节能降耗一直是火电行业备受关注的研究方向，是落实“双碳”战略的重要措施。提出了一种基于加权支持向量机(WSVM)的火电机组能效寻优及运行状态评价方法：首先，利用机组运行数据，依据稳定性判据和样本密度法筛除非稳态及异常工况下的运行数据；其次，采用能效择优方法确定模型训练时的样本权重，利用WSVM建立火电机组能效寻优模型；依据该模型，获得火电机组各工况下主要指标及参数最优值；最后，在机组运行过程中，将其与过程值进行比较，并通过构建的Mandani模糊评价模型，实现对机组运行状态的实时量化评价。通过对某660MW超临界火电机组应用实例表明，在30%~100%额定负荷区间，所提方法获得的最优供电煤耗较聚类算法低0.9~4.8g/kWh，能较好反映火电机组能效最优状态，为判断机组节能降耗空间提供依据，基于模糊模型的状态评价可为火电机组运行优化调整提供指导。

Energy conservation and consumption reduction have always been a research direction of great concern in the thermal power industry, and are important measures to implement the "dual carbon" strategy. A method is proposed for energy efficiency optimization and operation status evaluation of thermal power units based on weighted support vector machine (WSVM). Firstly, using the unit operation data, the operation data under steady-state and abnormal conditions are screened based on stability criteria and sample density. Secondly, adopting the energy efficiency optimization method to determine the sample weights during model training, and then using WSVM to establish an energy efficiency optimization model for thermal power units; Based on the model, obtain the optimal values of the main monitoring parameters for each operating condition of the thermal power unit. Finally, during the operation of the unit, it is compared with the process values and the Mandani fuzzy evaluation model is constructed to achieve real-time quantitative evaluation of the unit's operating status. An application example of a 660MW supercritical thermal power unit shows that the optimal coal consumption obtained by the proposed method is 0.9~4.8g/kWh lower than that of the clustering algorithm in the 30%~100% rated load range, which can better reflect the optimal energy efficiency state of the thermal power unit, providing a basis for judging the energy-saving and consumption reducing space of the unit. The state evaluation results based on fuzzy models can provide guidance for the operation optimization and adjustment of thermal power units.

国网湖南省电力有限公司科技项目(NO.5216A521N00H );国家自然科学青年基金项目(NO. 52006016)