基于多元线性回归与时间序列分析的燃烧室故障预警研究

Early Warning of Combustion Chamber Faults Based on Multiple Linear Regression and Time Series Analysis

为尽早发现电站燃气轮机的燃烧室在运行期间出现的故障隐患,发出报警信息提醒运行人员处理,在提取燃机操作台的监测参数之后,将时间序列分析与多元线性回归相结合进行故障预警。首先采用指数平滑方法对单时间序列指标透平平均排温进行拟合以及预测,通过透平排温的预测结果间接判断燃烧室的燃烧状况;然后选取合适的解释变量与被解释变量建立燃烧室的多元线性回归故障预警模型,该模型可计算出被解释变量的拟合值与残差值,并规定故障预警阈值的上下限,当燃烧室处于正常工作状态时,残差被预警阈值包络,当燃烧室处于故障隐患或者故障状态时,残差值超过预警阈值产生预警信号。仿真结果表明,该方法可提前数小时发现燃烧室存在的运行异常,运行人员处理后可减少因停机带来的经济损失。

In order to find out the hidden troubles of the combustion chamber of the gas turbine of the power station as early as possible,an alarm message is sent to remind the operation personnel to deal with it.After the monitoring parameters of the gas turbine console are extracted,the time series analysis and multiple linear regression are combined to provide fault warning.Firstly,the exponential smoothing method is used to fit and predict the average temperature of the turbine in a single time series index.The combustion state of the combustion chamber is indirectly judged by the prediction result of the turbine temperature;then the appropriate explanatory variables are selected to establish the combustion with the explanatory variables.The multi-linear regression failure warning model of the room,the model can calculate the fitted value and residual value of the interpreted variable,and specify the upper and lower limits of the fault warning threshold.When the combustion chamber is in the normal working state,the residual is predicted by the threshold package.Network,when the combustion chamber is in a hidden danger or fault state,the residual value exceeds the warning threshold to generate an early warning signal.The simulation results show that the method can detect the abnormal operation of the combustion chamber several hours in advance,and the operating personnel can reduce the economic loss caused by the shutdown.