Distributed Generation Planning In Distribution Network Based On Hybrid Intelligent Algorithm By SVM-MOPSO

While distributed generation(DG) such as distributed wind power generation,photovoltaic, etc., have the advantages to save energy and reduce environmental pollution andnetwork losses, improve power quality and power supply reliability and so on, but due to therole of climatic and environmental factors, the randomness, intermittent and volatility of DGpower output is vary significant. DG planning is one of the most key technologies in smartdistribution grid construction. Load flow will be change if a large number of DG connected todistribution network, the change will affect the whole normal operation of grid, degree isclosely related to the position and capacity of DG. Therefore, how to scientifically siting andsizing is one of the major hot spots which DG planning phase will face, is an importantmeasure to promote energy conservation at home and abroad, have an important significanceto ensure safe and stable, economical and reliable operation for distribution network. Thispaper focuses on the above questions to research as the following work.Firstly, this paper introduces the research background, summarizes the research statusand limitations at home and abroad for DG planning, introduces an equivalent circuit modeland flow calculation approach, and establishs a probability model of DG. Taking into accountthe characteristics of wind and solar power as the representative of distributed intermittentrenewable energy sources significantly influenced by the external environment, this paperuses probability theory based on two-dimensional continuous random variable jointprobability distribution theory to establish joint probability distribution function whichconsidering the timing characteristics of wind and solar, it will accurately reflect therandomness of two DG, and significantly reflects the random fluctuation of random variablesimpact on the power system. These works are foreshadowing for the later DG planning.Secondly, this paper introduces the basic theory of chance constrained programmingmodel and multi-objective chance constrained programming model, establishs a mathematicalmodel of DG programming based on the above theory, the model contains the objectivefunction and constraints, while the objective function have three sub-goals includeingeconomy, power quality, environmental protection.Thirdly, the paper introduces the basic theory of the stochastic simulation techniques,support vector machines (SVM), particle swarm optimization and multi-objective particleswarm optimization (MOPSO), uses random number generator generates the random numberssamples which obey to wind and solar joint probability distribution, applies input and outputdata generated by random simulation technology to be as training samples to obtain the regression models of SVM to approximate the uncertain function, and use multi-objectiveparticle swarm algorithm which based on crowding distance selection strategy to solve themulti-objective problem. For the chance constrained programming model, combined with thestochastic simulation technique, SVM technology and multi-objective particle swarmalgorithm based on crowding distance selection strategy, this paper proposes a hybridintelligent algorithm based on SVM-MOPSO for siting and sizing in DG planning, and givesthe detail algorithm steps and flowcharts. It also describes the ideal point method optimalselection strategy to obtained optimal choice from planning results.Finally, this paper uses Homer software to simulate and analyse the wind speed and lightintensity in an area of Guangdong Province based on wind speed, temperature, light intensityand average monthly weather datum, apply IEEE-33-bus distribution system to simulate andtest for the new algorithm. Simulation results show that the planning approach which takinginto account the stochastic nature, timing characteristics and connected-grid probabilitydistribution of DG, can improve the efficiency of algorithm, prove the reasonable andeffective of this algorithm, and the introduction of Pareto front give decision makers full placeto select, and has more engineered.