In light of problems such as the untimely condition monitoring and alarm, excessively large threshold bandwidth and inaccurate condition evaluation parameters of intelligent ship power system equipment, an adaptive threshold method is proposed to monitor, alarm and evaluate the conditions of such equipment.

First, a simulated annealing algorithm is used to optimize the support vector regression (SVR) machine prediction model to simulate the general state characteristic parameters of the power system equipment. Then, after the normal transformation of the modeling residual, combined with the sliding time window, the adaptive threshold model is constructed. Finally, the exhaust gas temperature of the ship's main propulsion diesel engine is selected as the research object for example verification.

The results show that compared with the traditional fixed threshold, the adaptive threshold model has more compact bandwidth and good adaptability, and can identify abnormal phenomena in power system equipment in advance.

This method improves the efficiency and threshold accuracy of monitoring and alarm systems, and provides an effective means of early fault diagnosis and a more accurate basis for system status evaluation.