of the user，thus effectively protect the user’s power consumption information.