Calculation And Analysis Methods Of Theoretical Line Loss Rate Based On Deep Learning

Grid power loss(abbreviated as network loss)is one of the important indexes for assessing the operating level of power supply companies.Theoretical line loss rate calculation and analysis are important methods for formulating loss reduction plans and improving line loss management and grid operation levels,and reducing theoretical line losses of operation of power supply companies is an important key initiatives to realize high efficiency,energy saving and environmental protection of power grid.At present,the power system is becoming larger and larger,and at the same time,the requirements for calculation and analysis speed are becoming higher and higher.Therefore,from the perspective of reducing the calculation scale of the power grid and increasing the speed of calculation,a method of calculation and analysis of theoretical line loss rate based on deep learning is presented is.The main contents are as follows:Firstly,studying the theory of deep learning and learning about the training methods of deep learning and a variety of deep learning model,basis on this,A deep learning model that is fit for calculation of The theoretical line loss rate based on deep blief network network and deep neural network combination is selected,and a deep learning model that is fit for analysis of The theoretical line loss rate based on deep blief network network is selected,and The problems of initial parameter sensitivity,convergence speed,overfitting problem and incomplete calculation result are respectively improved.Secondly,the influence factors of theoretical line loss rate are analyzed by using the theory of electrical network,and the influence of network parameter change and network frame structure change on theoretical line loss rate is analyzed.On this basis,the calculation model of the theoretical line loss rate based on deep belief network and deep neural network is builded,especially,the calculation model of the theoretical line loss rate based on deep belief network and deep neural network is builded under the condition of the network parameters and the network structure changes.A deep learning greedy layer training method using BP fine-tuning after layer-by-layer pre-training using Restricted Boltzman machine and Random small batch gradient descent method(MSGD)and Dropout method is used to train deep learing model,the validity of this method is verified by an example.Finally,using the method of the association rules of data mining technology to analyze the connection degree of each generator and load and theoretical line loss rate,and remove the generator and load which are weakly associated with the theoretical line loss rate.On the basis of this,a analysis method of theoretical line loss rate based on Deep Belief network(DBN)is proposed.The generator and load data mined by association rules that are strongly associated with theoretical line loss rate are used as input data of the deep belief network model,and the size of the deviation of the theoretical line loss rate with the optimal theoretical line loss rate is used as output.This method is used to analyze the the size of the deviation of the theoretical line loss rate with the optimal theoretical line loss rate,and the validity of the method can be verified by an example simulation.