Health Condition Identification and Quantitative Assessment of Emergency Diesel Generator Sets Based on Back Propagation Neural Network and Mahalanobis Distance

Abstract

To address the problem of a comprehensive assessment of emergency diesel generator set health status, a quantitative assessment method of emergency diesel generator set health status based on data-driven and distance measurement is proposed. BP neural network was used to extract the feature sequences of emergency diesel generator sets under different health states. An emergency diesel generator set health state identification model was established. To further quantitatively evaluate the health level of the emergency diesel generator set, the distance between each state of the emergency diesel generator set and the health state is measured by introducing the Mahalanobis distance algorithm, which realizes the unification of both qualitative and quantitative evaluation of the health state of the emergency diesel generator set. The operating states of diesel generator sets were subdivided into four states: healthy, sub-healthy, abnormal, and fault, and a simulation test study was conducted using the actual measurement data set of emergency diesel generator set operation. It is verified that the proposed method can classify the operating status of diesel generator sets more accurately and obtain the health assessment score index. The integration of qualitative and quantitative assessment of the health status of emergency diesel generator sets is realized, and the status warning is carried out accordingly, which provides an accurate basis for taking different maintenance and repair measures in a targeted manner.