Información de la Tesis Doctoral

This dissertation addresses the problem of designing robust and efficient underfrequency load-shedding (UFLS) schemes of small isolated power systems. UFLS schemes are a well-established remedy against frequency instabilities.

Frequency stability is a major issue for small isolated power systems. In small isolated power systems, frequency instability occurs in the form of continuous frequency decay or of sustained frequency swings leading to tripping of generating units. Main reasons for the continuous frequency decay are insufficient spinning reserve or the loss of a major generating unit. However, the ultimate reason for frequency instability comes from the fact that generating units are equipped with underfrequency protection devices which trip the generating units in case of underfrequency conditions. Thus, if additional counter-measures, such as UFLS, fail, underfrequency protection devices trip the generating units and the power system collapses due to frequency instability. In addition, increasing penetration of decoupled power generation enhances the risk of frequency instability due to its negligible frequency-control capabilities and the absence of inertial response. Decoupled power generation refers to those solar and wind-power generating units which are mechanically decoupled form the power system through a stage of power electronic converters.

UFLS schemes are implemented to avoid system collapses. These schemes trip loads to restore the power equilibrium and to arrest frequency decay. UFLS schemes can be grouped in conventional and advanced schemes. This dissertation is concerned with conventional UFLS schemes which trip predefined amounts of load at specified frequency thresholds. The instant of tripping is determined by underfrequency and rate of change of frequency (ROCOF) relays, monitoring continuously frequency and ROCOF and comparing them to underfrequency and ROCOF thresholds.

The design of such UFLS schemes is usually based on the evaluation of the responses of the power system to a set of contingencies. These evaluations in turn are realized by means of simulations of a model of the power system of interest. A simple but still accurate model has been developed to simulate frequency dynamics of a power system. This so-called non-linear multi-generator system frequency dynamics (SFD) model is a low-order model which reduces the computational cost with respect to a fully detailed model maintaining the essential frequency dynamics of the original model. The ability of the non-linear multi-generator SFD model to accurately reproduce short-term frequency dynamics has been checked by comparing the SFD model with more detailed power-system models.

Several methods have been reported in literature to design conventional UFLS schemes, but there exists no generally accepted method for the design of UFLS schemes. These methods are mostly based on simulations of a set of operating and contingency scenarios. However, these operating and contingency scenarios are usually determined by the common practice of scenario selection, which does not necessarily guarantee the selection of the most appropriate operating and contingency scenarios. Furthermore, some design methods do not guarantee an optimal UFLS scheme performance in terms of shed load. Thus, a method for the design of robust and efficient UFLS schemes of small isolated power system has been elaborated. Efficiency is tantamount to shedding a minimum amount of load, whereas robustness denotes efficiency for a set of operating and contingency scenarios. This proposed method for the design of robust and efficient UFLS schemes is based upon the SFD model approach.

To guarantee a robust UFLS scheme design, adequate operating and contingency scenarios need to be selected. A method based on Data Mining has been proposed to identify and select representative operating and contingency scenarios. Several Data Mining techniques such as K-Means, Fuzzy C-Means and KSOM have been applied. Efficiency of the UFLS scheme can be achieved by applying an optimization algorithm to adjust the parameters of the UFLS scheme, corresponding with the tunable parameters of the underfrequency and ROCOF relays. The main objective of the optimization stage is to minimize the amount of shed load without jeopardizing the system stability. Adaptive Simulated Annealing algorithm has been suggested as optimization algorithm.

The proposed method for the design of efficient and robust UFLS schemes comprises thus two tasks: the selection of adequate operating and contingency scenarios by means of Data Mining techniques and the optimal tuning of the UFLS parameters by means of adaptive Simulated Annealing optimization algorithm. To confirm the feasibility of the approach, the proposed method has been applied to the design of the UFLS schemes of the La Palma and the Gran Canaria power systems. In addition, the proposed UFLS scheme has been compared to a simply optimized UFLS scheme which uses operating and contingency scenarios determined by the common practice of scenario selection.

Finally, methods for the design of UFLS schemes are in general based upon certain assumptions concerning design conditions and power-system behavior and in particular, variability of UFLS scheme step sizes. Usually, implemented step sizes do not correspond with actual step sizes due to changing operating conditions. Thus, the impact of varying design conditions and varying step sizes has also been extensively studied.

The analysis of varying design conditions is addressed by designing different UFLS schemes resulting from a variation of one of the design conditions. These design conditions can be roughly classified into those affecting the selection of operating and contingency scenarios and those affecting the formulation of the optimization problem, i.e. the objective function, the constraints, the decision variables or the optimization algorithm. Some of these modified design conditions coincide with other design methods proposed in the literature.

In order to address step-size variations and step-size reduction in particular, a method for the design of UFLS backup steps has been proposed which first designs several combinations of different backup steps settings heuristically, based on simplified variations of UFLS scheme step sizes, and in a second step, the proposed designs are validated using a Monte-Carlo simulation approach. Monte-Carlo simulations make use of a probability density function, modeling step-size variation as combination of feeder-load variation and feeder outages and breaker failures. The necessity of backup steps has been reaffirmed by the Monte-Carlo simulation approach

**Descriptores: **Transmisión y Distribución

*Universidad Pontificia Comillas. Madrid (España)*

17 de diciembre de 2010