Research On Theoretical Line Loss Calculation Of Power Grid Based On Deep Learning

Theoretical line loss is an important economic indicator for power supply enterprises.It reflects the level of power system planning,operation and management.Therefore,power supply companies at all levels must perform line loss calculation and assessment.Accurate calculation of line loss helps to reasonably reduce energy loss,save resources,and formulate loss reduction plans to improve economic efficiency.This paper aims at the theoretical line loss calculation problem of low-voltage station areas and wind power grids,and combined with deep learning,this paper proposes a deep learning-based theoretical line loss calculation study for power grids.The specific contents are as follows:First,the development of deep learning is outlined from perceptrons and perceptrons to neural networks,from neural networks to deep learning;and deep learning theory is studied.The structural characteristics,basic units and training methods of the three neural networks,memory network and gated loop unit,were analyzed and studied.Secondly,a theoretical line loss calculation method based on deep boltzmann network network low voltage station area is proposed.The deep boltzmann network is used to describe the complex system,and the historical data is used to train the mapping model between the line loss and the sample feature value of the station area.In the training process,the model firstly uses the greedy algorithm to perform unsupervised pre-training on the network layer of the model layer by layer,and then performs supervised global fine-tuning on the deep confidence network model.In order to improve the accuracy of calculation,Adam optimizer is used in the training process.And using 2140 measured data of a certain area as a sample for simulation calculation,it is verified that the deep boltzmann network line loss calculation model has better generalization ability,better accuracy and faster speed than the shallow neural network.At the same time,the superiority of Adam optimizer compared with RMSProp and SGD in line loss calculation is also verified.Finally,for the wind power grid,a theoretical line loss calculation method based on long short-term memory grid is proposed.The long short-term memory model has a good effect on the processing of time-series data.A stacked long short-term memory network model is built,and the number of hidden layer layers,hidden layer neuron nodes,activation functions and training are determined through a large number of experiments.batch.It is verified by simulation that the stacked long short-term memory is superior to gated recurrent unit model in the calculation of wind-containing wire loss;and the stacked long short-term memory model can still be calculated in the absence of input sample data,and has certain data The fault tolerance of the model is verified by experiments.