建立有效的电站设备性能监测模型，对指导设备进行预测性维护具有重要意义。以热工过程中的SCR脱硝系统为例，基于海量且连续的运行数据，从概念漂移的角度对催化剂活性进行了监测以及劣化分析。首先建立了SCR系统健康状态的离线模型，其次将实时运行数据输入模型进行了测试，并构造了衡量模型性能的统计指标以反映催化剂性能的劣化程度。将模型仿真算例和SCR系统实际运行数据进行对比验证，验证结果表明，监测指标随时间推移呈指数分布变化关系，可有效跟踪和监测催化剂活性变化，为电厂运行人员适时更换催化剂提供科学指导。

The establishment of an effective performance monitoring model in the power plant is significant to guide the predictive maintenance for the key equipment.Taking the SCR denitrification system in thermal process as an example, based on the massive and continuous operation data, the activity of the catalyst in the SCR system was monitored, and performance degradation was analyzed from the perspective of concept drift.Firstly, the offline model of health state of the SCR system was established, and then the input model of realtime operation data was tested.Meanwhile, the statistical indicators measuring the model performance were constructed to reflect the activity of the catalyst and the related deterioration degree.The comparison between the simulation examples of model and the field operation data of SCR system shows that the variation of monitoring indicators with time in exponential distribution, which can effectively track and monitor the activity variation of the catalyst, so as to provide the scientific guidance for the operators in power plant to replace the catalyst at the right time.

TM621.1