Research On Data Driven Intelligent Identification Method For Private Capacity Expansion Of Distribution Transformer

Private capacity increase of distribution transformer not only infringes the interests of power enterprises,but also affects the safe and stable operation of power grid.At present,the inspectors of power enterprises only judge whether the user’s private capacity increase behavior is normal by comparing the user’s electricity consumption.The detection efficiency and accuracy of this method are low.After analyzing and summarizing the current status of transformer capacity detection technology and data-driven anti-theft technology,this subject has conducted a comprehensive and in-depth study on the intelligent identification of private capacity increase of distribution transformer based on data-driven.Based on the analysis of transformer structure and operation mechanism,a dimensionality reduction mapping method for transformer operation data is derived by combining the transformer basic balance equation.On this basis,considering the different application fields,the intelligent inspection method of private capacity increase of distribution change based on density clustering combined with limit learning machine algorithm(DBSCAN-ELM)and the intelligent monitoring method of private capacity increase of distribution change based on short and long term memory network nested fuzzy C-means clustering(LSTM-FCM)are proposed respectively.Finally,the effectiveness and superiority of the proposed method are verified by the transformer operation data.The main work of this project is reflected in the following aspects:(1)A dimension-reducing mapping processing method for transformer operation data is proposed,and the dimension-reducing formula and derivation steps for single-phase transformer and three-phase transformer are given respectively.On this basis,the distribution characteristics of dimension-reducing data on the two-dimensional plane are studied in depth,providing theoretical support for the data cleaning method proposed in the thesis later;(2)Based on the distribution characteristics of dimensionality reduction data,a capacity-increasing intelligent troubleshooting method based on DBSCAN-ELM is proposed to solve the on-site troubleshooting problem of private capacity-increasing behavior in multi-users under the background of big data.The specific derivation process of the method and the building principle of the model are given respectively.On this basis,combining with the time characteristics of private capacity-increasing,a capacity-increasing troubleshooting criterion considering time continuity is proposed;(3)On the basis of DBSCAN data cleaning,the abnormal data correction method is deduced to restore the original time continuity of the data,and a capacity-increasing intelligent monitoring method based on LSTM-FCM is further proposed to solve the real-time monitoring problem when users have private capacity-increasing behavior in the future.The specific derivation process of the proposed method and the building principle of the model are given separately,and the method flow and capacity-increasing monitoring criteria are summarized;(4)Using the transformer operation data,the comparative experiments of the data dimension-reduction mapping method,the capacity-increasing intelligent troubleshooting method based on DBSCAN-ELM,the abnormal data correction method and the capacity-increasing intelligent monitoring method based on LSTM-FCM are carried out,which verify the effectiveness and superiority of the proposed method.The research results of this thesis provide a new inspection method for the field of power consumption behavior inspection of power enterprises,and provide a certain theoretical and technical reference for the private capacity increase inspection of distribution transformer.