Research On Distribution Network Planning Based On Big Data Technology

With the development of economics and the improvement of people's living standard,the main contradiction of the current electricity market has been changed into the contradiction between increasing demand of power supply capacity and quality from the users and low power supply capacity and quality in reality.The distribution network is an important part of the power system,and the level of its planning directly affects the capability and quality of the power supply.The application of big data in power system to distribution network planning is one of the ideas to improve the level of distribution network planning.In order to improve the fineness and efficiency of distribution network planning,targeting at the data-intensive problem in analytical processes of heterogeneous data which is from multiple sources and in the planning model,research on distribution network planning based on big data processing and applications is carried out in this paper,and the main contents are as follows:(1)The characteristic and processing method of big data which is related to distribution network planning is sorted out and analyzed.Also,the MapReduce expression of text structuring,anomaly data mining and data normalization is analyzed,and the adverbs of degree in power system texts which are weak structured are scaled by classification in linguistics to structure texts.This text structuring method is partly based on semanteme.And these preprocessing processes can be applied on information feature extraction and mining in subsequent research.(2)A method for precisely predicting the spatial distribution of saturated load based on deep learning and multi-source information fusion is proposed.Firstly,the high dimension features in load cells are acquired,and the unstructured features in these cells are structured.Then Stacked Denoising AutoEncoder(SDAE)is applied to learn characteristics of the features.Finally,the spatial saturated load density of the area to be planned is predicted by the regression unit on the top layer of SDAE.The simulation results show that the prediction accuracy of the proposed method is higher than the prediction accuracy of feature extracting and predicting methods such as the method based on least squares support vector machine(LS-SVM)and the method based on principal component analysis(PCA)and BP neural network.(3)A method for substation's pre-locating based on deep learning algorithm and substation-line joint planning based on parallel computing is proposed.Firstly,convolutional neural network is applied to learn features related to the principles of substation's locating to carry out substation's pre-locating,and the result of which can be alternative sites for the follow-up planning processes.Then based on the results of substation's pre-locating,the substation-line bi-layer planning model is established,and the improved bird swarm algorithm based on the multi-processor master-slave parallel computing structure is applied to solve the planning model.The simulation results show that the judgment accuracy of the proposed substation pre-locating method is higher than that of substation pre-locating method based on shallow learning,and a reasonable planning scheme can be obtained through the simulation of the established planning model,also,the parallel computing structure can effectively improve the speed of the planning process when the structure is in a reasonable configuration.