Font Size: a A A

Research On Distribution Network Island Partition Considering Random Characteristics Of Distributed Resources And Load Fluctuation

Posted on:2017-06-08Degree:MasterType:Thesis
Country:ChinaCandidate:Z X LeiFull Text:PDF
GTID:2392330590468101Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
Islanding partition of distribution network including distributed resources is of great significance to improve distributed energy efficiency,and to increase the stability and economic operation of distribution network when the superior grid fails.Actual operation of island partition is separated into two steps,which are respectively off-line calculation and on-line judgement.Existing researches focus on-line judgement of islanding partition,lacking study of off-line calculation.From the perspective above,this thesis sets up an off-line calculation method of island partition in future periods.This off-line calculation method is based on a novel off-line judgement method and two highly-affected factors,which are random characteristics of distributed generation and load forecast.Main contents are as follows:1)Propose an island partitioning method based on voltage stability margin.The voltage stability margin index is based on Thevenin equivalent,and the method determines source nodes and breaking lines according to the index.After implementing power adjustment strategy by DG-cutting or load-shedding strategy,the whole method is complete.Taking the multi-microgrids structure in distribution network into account,this thesis takes the CIGRE MV standard model as the example.After establishing a multi-stage load situations,the example takes a traditional island partitioning method based on electrical distance as the comparison method,and compares the partition outcomes as well as indexes as voltage stability margin using two distinguished methods.Results verify that the method based on voltage stability margin possesses higher adaptability to load change,and is able to retain more load as well as keep voltage stability higher.2)Propose an AR-ANN short-term load forecasting algorithm and “from the whole to the node" multiple nodes load forecasting approaches.To carry out an AR-ANN forecast,ARMA should be executed first.Then replace part of training data in BP artificial neural network with ARMA predicting outcome along with other types of data in the same period.Finally,execute the BP artificial neural network load forecast."From the whole to the node" multiple nodes load forecasting approach rises from the practical problem of node layer data loss.After calculating the historical proportion of node layer load in the whole,this approach obtains the node level forecasting result with the help of the whole level load forecasting outcome.In order to apply AR-ANN,this thesis established three different samples based on three different forecasting ways: traditional single node load forecast,traditional multiple nodes load forecast and "from the whole to the node" multiple nodes load forecast.Examples verify the advantage of AR-ANN in forecasting accuracy,time consumption,etc.compared to BP artificial neural network.3)Propose the concept of day-ahead island partition,and apply it on CIGRE MV network.Those work is done based on the study above and the study of multiple wind turbines output modeling considering wind speed correlation.As to modeling part,this thesis establishes the model through a function between wind turbine output and wind speed.Moreover,the multiple Weibull distributed wind speed random variables' model is obtained by Latin hypercube sampling,which is also based on Copula theory and with the help of probability integral transformation.Finally,the outcome of AR-ANN "from the whole to the node" day-ahead multiple nodes load forecast and two correlated wind turbines output model are introduce to the CIGRE MV model.Apply modified island partitioning method based on voltage stability margin to the CIGRE MV model,and the day-ahead island partition is complete.Switching data and analysis data of distributed network in island mode can be extracted from the result above,which helps grid maintenance,dispatch and planning.Furthermore,sub-period result can contribute to dynamic planning and regulation of devices in power system.
Keywords/Search Tags:Islanding partition, voltage stability margin, Latin hypercube sampling, ARMA, artificial neural network, multi-node load forecast
PDF Full Text Request
Related items