The importance of the wide-area monitoring system is rapidly increasing and many transmission system operators have already made several steps toward implementing wide-area monitoring, protection, and control. Underfrequency load shedding (UFLS) is undoubtedly one of areas that can be significantly improved within such an upgrade. However, in order to pursue finding a concept for practical implementation, as little measurements as possible should be used for the purpose. In this paper, an approach for obtaining a few-seconds-in-advance frequency prediction is described. It enables making decisions about the amount of load to be shed and, at the same time, being aware that inaccuracies in the procedure are inevitable. An extensive testing of the approach was performed on the 110 kV part of the Slovenian power-system model and the predictive UFLS appears to be highly superior to traditional UFLS on evoked frequency conditions as well as the total amount of disconnected loads. Consideration of practical limitations, such as all major time delays, unveiled a limitation for using the procedure at extreme values of system-wide frequency gradients. Consequently, the proposed algorithm is to be considered a “smart” supplement to already operational traditional UFLS.

The actual activation of underfrequency load shedding is rarely needed in a power system. However, this does not diminish its significance, as it is the last barrier before a system blackout if underfrequency conditions occur. Like in all other areas of application, a tendency also exists to maximize the efficiency of the underfrequency load-shedding scheme and to bring its operation to the limits. By default, the scheme should be able to stop a frequency fall under any circumstances. However, trying to lower the quantity of the disconnected load in the same process is the purpose of this paper. After the theoretical background is summarized, in terms of the reaction of individual generators in the islanded system to a certain power imbalance, a procedure is described showing how it is possible to adjust any of the adaptive load-shedding schemes to utilize as much primary frequency control as possible. As a reference, a theoretically optimal underfrequency load-shedding scheme is presented.

Underfrequency load shedding is one of the most important protection systems, as in many cases it represents the last chance to prevent a system blackout after a serious disturbance occurs in a power system. In order to improve traditional schemes, many efforts have been concentrated on attempts to use the frequency gradient as an indicator for determining the lack of active power in a system. This paper concentrates on analyzing the factors that influence the gradient. Analyses have shown that the gradient can give misleading information about the active-power deficit, as long as certain factors are ignored or assumed to be constant. A draft of a gradient-based underfrequency load-shedding scheme is presented, which also highlights the problems associated with the use of gradient. On the one hand, a frequency gradient alone does not appear to be sufficient for the active-power deficit estimation. Nevertheless, for the actual load-shedding procedure it is found to be very useful.