基于神经网络的凝汽器变工况模型

Off-Design Model for the Condenser Based on Neural Network

汽轮发电机组冷端系统运行优化问题中,因凝汽器设备运行一段时间后污染结垢及设备老化而性能改变,传统的凝汽器变工况特性模型计算值与实际值偏差较大,影响优化效果。针对上述问题,以600MW汽轮机组凝汽器为研究对象,在大型历史数据集的基础上,采用BP神经网络建立了该机组凝汽器变工况特性模型,仿真结果表明机组背压的预测计算结果与实际数据误差在4.5%之内,大部分误差不超2.0%。基于上述模型对凝汽器变工况特性进行了计算及敏感性分析,结果表明机组背压对循环冷却水进口水温变化最为敏感,其次是负荷率变化,最后是循环水流量变化;在机组高负荷率和入口冷却水温较高时,增加循环水流量降低机组背压效果更加明显。

In the optimization of the cold end operation of the steam turbine unit, due to problems such as contamination, scaling and equipment aging after the operation of the condenser equipment for a period of time, the traditional off-design condition characteristic model of condensers has a large deviation between the calculated value of the model and the actual value, which lead to deviations in optimization results. To address the above problem, taking a condenser of a 600 MW steam turbine unit as the research object, based on a large historical data set by using data mining technology, the off-design characteristic model for the condenser of the unit was established based on BP neural network. The simulation results show that the error between the predicted calculation result of the unit back pressure and the actual data is within 4.5%,and most of the error does not exceed 2.0%. Based on the above model, the off-design characteristics of the condenser were calculated, and the variation law and sensitivity analysis of the unit back pressure with the main influencing factors were analyzed. The results show that the unit back pressure is most sensitive to the variation of circulating cooling water inlet temperature, followed by the load rate, and finally, the circulating water flow. When the unit load rate is high and the inlet cooling water temperature is high, increasing the circulating water flow improves the unit vacuum more obviously.