Abstract: Timely detection and control of abnormal production information is among the key and difficult points in the continuous power production. A method based on alarm frequency, information features, and other optimization of voice settings is proposed to address the issue that the central control center is not able to detect anomalies in a timely manner, which increases the risks of system operation. This method enables the center to analyze the frequency, duration, spatial distribution and other characteristics of voice broadcasting, optimize voice configuration, improve text expression, and adopt automated generation technology so as to improve the timeliness and effectiveness of production alarms, of which the effectiveness has been proven upon operation.