The Analysis Of Non-normal Power Line Status Using Deep Learning Algorithms Based On Embedding Parallel Computing Architecture

The power line is responsible for real-time transportation and distribution of electric energy in the power system.Its safe and stable high-quality operation state is related to the stable operation of the entire power system.According to the function of the power line in the power system,it can be divided into: distribution line and transmission line,which respectively undertake the point distribution and transmission functions.The expansion of power grid scale and the rapid development of power electronic equipment have led to more complicated power line conditions and greater potential risks.At present,there are three problems in the abnormal analysis of power lines: the increasingly complex power distribution lines generate more complex power quality problems,the expansion of the power grid needs to update the transmission line short-circuit fault diagnosis algorithm and the power line online monitoring system has intelligent requirements.The basis for solving power quality problems is to accurately identify and classify the types of disturbances.In this paper,the deep learning algorithm based on deep confidence network is applied to the identification and classification of power quality disturbance,and the disturbance characteristics are directly learned from the disturbance signal,which reduces the subjective dependence on the artificial extraction feature,and the accuracy and algorithm anti-noise in recognition and classification.Excellent performance in sexuality.In order to solve the structural design problem of deep confidence network,this paper firstly uses t-SNE dimensionality reduction to analyze the data characteristics of high-dimensional disturbance signals,so as to solve the hidden layer number problem of deep confidence network;then proposes self-organizing structure based on feature migration.The optimization mechanism automatically finds the optimal number of hidden layer neurons to realize the deep confidence network structure optimization problem.Secondly,through comparative experiments,the better training parameters are obtained,thus improving the training efficiency.Finally,compared with the existing algorithms,the proposed algorithm Excellent performance in classification accuracy and anti-noise.Aiming at the problem of parameter variation of transmission line with grid expansion,this paper proposes a short-circuit fault diagnosis algorithm based on long-term and short-term memory loop network,which not only obtains good diagnostic results in the case of variable parameters,but also satisfies the real-time nature of rapid diagnosis.Claim.In addition,in order to further improve the accuracy of short-circuit fault classification and reduce the error of fault distance judgment,this paper proposes a long-short-term memory loop network training method based on local prediction.Compared with the existing algorithms,it has the accuracy and distance judgment error.Great advantage.In addition,in order to further improve the intelligence of the existing online monitoring system for abnormal power lines,this paper proposes an online monitoring system based on embedded deep learning architecture chip intelligent terminal.In this paper,the embedded deep learning architecture chip is used as the arithmetic processor to offline and run the above-mentioned deep learning algorithm based on the deep learning algorithm.This reduces the dependence of the online monitoring system on the real-time performance of the network.The method of abnormal sample uploading,offline training,and network update intelligent terminal algorithm enables the online monitoring system to continuously learn the characteristics of the abnormal state of the power line with the development of the power grid,thereby improving the intelligence of the online monitoring system.In this paper,the real-time digital system is used to generate the short-circuit fault signal of the transmission line,and the intelligent terminal is used to conduct the online diagnosis of the short-circuit fault of the transmission line.The experimental results show that the proposed algorithm can run in real time on the intelligent terminal and obtain better performance.In summary,this paper analyzes the abnormal state of power lines,and proposes a power quality disturbance recognition and classification algorithm based on self-organized deep confidence network and a short circuit fault diagnosis algorithm based on local prediction to strengthen the long and short time memory loop network.An embedded parallel computing architecture deep learning intelligent terminal,and successfully applied the proposed algorithm to the intelligent terminal,and finally got better performance.