Research On Network Loss Rate Calculation Based On Deep Migration Learning And Optimization Of Wind Power Delivery System

As a new type of renewable energy,wind power has become an important part of sustainable development strategies.However,wind power integration also brings new challenges to the power system.Due to the characteristics of volatility,intermittentness and randomness of wind power output,which makes the shortcomings of traditional power flow calculation easy to have no solution,no convergence and poor data fault tolerance more prominent,it also have an impact on network loss rate and power flow.Therefore,it is of great significance to establish a high-precision and strong data fault tolerance network loss rate calculation model,and optimize the loss reduction research for large-scale wind power transmission grids.In this paper,the relevant analysis and research are carried out,and the specific contents are as follows:First of all,the theory of deep learning is studied,the theory and characteristics of training and learning of three basic structural units commonly used in deep learning models are analyzed;The necessity of migration learning is studied,also the theory features and principles of four learning models in migration learning are analyzed;the importance of metric data distribution differences in the process of big data research is analyzed,and three methods for measuring the difference in data distribution is analyzed.Then,a large-scale wind power transmission network loss rate calculation model based on deep migration learning is proposed.The deep migration learning(Deep Boltzmann Network-Deep Neural Network,TDBN-DNN)network loss rate calculation mode is obtained based on the deep learning model(Deep Boltzmann Network-Deep Neural Network,DBN-DNN)having being trained that is fine-tuned by samples from the source training data are more closely related to the data to be calculated by defining the maximum mean difference coefficient.Simulation results show that the DBN-DNN calculation method has better nonlinear fitting ability than the traditional shallow BP neural network calculation method;the TDBN-DNN has higher precision than the DBN-DNN model,moreover,the TDBN-DNN model can still be calculated in the absence of calculated data,and has certain data fault tolerance,which verifies the validity of the model.Finally,the optimal power output of other power sources is sought when the network loss rate is minimum by the optimal power flow model based on the particle swarm optimization algorithm for large-scale wind power transmission grid.The correlation rules and K-means method are used to analyze the influence on the space for loss reduction of optimization variables and wind power.In the simulation example,the results of optimal power flow is analyzed in comparison with the actual operation;the optimal network loss rate is calculated by TDBN-DNN model,and the excellent calculation performance of TDBN-DNN model is verified again.;the power grid is easy to deviate from the optimal operating state when the wind power fluctuates greatly,and has further space for loss reduction at this time.