Optimization and Implementation of Deep Learning Algorithm in Power Dispatch Speech Recognition

With the rapid development of smart grid, the application demand of speech recognition technology for power dispatch is increasing. In this paper, a deep learning-based speech recognition technology for power dispatch is proposed to solve the problem of recognizing professional terms and regional terms in smart grid dispatch. We constructed a base corpus containing special terms for power dispatch and dispatching terms for a regional power grid, and used deep Neural network Hidden Markov Model (DNN-HMM) for acoustic model training to improve the recognition accuracy. In the process of research, we discussed the whole process of making speech materials, speech model training, endpoint detection, interaction with D5000 system and speech-to-text in detail. The practical results show that the accuracy of this technology reaches 94.63%, which is significantly better than the traditional Gaussian mixture hidden Markov model. In addition, the power dispatch speech recognition system developed based on the proposed method has been successfully applied in a regional power grid control center, which proves its feasibility and superiority. This achievement not only provides a convenient speech recognition tool for power dispatching personnel, but also improves the efficiency and accuracy of power dispatching, which is of great significance for the dispatching and management of smart grid. In short, the speech recognition technology of power dispatch based on deep learning proposed in this paper has achieved remarkable results in solving the problem of speech recognition in the field of power dispatch.