Adopting neural network technology with self-learning ability, establish the structural model of neural network algorithm, use unlabeled data set to train sparse autoencoder, use sparse autoencoder to extract features from labeled training sample set and test sample set, extract training set features and their label sets, train feature classifier, accurately identify control strategies for resistive heating loads (malignant loads), and do not affect the use of compliant loads. Different from traditional power control, the identification is more accurate and smarter. With load AI identification technology, it can identify all kinds of prohibited campus electrical appliances from 10W to 6000W, including voltage regulating sockets that are rampant in the market. At the same time, students' normal life and study appliances can be used normally regardless of power (such as: desktop computers, notebooks, desk lamps, fans, etc.)

Working voltage AC 110V ~ 265V

Rated current 25A

Power consumption ≤1.5VA

Rated power 6000W

Automatic power-on times after power failure 3 times (adjustable)

Automatic power-on time after power failure 5 minutes (adjustable)

Sound and light alarm

Multiple communication methods: MODBUS-485 or 4G

It is used in collective dormitories in schools, factories, mines, construction sites, hospital wards and other places where the use of malicious loads is prohibited.