Abstract
A method for evaluating a predictor-based alarm system is studied in this paper. The predictor is composed of a deterministic component reflecting external information and a statistically based component for the error between the measurements and the external predictor. The aim of the predictor study is twofold: it is a means of interpreting the connections between the alarm and the catastrophe, and it can be used to select suitable alarm levels. As an application, the performance of a water-level predictor as part of a flood warning system has been evaluated. The result of this analysis shows that an alarm system which operates when the predictor reaches a certain level will tend to give either too many alarms or alarms that are out of phase with the catastrophe.''
| Original language | English |
|---|---|
| Pages (from-to) | 1-18 |
| Journal | Journal of Time Series Analysis |
| Volume | 11 |
| Issue number | 1 |
| DOIs | |
| Publication status | Published - 1990 |
Subject classification (UKÄ)
- Probability Theory and Statistics
Free keywords
- flood warning system
- modelling
- alarm system
- Prediction
- performance evaluation
- level crossing
- extremes.