Model And Algorithm Of The Optimal Island Partition Of Smart Grid Based On Graph Theory

Island operation(or independent sub-system operation)is a special operation mode for interconnected transmission power system and distribution power system integrated with distributed generation (DG). Many blackouts or even system collapse caused by a local disturbance would have been avoided if the interconnected transmission power system is safely and quickly split into several islands in time. For the distribution power system, its supply reliability can be not only improved through flexible and optimum island operation, but also the usage of DG can be taken fully advantage of. To focus on the deficit and difficulty existed in current study on the optimum island partition of power system, with the related theories of graph theory to be tools, the model and algorithm for optimum controlled separation surface searching of smart transmission system and optimum island partition of smart distribution system has been thoroughly studied in this thesis, and many satisfactory achievements have been obtained. The main achievements and research contents are listed in the following. First, a new graph theory problem--connected graph constrained knapsack problem (CGKP) is built, and an efficient approximation algorithm is proposed to solve the CGKP. In this thesis, four node sets related to graph connectivity are first introduced. The characteristics of the sets are studied, also the searching methods of the sets are proposed. Based on those study achievements on the characteristics and searching algorithms of the new node sets, an approximation algorithm for CGKP is presented by the extension of the approximation algorithm of graph constrained knapsack problem (GKP). Second, the mathematic model of the optimum controlled separation surface searchingproblem of interconnected transmission power system is constructed, and a two-stage strategy of"searching and regulation"is designed to solve this problem. The model is decomposed into an optimum balance graph partition sub-problem and a regulation sub-problem based on optimal power flow. These two sub-problems are successively solved by the CGKP approximation algorithm and optimal regulation measure based on the solution of optimal power flow to acquire a final controlled separation surface. Third, electrical distance constraint is proposed to improve the CGKP-based method of the optimum balance graph partition sub-problem. The load nodes which are closer to a specific coherence generator group are preferentially allocated to the island containing the specific coherence generator group by the improved method, so that the searching space of the CGKP approximation algorithm is reduced evidently. At the same time, the configuration of the sub-system obtained is also more reasonable. Fourth, an optimum controlled partition method based on the alternate optimization of the master problem and sub-problem is proposed. The optimum balance graph partition sub-problem and optimal power flow sub-problem which are successively solved in the"searching and regulation"strategy are set to be master problem and sub-problem, respectively. The result fed by the sub-problem is used to adjust the parameters and solving process of the master problem. Therefore, a more optimal controlled separation scheme will be obtained by repeating the alternate optimization of the master problem and sub-problem for several times.Fifth, in this thesis, a novel optimum island partition model is presented for the distribution system integrated with DG, and the"search and regulation"strategy is applied to solve this problem. Initial optimum island partition scheme is gained through island partition procedures including multiple tree knapsack problems (TKP) and island combination procedures based on dynamic programming algorithm and branch and bound algorithm. The final island partition scheme is obtained after feasibility checking and adjustment. Furthermore, through the comparison of the optimum island partition strategies based on dynamic programming algorithm and branch and bound algorithm, application scopes of those two methods are shown. Last, a new reliability evaluation method using the Monte Carlo time sequential simulation for the distribution system with DG including wind power and photovoltaic power generation system and etc. is presented based on the simplified equivalent network The zone element failure effect table is obtained only at the simplified network; then, the impact of the DGs'stochastic power output, running/failure statuses of the devices and the stochastic capacity of the loads on the distribution system reliability are analyzed and calculated by the Monte Carlo time sequential simulation. The results of samples demonstrate the validity of the method proposed, fourthly, it also strongly proofs the practical significance of the optimum island study on the smart distribution system.