Research On Multi-objective Optimal Placement Of PMU In Smart Distribution Network Considering The Accuracy Of State Estimation

Smart distribution network is the trend of the development of distribution network,and it is an important component of smart grid.Due to development and integration of distributed generations and interaction between distribution network and consumers,the operation state of smart distribution network becomes more and more complicated.Original measurement systems of distribution network cannot depict the states of smart distribution network.To improve controllability and situation awareness ability of smart distribution network,phasor measurement units should be developed,and the state estimation of smart distribution network should be dealt on the basis of phasor measurement units.However,it is not economically and technically feasible to deploy phasor measurement units on a large scale.The research of phasor measurement units multi-objective optimal placement can improve not only the accuracy of state estimation,but the controllability of smart distribution network.What is more,it can give more feasible information for the planning and operation of smart distribution network.In this paper,a research on multi-objective optimal placement of phasor measurement units is carried on.Firstly,this thesis constructs the model of mixed measurements state estimation in smart distribution network,on the basis of weighted least square method.The adaptive Levenberg-Marquardt method is used to solve the nonlinear equation.Besides,a two-step state estimation model is constructed by taking the measurement from phasor measurement units as the redundant correction.The errors between the traditional model of mixed measurements state estimation and the two-step model are analyzed.Secondly,the multi-stage optimal placement of phasor measurement unit in smart distribution network is researched,in which the problem caused by voltage of buses exceed the limit is considered.The probability models of load and distributed generations are established.Specifically,the probability distribution functions are used to represent the random input variables,the correlation matrix is used to denote the correlation of input variables.What is more,quasi Monte-Carlo method is used to obtain the input samples,and the kernel density estimation method is used to fit the probability distribution functions of buses' voltage magnitude.The weights of buses are designed to identify the probability of buses' voltage magnitude exceed limits,and the model of multi-stage optimal placement of phasor measurement units is constructed which take the weights as the objective.The effectiveness of the model is analyzed from the aspects of correlation,zero-load buses and different method of multi-stage optimal placement.Thirdly,the optimal placement of phasor measurement units in smart distribution network which considering the accuracy of state estimation is studied.The cost of phasor measurement unit and the accuracy of state estimation for a variety of scenarios are considered as the objective.Besides,the model combines the measurements from phasor measurement units and the traditional measurement systems.Extended binary bat algorithm and extended adaptive binary differential evolution algorithm are used to obtain the optimal solution.The results show the superiority of the model and the algorithms are compared with the generic algorithm,the particle swarm optimization algorithm and the glowworm swarm optimization algorithm.Finally,the multi-objective optimal placement of phasor measurement units is studied.On the basis of the above-mentioned two chapters,the model of multi-objective optimal placement is constructed,which takes the cost,the accuracy of state estimation for a variety of scenarios and the sum of placed buses' weights as the objectives.Extended multi-objective binary bat algorithm and extended adaptive multi-objective binary differential evolution algorithm are used to obtain the Pareto frontier.The two algorithms are compared with the non-dominated sorting genetic algorithm with elite strategy,the multi-objective particle swarm optimization algorithm and the multi-objective glowworm swarm optimization algorithm.The results show the feasibility of the model and the algorithms.