If the undesired situations such as a transient or an accident improperly affecting normal operation occur in nuclear power plants (NPPs), accurately checking the NPP state by the operators using temporary trends of several instrumentation signals in a short time can be constrained. Therefore, this study was carried out to provide the transient identification information to the operators in a short time after the reactor trip according to the abnormal circumstance occurrence using the deep learning since the diagnosis of the NPP states is prior for effective accident management. To establish the deep learning model identifying the initial events of the NPPs, the simulated accident data were applied to train the deep learning model. These data were obtained by simulating the postulated scenarios using the modular accident analysis program (MAAP). The data from the MAAP code are used to calculate the time-integrated values of the simulated instrumentation signals. That is, the deep learning model is trained to find the optimized classifier to identify the events using the simulated signals of the accident data showing the behaviors of each accident circumstance. Utilized simulated signals were considered as some of the highly correlative accident monitoring variables. In this study, deep neural networks (DNNs) were used for identifying the transients of the NPPs. The identification performance of the DNN model, and moreover the support vector machine (SVM) model in the previous study is able to be checked in this paper. In addition, performance of the artificial intelligence methods as advanced technologies monitoring and diagnosing the NPP states can be assessed.