Gis partial discharge diagnosing method, model training method, device and system

A GIS partial discharge diagnosing method applied to a GIS partial discharge diagnosing device, wherein the GIS partial discharge diagnosing device comprises a monitoring host, a network relay device and a plurality of sensor modules distributed on a GIS, the monitoring host is computer equipment or server equipment, and the sensor modules are used for monitoring ultra high frequency signals generated when GIS partial discharge occurs, and the ultra high frequency signals are electromagnetic wave signals in a range of 300-3000 MHz;

• each of the sensor modules comprises a wireless transmission device and is in communication with the network relay device, and the sensor modules are connected with each other through an ad-hoc network of the wireless transmission devices to form a wireless transmission network, each of the sensor modules obtains a distance from other sensor modules through a wireless communication time, so as to obtain sensor network position distribution data of each of the sensor modules in the wireless transmission network, and the network position distribution data are sent to the network relay device, and the monitoring host is in communication with the network relay device;
• the diagnosing method comprises:
• obtaining, by the monitoring host, monitoring data of each sensor module, the monitoring data comprising partial discharge ultra high frequency signals monitored by each sensor module, receiving time information of the partial discharge ultra high frequency signals monitored by each sensor module and sensor network position distribution data of the wireless transmission network;
• inputting, by the monitoring host, the monitoring data into a partial discharge diagnosing model for prediction processing, and determining a predicted partial discharge type;
• wherein, the partial discharge diagnosing model is obtained by the following training method:
• obtaining, by the monitoring host, the partial discharge diagnosing model from a training server;
• obtaining, by the monitoring host, training data, the training data comprising training feature data and labeling information;
• the training feature data comprising the partial discharge ultra high frequency signals monitored by each sensor module, the receiving time information of the partial discharge ultra high frequency signals monitored by each sensor module and the sensor network position distribution data of the wireless transmission network;
• the labeling information comprising a partial discharge type;
• inputting, by the monitoring host and after vectorizing the training feature data, the vectorized training feature data into the partial discharge diagnosing model for prediction processing, and determining a predicted partial discharge type;
• and
• correcting, by the monitoring host, the partial discharge diagnosing model according to the predicted partial discharge type.