Research On Restoration Strategy Of Power Distribution Network Considering Distributed Generation And Controllable Load

With the development of the automation level of smart distribution network, the ability of load management for the power system is improved. In this case, more and more controllable loads get into the distribution network, which changes the restoration model. So it is significant to research how to improve the reliability of power supply through switching the controllable load reasonably and optimizing restoration program. For the long-term fault restoration, static restoration is not only unrealistic but also easy to overload since load and power output always change constantly during the fault. Therefore, it is important to formulate the dynamic restoration strategy with the power output and load uncertainty considerd in order to guarantee the safe and reliable operation of the distribution network. To solve these problems, this paper has done the following work:Firstly, the graph theory is applied to simplify the distribution network, and the adjacency matrix was used to represent the relationship between load nodes. Based on this, the back/forward sweep algorithm was introduced though reasonable equivalent of distributed generations, and the nearest neighbor clustering and Elman neural network methods were analysised.Secondly, for the short-term restoration, a method of combinating the equal possible path was designed to restore the loads. Besides, the relative fitness of loads and degree of risk index were proposed to optimize the restoration procedure. The new problem for distribution network with taking account of the degree of importance and controllable load can be solved by the method and indexes effectively, For the independece and the relationship of the DGs and reserve tie-lines during the process of optimization, the multi-agent system including information layer, executive layer,equipment layer and coordination layer was introduced to make full use of the information to parallel compute and optimize scheme.Finally, for a long-term fault restoration, the uncertainties of photovoltaic power output and load were considered. Light intensity was assumed to approximately obey the Beta distribution in a period of time. The power output was determined by analysising the light intensity data in different cloud intervals. For the load uncertainty problems, the nearest neighbor clustering method was used to process the load forecast samples to find out the similar samples from the historical data. Then the method of Elman neural network was used to forecast the load power. The distribution network dynamic restoration with the photovoltaic power and energe storage device was realized through the method of combinating the equal possible paths.