Research On Comprehensive-benefit Evaluation Method Of Distribution Network Planning

With the gradual expansion of new energy access and the continuous improvement of intelligence in the distribution network,the complexity of the distribution system is also becoming increasingly apparent.The planning of the distribution network has gradually developed towards the direction of systematization,streamline,conservation,greenization and integration.Comprehensive evaluation of the planning scheme is an important measure to ensure the effective realization of the planning objectives.Therefore,the following research work is carried out for the evaluation method and evaluation index system of distribution network planning:(1)Systematically sorting out the research status of domestic and foreign on evaluation indicators of distribution network planning.A series of indicator systems adapted to the technical and economic evaluation of distribution systems have been proposed based on the basic characteristics and development trend of distribution network in China.Taking into account the rationality of the grid structure,security,reliability,national economy,financial benefits,environmental emissions,energy-saving and low-carbon properties in distribution network,we quantified the planning project and proposed a series of new indicators,such as “PB”,“DB”,“CS”,“ERSI”,“IFSI”,“SCSEI”.(2)For the problem of rapid assessment in distribution network reliability,this paper introduces a machine learning algorithm to improve the traditional Monte Carlo simulation.We adopts pattern recognition method to judge the running status of distribution network.Compared with the traditional MCS simulation which using the optimal power flow to identify the operating status of the distribution network,this method greatly improves the computational efficiency.In addition,for the problem of false recognitions called “false positive judgment” and “missing judgment”,we analyzed its influence on the error and convergence of the reliability index and prove that it should try to avoid “misjudgment” samples to reduce the integration error.Introduce and improve the Adaboost.M1 algorithm framework to integrate traditional single machine learning algorithms,it due to that the improved classifier minimizes “misjudgment” samples while ensuring classification accuracy.Nataf transformation is used to process the random variables(distributed power output and load fluctuation)that have correlation in the distribution network,so that we can consider the influence of the correlation on the reliability in the distribution network.Through the simulation result proved that the method saves the iterative time compared to the classical MCS and also ensures the accuracy of the reliability index.(3)To propose a weight system model that can reflect objective information of the indicators and subjective opinions of experts.Using the maximum deviation to reflect the difference of the indicators between different evaluation plans.Using information entropy to quantify the amount of decision information from the sub-indicator layer to the upper-level.And then,propose an objective function constructed by information entropy and maximum deviation.The subjective constrains mainly include the opinions of the experts on the relative importance of index weights and the constraints on the fluctuation range of weight values.Using Binary contrast method to calculate subjective weight and intuitionistic fuzzy set to calculate objective weight.At last,Forming a comprehensive decision model that includes subjective and objective information.