Search

In the field of computer vision, human behavior recognition technology is still a hot topic. Especially in the field of electric power, in order to identify the behavior of the staff in the power grid area, it is necessary to use human behavior recognition algorithm to establish behavior recognition model. In order to solve the problem of behavior recognition of Substation Scene staff, this paper proposes the use of deep learning action classification technology to realize the behavior analysis of personnel, and selects 3D convolution neural network to identify the behavior of personnel in the video. Firstly, the influence of different feature extraction networks and different network parameters on behavior recognition is studied. The network is improved by adjusting network parameters and model optimization method, and the 3D network parameters that are most conducive to behavior recognition are selected. Then, the spatial flow information and time flow information in video data are extracted as the input of behavior recognition model, so as to further improve the behavior recognition Finally, the video data collected from substation is used to verify the model. The experimental results show that: the deep learning based human behavior recognition technology can accurately identify the behavior of people from the video region.

Real-time pricing (RTP) is the most ideal pricing mechanism in demand side management, but when in the face of short-term power shortage, its price will become extremely high. In this paper we propose an algorithm to improve the realtime pricing mechanism by offering compensation price to users who could provide adequate power consumption reduction. By utilizing the theory of Stackelberg game, we establish the optimal pricing model between the demand side and the supply side. Simulation results confirm that the proposed distributed algorithm can achieve a better performance during the power shortage.