Abstract：

　　Faced with increasingly complex chemical process plants, improving the performance of chemical process alarm systems is important. The traditional chemical process parameters alarm threshold setting method generally considers only false positives, but not taking both false positives and false negatives into account, leading to a lot of false alarms in alarm systems. To solve these problems, we used the alarm threshold optimization method based on an adaptive composite indicator. We used the kernel density estimation method to estimate the state of the process alarm based on historical data, integrating the false positives rate and false negatives rate to establish an objective function for optimal alarm thresholds. The numerical optimization algorithm was embedded in a particle swarm optimization algorithm, forming a new algorithm to solve the function. In case, this method was applied to the TE process. The results showed a false positive rate of 0, and a false negative rate of 0.78%. Compared with the traditional 3σ method, this method can effectively reduce the rate of false positives with a low false negative rate, and improve the performance of the chemical process alarm system. This will reduce stress on site operators, as well as the risks of loss of life and property.

Key words：adaptive composite-indicator false positives false negatives kernel density estimation particle swarm optimization algorithm

Received： 15 November 2016

Corresponding Authors：

Cite this article：LUO Jing,HU Jinqiu. A study of adaptive composite-indicator alarm threshold optimization of chemical process parameters [J]. 石油科学通报, 2016, 1(3): 407-416.

URL：