Research On The Method Of Light Massive Data In Smart Grid

With the rapid development of smart grid in China,the degree of digitalization and informatization of power grid is getting higher and higher,and the safe production and operation of power grid rely on a lot of comprehensive information more and more.Smart grid accurately collects and efficiently transmits all kinds of real-time and non-real-time panoramic state information in wide area,and realizes"three flow"integration,high integration and sharing.Compared with traditional power grid monitoring system,smart grid wide area monitoring range,monitoring node number,monitoring information type and monitoring information amount are significantly increased.In the evolution process from smart grid to energy Internet,the vigorous development of new information makes the information of electric power communication network become complex and diversified.After the information converges layer by layer,it is transmitted through electric power communication network,which puts forward higher requirements for electric power communication network.With the development of smart grid,information system,marketing system,etc.,a large amount of data interaction has been generated,and the demand for bandwidth has risen sharply.The existing transmission network has been unable to meet the requirements,resulting in the failure to realize the advanced application system functions of the smart grid,which seriously affects the safe and stable operation of the grid.In order to meet the needs of on-line monitoring,transmission and storage of massive data in smart grid,realize the advanced application system functions of smart grid,and ensure the safe,stable and economic operation of power grid,this paper focuses on the light of massive data transmission and storage in smart grid,including compression sampling,non-uniform sampling based on low rank Hankel matrix,minimum feature information extraction,light protocol data generation.The main contents of this paper are as follows:(1)In view of type I signal(i.e.the data contains pulse signal or oscillation signal),a method of light sampling of large amount of data in smart grid based on compressed sensing theory is proposed.Based on the event triggering mechanism,the data in one observation window before and after the disturbance is recorded on-line,and the complete sampling record is made.The disturbance detection method is used to accurately locate the disturbance time.Then,the deep learning network is used to recognize the data quickly.The data is a type I signal,and each component of the data selects the strong correlation atom library in a certain order for sparse decomposition.The strong correlation atom library is a redundant atom library constructed according to the dynamic characteristics of the data components and the mathematical model,which can effectively improve the sparsity of the data components,reduce the total sparsity of the data and optimize the measurement matrix scale makes the amount of data compressed and sampled less.In order to enhance the real-time performance of data compression and sampling,measures such as setting inner product constant to reduce the size of the subset of the atomic library,using intelligent algorithm to speed up the atomic matching pursuit and using orthogonal projection matrix updating instead of sparse decomposition least square method to avoid solving the inverse matrix many times are taken to reduce the time of sparse decomposition of data components.Finally,Resonance-Based Sparse Signal Decomposition is used to enhance the extraction of the pulse data components,and the amplitude of the weak data components to be monitored is increased,and Harmonic Filtered algorithm is used to filter out,effectively improving the accuracy of data component reconstruction.The accuracy and effectiveness of compression sampling for type I signal are verified by an example.(2)In view of type II signal(i.e.data only contains similar fundamental wave signal and short-term variable signal),a light sampling method of smart grid massive data is proposed by using low rank matrix completion theory.The deep learning network is used for fast pattern recognition.The data is type II signal.The data is composed of Hankel matrix by operatorsΞ.The data is non-uniform sampled based on the low rank Hankel matrix to reduce the amount of data sampling.Firstly,the low rank property of Hankel matrix composed of signal data is proved mathematically,and it is concluded that the large amount of smart grid data has low rank property.The non-uniform sampling point is composed of the Hankel matrix whose elements are observed to be recovered by operatorsΞ.The matrix is completed with the recovery algorithm to recover the matrix,and the average value of the subdiagonal elements of the recovery matrix is obtained to recover the signal.When the relative error of the recovered signal meets the requirements,the position of the non-uniform sampling point is formed into a bit string u_v,which can be used to quickly determine the non-uniform sampling point of the data behind the observation window.Finally,in view of the disadvantages of randomness and large amount of calculation,Absolute Value of Gradient Difference method can be used to quickly determine the non-uniform sampling points,and low rank matrix completion can also simplify the calculation of the non-uniform sampling points by using the characteristics of data periodicity,symmetry,parity,etc.The accuracy and effectiveness of the non-uniform sampling method based on low rank Hanel matrix for type II signals are verified by an example.(3)Aiming at the light transmission of massive data in smart grid,a light transmission method of massive data in smart grid based on the principle of least characteristic information transmission is proposed.The data is type I signal,the data compression measured value uses reconstruction algorithm to reconstruct the data sparse representation vector,analyzes and interpolates the sparse representation vector to obtain the data component characteristic parameter,extracts the characteristic parameter and the sparse representation non-zero coefficient,the parameter less is the data component minimum characteristic information;the data is type II signal,the data non-uniform sampling point is operated by the operatorΞ.The low rank Hankel matrix whose component elements are observed to be recovered is recovered by matrix completion recovery algorithm,and the non-zero singular value is the minimum characteristic information of the data through the Singular Value Decomposition of the matrix.Then,we define the characteristic pattern group coding and pattern eigenvector mapping rules,use the fusion technology to fuse the heterogeneous minimum characteristic information,use the sampling value transmission protocol to encapsulate the message,generate the sampling minimum characteristic value light protocol data according to IEC61850-9-2 standard for transmission,and realize the communication network data standardization,high integration and sharing.By setting Virtual LAN Identifier to avoid network storm and congestion caused by large-scale broadcast transmission of data frames,and to save network resources,Dynamically Bandwidth Assignment algorithm is used to dynamically allocate network bandwidth,to ensure real-time transmission of high priority messages,reduce transmission delay and jitter,and to achieve reasonable network bandwidth for all kinds of message transmission.Finally,the paper introduces the transfer of light protocol data at the destination,which performs the opposite operation of generating light protocol data at the source.By unsealing the light protocol message,analyzing the characteristics,reconstructing or recovering the pattern components according to the pattern feature vector mapping rules and composition them to quickly and accurately recovery the original massive data.The accuracy and validity of the light transmission method for massive data in smart grid based on the principle of least characteristic information transmission are verified by an example.(4)Based on OPNET Network simulation platform,the communication network model of intelligent substation is constructed,the massive data network transmission simulation experiment is carried out,and the network transmission performance is analyzed and evaluated.Light transmission can effectively reduce the network transmission load and transmission delay.Then,the light data transmission physical prototype system built in the laboratory is compared with the traditional data transmission system for simulation test.The light transmission system can effectively reduce the network transmission traffic,and the transmission data compression ratio increases with the increase of analog sampling frequency.The results of network simulation experiment and physical prototype system dynamic simulation test verify the feasibility,reliability and superiority of the proposed light mechanism of smart grid massive data.At the end of the paper,the conclusion of this paper is summarized,and the future research work is prospected.