Study On High-Speed Method For Sloving Large-Scale Security Constrained Optimal Power Flow

Security constrained optimal power flow (SCOPF) is one of the important tools to analyze the security and the economy of power system. With the SCOPF we can ensure the safe operation of power grid persistently. As it take into account the contingencies and pre-contingency configurations together in the perspective of mathematics, we can deal with the unexpected accidents of the power grid effectively in theory. But actual system often need to consider so many contingencies, which let the traditional SCOPF methods be limited for the realistic systems due to the high memory consumption and low computation efficiency. This paper proposes a high speed multi-core parallel algorithm for overcoming these disadvantages.We use the optimal power flow (OPF) tool under normal operation to check all of the contingencies and determine which of those uncontrolled at first, and then the uncontrollable contingencies are used to correct the normal operation status.Form the two process alternates until all state are satisfied we can get the optimal solution.We call this basic method.As the power flow (PF) tool has higher computational efficiency than the OPF tool, we use a PF tool to check the contingencies before the OPF tool. We call this kind of modified method for improvement method1.With the contingency filtering technology, we get the dominant contingencies from the calculation result of the PF tool. Only the dominant contingencies should be checked by the OPF tool, so there are less contingencies which need been checked.We call this improvement method2.Analyzing the characteristics of the modular structure of the improvement method2. Multi-core parallel technology is used to speed up the calculation, then we get improvement method3, a high-speed multi-core parallel algorithm.The proposed method has been successfully tested on three systems, which include a realistic provincial3301-bus system with693contingencies. The numerical results indicated that the proposed method is suitable for solving large SCOPF problems because of fast and steady computation time. Furthermore, the memory consumptions are insensitive to the size of the system compared with the traditional SCOPF methods. That is, the memory consumptions of the proposed method are1%less, and the computation speeds are three orders of magnitude faster than that of the traditional ones as the system scale enlarges.