Online Application of Local OOS Protection and Graph Theory for Controlled Islanding

Despite layers of protection at work, power systems have experienced major blackouts in the past due to cascading outages. Such outages can be prevented with controlled islanding, which consists of two problems: “when” and “where” to island. The majority of the existing offline and real-time methods for the “when” aspects are based on wide area measurement systems (WAMSs), which depend on the system topology to perform periodic offline simulations or real-time computation of dynamic system equivalents. In this paper, a simplified “when” approach is proposed based on local generator out-of-step (OOS) protection and generator coherency using fault location. The novelty of the work lies in the use of local OOS protection relays to identify the overall system instability, which eliminates the need for WAMS communication and the complex computation required in real-time methods to form dynamic system equivalents. As this approach does not require any equivalent networks, the methodology is independent of system topology. Unlike existing WAMS-based methods, the proposed methodology simplifies the communication between the central “when” unit and each generator's OOS protection relay using status flags communicated with the IEC 61850 R-GOOSE protocol. The proposed “when” methodology is combined with the “where” method based on graph theory to test the overall controlled islanding scheme in the IEEE 39-bus system using the real-time digital simulator (RTDS).

I. Introduction

Different incidents in recent history can be recalled in which cascading failures resulted in wide area blackouts in power systems [1]. Researchers have proposed controlled islanding [2] as a last resort to prevent these blackouts, splitting the network into intentional islands before the cascading outage progresses through the system. The two major problems involved are identifying the condition when controlled islanding action is required and finding the possible locations where the system can be split for a clearly unstable case [3], [4]. These problems are termed the “when to island?” and “where to island?” problems of controlled islanding. This paper focuses on the “when” aspect, but also implements the “where” aspect to develop an overall controlled islanding scheme. The “when” aspect of islanding must predict system-wide instability and initiate controlled islanding procedures before it is late for any remedial action. Detection of an unstable power swing, also referred as an out-of-step (OOS) condition, plays a key role in terms of the “when” aspect [5]. Different methods exist in the literature to predict the OOS condition using either local or WAMS measurements. The majority of the existing methods in the literature for determining “when to island” are based on synchrophasor-based WAMS measurements as the local OOS methods face significant challenges with respect to identifying instability of the overall system. The main challenge is the lack of information about the rest of the system at the local level. With local measurements, however, the communication bandwidth and reliability could be improved along with simpler implementation.