基于XGBoost模型的发电机组瞬态工况的识别

Identification of Generator Operation Modes Based on XGBoost Model

电厂发电机从开始启动到进入稳态发电会经历多个瞬态工况的变化,利用实时数据在线、自动识别各种瞬态工况,对机组的数字化运维、智能化健康管理等平台具有重要价值和意义。本文提出了一种基于XGBoost机器学习分类算法的发电机组瞬态工况识别方法,通过构建样本高维特征数据集离线训练XGBoost模型,并根据特征重要性筛选建模特征,建立XGBoost在线模型并进行模型验证,最终实现了在线的发电机组瞬态工况识别。与一些常见的机器学习算法相比,XGBoost算法的识别准确率优于其他识别算法,同时与串行运算的传统决策树算法相比,所用时间也大幅缩短。

There are multiple operation modes during the process from startup to steady-state operation of power plant generators.Using online real-time data to automatically identify different operation modes is important and valuable for generator digital operation and maintenance of the units and intelligent health management platforms.This paper proposes a method for identifying operation modes of generators based on XGBoost machine learning classification algorithm.The XGBoost model is trained offline by constructing high-dimensional features of samples,then the modeling features are selected based on the feature importance to establish online model.Model validation is performed,achieving the online identification of operation modes of generators.Compared with some common machine learning algorithms,XGBoost algorithm has better identification accuracy than other recognition algorithms,and the time required is significantly reduced compared to traditional decision tree algorithms.