摘要：【中文】本申请公开了用于解决对时误差的暂态同源数据比对方法，包括获取同一故障采样单元和保护装置的录波文件，获得采样单元的采样零时刻点以及保护装置对应的采样零时刻点，计算角度误差，通过角度误差修正采集单元通道和保护装置通道的数据，进行比对报警。本发明能够通过获取采集单元和保护装置每次启动的录波文件来实时监测同数据源的差值情况，并且有直观告警提示，提高智能变电站运维检修效率。 【EN】The application discloses a transient homologous data comparison method for solving time setting errors, which comprises the steps of obtaining a recording file of a same fault sampling unit and a protection device, obtaining a sampling zero-time point of the sampling unit and a sampling zero-time point corresponding to the protection device, calculating an angle error, correcting data of a channel of the sampling unit and a channel of the protection device through the angle error, and performing comparison and alarm. The method and the device can monitor the difference condition of the same data source in real time by acquiring the recording file started by the acquisition unit and the protection device each time, and have visual alarm prompt, thereby improving the operation maintenance and overhaul efficiency of the intelligent substation.

摘要：【中文】本发明的目的是在于针对现有技术的缺陷和不足，提供基于发音量化和电力专用词库的离线语音识别装置，还提供基于发音量化和电力专用词库的离线语音识别方法，从而提高智能语音识别的效率和准确率，避免由于识别产生误差造成的损失，同时也提高了语音识别的效率和准确率。对此本发明提供了一种基于发音量化和电力专用词库的离线语音识别装置，包括与音频采集系统连接的通用语音识别模块，通用语音识别模块输出语音词语，还包括：输入特征提取模块、电力专用词库特征提取模块、语音词匹配检索模块、识别结果输出模块。本发明能够根据具体应用，修改编辑电力专用词库，也能够根据具体语音噪声环境修改量化值权重值，进而优化识别结果。 【EN】The invention aims to provide an offline voice recognition device based on pronunciation quantification and a power special word stock and an offline voice recognition method based on pronunciation quantification and a power special word stock aiming at the defects and the defects of the prior art, so that the efficiency and the accuracy of intelligent voice recognition are improved, the loss caused by errors generated by recognition is avoided, and the efficiency and the accuracy of voice recognition are also improved. The invention provides an off-line speech recognition device based on pronunciation quantification and a special word bank for electric power, which comprises a general speech recognition module connected with an audio acquisition system, wherein the general speech recognition module outputs speech words, and the off-line speech recognition device also comprises: the system comprises an input feature extraction module, an electric power special word bank feature extraction module, a voice word matching retrieval module and a recognition result output module. The method can modify and edit the electric power special word stock according to specific application, and can also modify the weight value of the quantization value according to specific voice noise environment, thereby optimizing the recognition result.