基于稳态检测的电厂数据预处理研究

发布时间：2018-05-13 00:03

本文选题：稳态检测 + SSD算法　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：稳态检测对热工过程中评价设备的性能优异、系统过程建模及优化、故障检测机制建立及过程辨识等均体现出非常重要的意义。在热电厂海量历史数据预处理过程当中,只有稳态工况下的数据才能真实客观的反映出系统,并且在以后的系统建模及辨识中都要获得稳态工况,所以将稳态工况下变量的数据从海量数据库中挑选出来非常重要。SSD(Steady State Detection)算法作为一种有效的稳态检测算法得到了广泛的应用,将SSD算法与数据滤波相结合,可以显著提高监测点稳态状况判断的准确度。将EWMA(Exponentially Weighted Moving Average,EWMA)滤波与SSD算法结合,形成EWMA-SSD方法,作为本文的研究对象进行分析研究。本文首先分析了SSD算法和EWMA滤波进行稳态检测的机制,推导出SSD算法中的窗宽n和EWMA的滤波因子?之间存在的等价关系。考虑到实际工况中监测数据分布具有不确定性,在SSD算法的基础上,对算法在控制限的求取和监控指标方面做出了改进。采用基于核函数的非参数控制限算法,用核函数拟合出监控指标的概率密度函数,然后计算出该概率密度函数下满足检验水平α的控制限,并结合EWMA滤波对监控指标进行改进后,对过程数据进行稳态检测。最后将SSD稳态检测法应用到实际过程当中,本文以某火电厂600MW机组1年的历史数据进行稳态检测,改进后的SSD算法相比较传统而言,稳态检出率显著提高。从而验证了方法的有效性。最后将多变量稳态检测后的相近负荷下的稳态数据段筛选出来,对涉及的变量两两组合后进行聚类分析,发现在负荷相近下稳态工况可以划分,对不同的工况下运行的状态及经济性分析和进阶的优化性运行进行分析后,发现不同聚类群的锅炉效率存在明显差异,为后续针对这种差异所做的研究工作提供了基础。
[Abstract]:Steady-state detection is very important for evaluating the performance of the equipment in thermal process, modeling and optimizing the system process, establishing the fault detection mechanism and identifying the process. In the preprocessing process of massive historical data in thermal power plant, only the data under steady condition can reflect the system objectively and realistically, and the steady state condition should be obtained in the later modeling and identification of the system. Therefore, it is very important to select the data of variables under steady condition from the massive database. The algorithm is widely used as an effective steady-state detection algorithm. The SSD algorithm is combined with the data filtering. The accuracy of judging the steady state of monitoring points can be improved significantly. The EWMA-SSD method is formed by combining the EWMA(Exponentially Weighted Moving average EWMA(Exponentially Weighted Moving filtering with the SSD algorithm, and it is analyzed and studied as the research object of this paper. In this paper, the mechanism of steady-state detection based on SSD algorithm and EWMA filter is analyzed, and the window width n and the filter factor of EWMA in SSD algorithm are deduced. The equivalent relation that exists between. Considering the uncertainty of the monitoring data distribution in the actual working conditions, based on the SSD algorithm, an improvement is made in the calculation of the control limit and the monitoring index. The nonparametric control limit algorithm based on kernel function is used to fit the probability density function of the monitoring index, and then the control limit satisfying the test level 伪 is calculated under the probability density function. After improving the monitoring index with EWMA filter, the steady state detection of the process data is carried out. Finally, the SSD steady-state detection method is applied to the practical process. The steady-state detection rate of the improved SSD algorithm is significantly higher than that of the traditional one year history data of the 600MW unit in a thermal power plant. The validity of the method is verified. Finally, the steady-state data segment under the similar load after multivariable steady-state detection is screened out, and the cluster analysis is carried out after the combination of the two variables involved, and it is found that the steady-state working condition can be divided under the similar load. After analyzing the state and economy of operation under different operating conditions and the advanced optimal operation, it is found that there are obvious differences in boiler efficiency among different cluster groups, which provides a basis for further research on this difference.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM621

【参考文献】