A Next Generation Alarm Processing Algorithm Incorporating Recommendations and Decisions on Wide Area Control
Abstract
The number of alarms for a typical power systemevent may be overwhelming to power system operators and maydelay the operator from taking appropriate corrective action.Worldwide, a number of alarm processing techniques are used toreduce the number of alarms that the operator ‘sees’, so as tobetter comprehend the situation at hand and make accurate decisionsfaster. This paper proposes an alarm processing algorithmthat goes beyond the prioritization of alarms. The conceptualalgorithm proposed has additional features that offer the operatorrecommendations and decisions for the event that has causedthe alarms, as well as a feature that may execute controls if theevent is non critical. In this sense, the proposal is a bridge fromwide area measurement to wide area control systems.Index Terms— alarm processing, power system control, reactivepower control, sensors, switched capacitors, transformer tapchanging, wide area control systems, wide area measurementsystems.
I. INTRODUCTION
POWER system operators may encounter an overwhelmingnumber of alarms due to system-wide events in their areaof responsibility. These alarms can be a source of confusionfor the operator, especially if the system events that havecaused the alarms require immediate action. For example, atransformer fault can possibly have 150 alarm messages thatare displayed in only two seconds [1]. The need for the developmentof alarm processing and prioritization algorithms hasbeen recognized since the 1960s [2]. This need is even morepressing today due to the increased size and complexity ofpower systems, the interconnections between large geographicalareas, and the deregulation of power systems which intro duces complexity in the control decisions. As an example,under deregulation, transmission circuits may be operated atrelatively high power levels more frequently in order to fullyimplement power marketing. These high levels of powertransfer generally suggest that high speed and accurate operatoractions are imperative. Further, there is a trend of performingwide area control and monitoring activities in a centralizedfashion by having a small number of control centers [3].A number of alarm processors have been developed and arein use today. Some examples include [4] and [5]. Alarm processorssuch as these provide a number of tools including alarmsuppression, alarm prioritization, reduction of alarms on thebasis of area of responsibility, and control recommendationsto the operator. For example, De Souza et al. [6] present amethodology that combines the use of artificial neural networks(ANNs) with fuzzy logic to form alarm processingidentification of faulted components. Inputs to the ANNs arealarm patterns while fuzzy relations are established to form adatabase employed to train the ANNs. Each neuron of theANN is responsible for estimating the degree of membershipof a specific system component into the class of faulted components.Results of an application on an actual system showthat the proposed method allows for good interpretation of theresults, even in the presence of corrupted alarm patterns.Lin, Lin, and Sun [7] present fault detection and alarmprocessing in a loop system using a fault detection system(FDS). The FDS has an adaptive architecture with a probabilisticneural network (PNN). The training of the PNN uses theprimary/backup information of protective devices to create thetraining sets. When the topology changes, the PNN can beretained and estimated effectively. Computer simulations wereconducted to show the effectiveness of the proposed system.The basis of neural network solutions to alarm processingappears to be the recognition of events seen in a training set.In many cases, this approach is satisfactory when the trainingset spans a large range of operating scenarios.


Download full report
http://ieeexplore.ieeeiel5/4275198/42751...er=4275419