Energy Reports (Nov 2022)
Harmonic current control strategy of DC distribution network based on deep learning algorithm
Abstract
In today’s power grid configuration system, due to the unique advantages of high efficiency, strong power supply capacity, and good energy utilization, DC distribution network gradually replaced AC distribution and became a more popular distribution system. However, the harmonic current control method in the DC distribution network often cannot keep up with the performance requirements of the entire power system. The design of the DC distribution network structure, the selection and matching of current control methods, which has a significant impact on the healthy operation of the entire system. Traditional current control methods often cannot take into account both quality and efficiency in current effect, and in recent years, current control methods have been difficult to innovate. Aiming at the above problems, this paper combines PI control and repetitive control and optimizes on this basis, and creatively proposes a new harmonic current control strategy. And based on deep learning related algorithms, a convolutional neural network model that is more in line with the current control concept is selected to build, and ADMM algorithm is introduced to improve the model. Experiments show that the accuracy of the optimized deep residual convolutional neural network model can reach up to 93.1%, the minimum value in error control is 0.06, which well meets the needs of the current control strategy.
