Low-voltage Overhead Line Loss Assessment Under The Disturbance Of Power Quality Based On Radial Basis Function Neural Network

With the continuous emergence of new power electronic devices such as distributed energy and electric vehicles,the power quality problem in power system is becoming more and more serious and complex,and the loss of network is increasing.Considering the factors of power quality,accurately evaluating the loss of low-voltage overhead lines has theoretical guiding value and engineering significance for the power industry to carry out power quality management,benefit analysis and promote energy-saving and loss reduction measures.This article summarizes the most common harmonic and three-phase unbalanced problems in power system,and deduces the calculation model of overhead line loss under their separate and joint action.The influence of power quality factors on overhead lines is analyzed by simulation.From the derivation process of the analytical calculation model,it can be seen that the physical analytical model is subject to the accuracy of parameter calculation,the calculation process is complex and the calculation difficulty increases with the increase of power quality problems.Aiming at the problems existing in the physical analysis model,a TRBF-CPSO line loss intelligent evaluation model based on Transformer network,radial basis neural network（RBF）and Crisscross particle swarm algorithm（CPSO）is proposed.The input of the model fully considers the influence of power quality factors,and is composed of 8-dimensional characteristics of phase and neutral line fundamental current,three-phase average total harmonic distortion rate and current unbalanced of each phase.In this article,it is the first time that Transformer network based on the Multi-Head Attention mechanism is used to extract the features of overhead line data,and the deep expression of the input features is obtained with the multi-level and multi-directional feature extraction ability.Then the extracted deep features are input to the RBF network to build a TRBF model,which makes up for the weakness of feature extraction of the RBF network,and evaluates the line loss by exerting its local response characteristics and strong nonlinear fitting ability;fully combines the advantages of supervised learning method and swarm intelligence optimization of parameter learning methods.After using the supervised training method to optimize the model parameters,the weights and biases of the output layer of the TRBF model are optimized by the CPSO algorithm and finally complete the construction of the TRBF-CPSO model.The optimal network structure of the TRBF-CPSO model is determined by experiments,and the effectiveness of the proposed model is verified by comparison.The results show that the feature extraction through Transformer network is conducive to improving the learning ability of TRBF network,and the loss evaluation accuracy is higher.The addition of CPSO intelligent algorithm for parameter optimization can effectively improve the loss evaluation ability of TRBF model and obtain the more accurate line loss evaluation result.The proposed TRPF-CPSO line loss evaluation model fully considers the influence of power quality,and can perform high-precision automatic evaluation of line loss.