Research on Anomaly Detection Method for Feed Water Pumps in Gas Power Plants Based on Similarity Principle
Equipment operation monitoring is an important work task for gas power plants.It is crucial to identify the possible abnormalities in equipment operation in advance and provide reliable warnings to ensure the safe and stable operation of the power plant.Taking the feed water pump of a gas power plant as the research object,the article develops an anomaly detection algorithm model based on the similarity-based modelling(SBM)and sequence probability ratio test(SPRT)based on a large amount of operating historical data.By establishing a similarity model between the current observation vector and the matrix representing the healthy operation state of the equipment,the estimated value of current each feature is obtained.The mean test is performed on the model prediction residuals of each feature at different times to identify equipment operation anomalies.The results indicate that this warning method can issue warnings when the parameters deviate from normal operating conditions before exceeding the alarm value of the control system,compensating for the problem of perception lag and insufficient monitoring caused by fixed value alarms in the original industrial control system and manual monitoring,and providing technical support for predictive maintenance of equipment.
feed water pumpoperation monitoringsimilarity-based modellingsequential probability ratio testanomaly detection
NETL
NSTL
万方数据
