On the use of the multivariate autoregressive and relative power contribution models for determination of transient root cause

Abstract

The multivariate autoregressive (MAR) model and the relative power contribution (RPC) ratio are used in this work to determine the root causes of a power oscillation event and an apparent positive reactivity insertion transient occurred at the at the BWR/5 Units of the Laguna Verde Nuclear Power Plant (LVNPP) of México. The application of the MAR and PRC models leads to identify dominant frequencies and the contribution from other different signals to the dominant frequencies. The methodology firstly uses a linear model to estimate the response characteristics of the system and the spectra of the noise sources. The estimate of the process linear predictor is obtained by the ordinary least squares method. Then, the model performs a MAR analysis, and the RPC ratio is computed to determine the inter-relationship between the different reactor noise signals. The RPC ratio is an indication of how the fluctuation of one variable depends on other variables, at each frequency. Reactor signals acquired during the two transient operational events are used in the analysis. In the first event, a problem on the position controller of the flow control valve's stem induced the appearance of a power peak of 12% amplitude on the average power range monitors. Actual insertion of positive reactivity did not occur. The signals used for the analysis come from an average power range monitor, the position of the stem in the valve, controller of stem position, and controller of the recirculation flow. For the second transient, power oscillations of about 12% amplitude occurred. Signals from an average power range monitor, total flow through the core and flow through the 10 jet pump of each loop are analyzed. In both cases, some other signals were also used, but since they did not show appreciable influence on the RPC results, they were not considered for final analysis. The RPC results obtained in here confirmed previously known facts about the origin of the transients analyzed. Specifically, for the first transient, a dominant frequency of 1.7 Hz appeared on the power spectral densities of different signals from sensors on the recirculation loop B. At this 1.7 Hz frequency, the RPC ratio showed influence of such loop B spectra to the average reactor power spectrum, but no influence at all of the average reactor power to any of such loop B signals. The root cause, although not a real power transient event, therefore was not of neutronic nature, but related to recirculation flow. In the second transient, the prominent power oscillation frequency (0.54 Hz) was tracked within the spectral data of other signals. The RPC results for this case showed a strong influence of the average reactor power on the flow signals, but only a modest contribution from the flow signals to the average reactor power. The root cause therefore was of neutronic nature, due mainly to a combination of a particular core configuration and control rod pattern change at the moment of the event.