火电机组热力设备部件建模是实现机组性能监测以及电厂数字孪生的关键。火电热力设备内部结构复杂,纯机理建模困难,而数据驱动模型无法明确设备内部特性的变化以评价设备的健康状况。本文考虑火电厂储存有大量运行历史数据的特点,将运行数据与热力设备内的物理过程相结合,采用主导因素建模法,实现对回热加热器、级组、管道的数理混合建模。在建模过程中,本文首先由机理公式推导出决定各热力设备工作性能的特性参数,再寻找影响特性参数的主导因素,通过大量运行历史数据拟合出特性参数与主导因素之间的数学关系式。为便于所建各热力设备模型在监测系统中的应用,本文以工质流入设备的热力参数为模型输入量,以流出设备的热力参数为模型输出量。通过对比各模型输出应达值与电厂测量值,相对误差不超过1.0%。针对汽轮机低压缸湿蒸汽比焓不能由温度、压力确定,造成加热器、级组模型在汽轮机低压区不能通用的问题,本文对同一工况分别采用能量守恒法和汽轮机低压缸膨胀线外推法计算湿蒸汽比焓,并进行对比。考虑到运行数据流量测点随机误差大的特点,本文推荐采用低压缸膨胀线外推法确定低压缸排汽焓及末级抽汽焓。对于数理混合建模的模型,对热力设备内机理的把握和运行数据的质量是制约模型精度的两大因素。本文针对运行测量数据误差大的问题,提出采用数据协调的方法,通过系列消去检验,剔除测点中的显著误差。再利用测量变量在空间上的冗余,通过约束方程实现随机误差的相互抵消,降低基准流量的测量不确定度。最后,本文就回热加热器模型在机组处于非稳态工况,出现模型计算应达值超前于测量值变化的情况,改进了加热器模型。在改进过程中,考虑了加热器本体对能量的蓄积效应,在加热器模型中新增特性参数,引入了时间尺度,考虑了加热器模型输入量在时间上的变化率对模型输出量的影响。验证结果显示,新增特性参数能提高加热器模型的精度。
The modeling of thermal equipment components of thermal power unit is the key to realize unit performance monitoring and digital twin of power plant. The internal structure of thermal power equipment is complex, and it is difficult to build pure mechanism model. However, the data-driven model can not identify the changes of internal characteristics of the equipment to evaluate the health status of the equipment. In this paper, considering the characteristics of a large number of operation history data stored in thermal power plant, the operation data is combined with the physical process in thermal equipment, and the dominant factor modeling method is adopted to realize the mathematical hybrid modeling of regenerative heater, stage group and pipeline.In the process of modeling, this paper first deduces the characteristic parameters that determine the working performance of each thermal equipment from the mechanism formula, and then finds out the dominant factors that affect the characteristic parameters. Through a large number of historical operation data, the mathematical relationship between the characteristic parameters and the dominant factors is fitted. In order to facilitate the application of the established thermal equipment model in the monitoring system, this paper takes the thermal parameters of the working medium flowing into the equipment as the model input, and the thermal parameters of the outflow equipment as the model output. By comparing the expected value of each model with the measured value of power plant, the relative error is less than 1%.In view of the fact that the specific enthalpy of wet steam in low pressure cylinder of steam turbine can not be determined by temperature and pressure, so that the heater and stage group model can not be used in the low pressure area of steam turbine, this paper calculates the specific enthalpy of wet steam by energy conservation method and expansion line extrapolation method of low pressure cylinder of steam turbine under the same working condition, and compares them. Considering the large random error of flow measurement points in operation data, this paper recommends the extrapolation method of expansion line of low pressure cylinder to determine the exhaust enthalpy and final stage extraction enthalpy of low pressure cylinder.For the model of mathematical hybrid modeling, the grasp of the internal mechanism of thermal equipment and the quality of operation data are two major factors that restrict the accuracy of the model. In this paper, aiming at the problem of large error in operation measurement data, the method of data coordination is proposed to eliminate the significant error in measurement points through a series of elimination tests. Then, the redundancy of measurement variables in space is used to realize the mutual cancellation of random errors by constraint equation, so as to reduce the measurement uncertainty of reference flow.Finally, this paper improves the heater model when the unit is in non steady state and the calculated value is ahead of the measured value. In the improvement process, the energy accumulation effect of the heater body is considered, the characteristic parameters are added in the heater model, the time scale is introduced, and the influence of the change rate of the heater model input in time on the model output is considered. The results show that the new characteristic parameters can improve the accuracy of the heater model.