Heating, ventilation, and air-conditioning (HVAC) systems consume approximately 60% of the total energy consumption in public buildings, and an effective way to reduce HVAC energy consumption is to provide accurate load forecasting. This paper proposes a load forecasting model CEEMDAN-Conv1D-BiLSTM-AM which combines empirical mode decomposition and neural networks. The load data are decomposed into fifteen sub-sequences using complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). The neural network inputs consist of the decomposition results and five exogenous variables. The neural networks contain a one-dimensional convolutional layer, a BiLSTM layer, and an attention mechanism layer. The Conv1D is employed to extract deep features from each input variable, while BiLSTM and the attention mechanism layer are used to learn the characteristics of the load time series. The five exogenous variables are selected based on the correlation analysis between external factors and load series, and the number of input steps for the model is determined through autocorrelation analysis of the load series. The performance of CEEMDAN-Conv1D-BiLSTM-AM is compared with that of five other models and the results show that the proposed model has a higher prediction accuracy than other models.

The effectiveness of a cruise ship’s emergency evacuation is greatly influenced by the way people interact; this paper uses the social force model to simulate two different evacuation scenarios considering the impact of small groups. It uses an agent to simulate the behavior of a single occupant, and leverages the social force model to quantify the effect of group behavior on the group members. According to the influence of the group on the members, this paper corrects the expected speed of the members to determine the speed of crowd evacuation. It uses the SAFEGUARD cruise ship as the evacuation platform to simulate the process of evacuating the passengers to the boarding station and assembly station, respectively, and calculates the evacuation time, congestion area, and congestion duration of passengers under the action of groups. The simulation results of the two scenarios show that the group effect increases the average evacuation time by 15.29% and 21.79%, and increases the average detour distance by 24.54% and 17.89%, respectively.

(1) Background: Voiceprint recognition technology uses individual vocal characteristics for identity authentication and faces many challenges in cross-scenario applications. The sound environment, device characteristics, and recording conditions in different scenarios cause changes in sound features, which, in turn, affect the accuracy of voiceprint recognition. (2) Methods: Based on the latest trends in deep learning, this paper uses the perceptual wavelet packet entropy (PWPE) method to extract the basic voiceprint features of the speaker before using the efficient channel attention (ECA) block and the Res2Net block to extract deep features. The PWPE block removes the effect of environmental noise on voiceprint features, so the perceptual wavelet packet entropy-guided ECA–Res2Net–Time-Delay-Neural-Network (PWPE-ECA-Res2Net-TDNN) model shows an excellent robustness. The ECA-Res2Net-TDNN block uses temporal statistical pooling with a multi-head attention mechanism to weight frame-level audio features, resulting in a weighted average of the final representation of the speech-level feature vectors. The sub-center ArcFace loss function is used to enhance intra-class compactness and inter-class differences, avoiding classification via output value alone like the softmax loss function. Based on the aforementioned elements, the PWPE-ECA-Res2Net-TDNN model for speaker recognition is designed to extract speaker feature embeddings more efficiently in cross-scenario applications. (3) Conclusions: The experimental results demonstrate that, compared to the ECAPA-TDNN model using MFCC features, the PWPE-based ECAPA-TDNN model performs better in terms of cross-scene recognition accuracy, exhibiting a stronger robustness and better noise resistance. Furthermore, the model maintains a relatively short recognition time even under the highest recognition rate conditions. Finally, a set of ablation experiments targeting each module of the proposed model is conducted. The results indicate that each module contributes to an improvement in the recognition performance.