Intelligent Topology Identification And Voltage Management Optimization Of Low Voltage Distribution Network Based On Data Mining

On the background of "Internet + Smart Energy" as the energy development strategy and "carbon neutrality" as the energy development goal,a growing number of distributed generation and charging pile access to low-voltage distribution network(LVDN).Intermittent voltage overrun will become one of the main challenges of LVDN operation safety,and the supporting basis of LVDN voltage management is intelligent topology recognition.Hence,the intelligent topology identification and power quality management of LVDN have become the key technological requirements for the power grid industry to promote the connection of distributed generation,energy storage,electric vehicles and other equipment to the grid,and build a high-quality modern power supply service system.Based on the measurement data of power consumption in LVDN,this paper focuses on the identification of the "transformer-phase-line-household" relationship and voltage quality management of LVDN.The details are as follows:(1)For identification of LVDN's transformer-household relationship,a knowledge-driven recognition method is proposed to address the shortcomings of the existing researches on data discrimination and noise robustness.Firstly,based on the power flow model,the prior knowledge related to the relationship between transformer and households are deduced;Then,the Z-score and principal component analysis are combined to standardize and extract the original data to amplify the differences between the data and reduce the impact of data noise;Moreover,a knowledge-driven transformer-household identification model which comprehensively utilizes the voltage correlation characteristics of users close to and not close to the transformer distribution is proposed.Case studies show the effectiveness of the proposed method.The performance of the proposed method in different scenarios is discussed.Compared with the existing methods,the proposed method can achieve higher identification accuracy,and has a better robustness to data discrimination and noise.(2)Aiming at the problem of the effect of insufficient consideration of the diversity of user electricity features on the recognition performance in LVDN phase-line-household relationship identification,an identification model considering adaptive clustering of vacant users is proposed.Firstly,based on the temporal and spatial distribution of node voltages,the correlation characteristics among users are deduced,and an adaptive user clustering method which can associate vacant meters with electricity meters is proposed.Then,the quadratic programming model for identification is established with the goal of minimizing the difference between the sum of the current of the low-voltage bus and the connected meter.Further,the probability distribution model of recognition results based on Monte Carlo is designed to improve the robustness of the recognition algorithm to the metering error of smart meters.Case studies show the effectiveness of the proposed method,and the influence of threshold coefficient on the identification accuracy is discussed.In the diversified scenarios with vacant users and smart meter measurement errors,the proposed model can achieve higher identification accuracy.(3)Aiming at the problem of the significant influence of data quality on the recognition performance in LVDN phase-household relationship recognition,a multi-dimensional correction method for phase identification considering data imperfectionan is proposed.Firstly,based on the voltage characteristics of LVDN nodes,the multi-dimensional correlation characteristics between low-voltage users and buses are deduced.Then,combining with the positioning index and the multi-dimensional correlation characteristics,a multi-dimensional correction theory that can correct the phase connectivity of users far away from the low-voltage buses is described.Moreover,by integrating a variety of phase sequence recognition theories,a diversified phase identification model is proposed to deal with data incompleteness.Case studies show the effectiveness of the proposed method,and the influence of the threshold coefficient on the recognition accuracy is discussed.In the case of incomplete data,the improved model proposed in this paper can achieve higher recognition accuracy than the existing methods.(4)For the problems of insufficient means in the treatment of voltage magnitude and unbalance limit violation in the LVDN,the idea of fully mobilizing the resource potential of resident demand is put forward,and a LVDN voltage management model based on multi-objective optimization of thermostatically controllable appliances(TCA)is proposed.Firstly,the response models of TCAs in LVDN voltage management are established.Then,a multi-objective optimization model of LVDN voltage management is established,which jointly minimizes the residential demand response cost,network loss and user discomfort.Of which,the power consumption of electric water heater and air conditioner are taken as the decision variable.Case studies show that the proposed method can take account of the adjustment cost of grid companies and user comfort,effectively control voltage amplitude and imbalance of LVDN within the qualified range,and improve the capability of LVDN to integrate PV.The research results of this paper enrich the theories and methods of intelligent LVDN topology identification and voltage quality management,which is of great significance for promoting the construction of China's "building a new power system with new energy as the main body" and the improvement of users' electricity experience.