Automatic Control of Low Voltage Load in Power Systems Based on Deep Learning

Abstract

Due to the interference of false data, there is a large error in the mining results of low voltage loads in the power system. In response to this problem, the author proposes a design of an intelligent mining system for low voltage loads in the power system based on deep learning. Using ARM+DSP dual CPU structure, initializing the adapter agent, and using dual arm spiral antennas, designing a low-voltage load monitor to detect partial discharge signals in the 500-1500 MHz frequency band and suppress noise interference; By transmitting monitoring information to the intelligent switch through CAN bus or 485 bus, remote monitoring can be achieved; Based on the contact points and current characteristics of the circuit breaker, a current transformer has been designed to reduce the range of induced voltage variation; Construct a continuous set of functions MMD in the space, adjust the original network structure, establish a deep learning mining model, initial network parameters, eliminate false data in the network, optimize the network using target domain data, and combine mining engines to achieve intelligent data mining. According to the experimental results, the maximum difference between the load of phase A of the data processing system based on numerical simulation and the actual data is 1000 kVA at a time of 6 seconds; When the load of phase B is 4 seconds, the maximum difference between it and the actual data is 2000 kVA; When the load of phase C is 8 seconds, the maximum difference between it and the actual data is 2000 kVA. It has been proven that the mining error of the system is 0, and it has a precise mining effect.