为解决凝结水蓄热辅助火电机组一次调频时,负荷响应滞后的问题。本文首先使用变分模态分解(variational modal decomposition, VMD)对电网频率进行模态分解,去掉频率数据中的干扰因素。对各分模态量使用最小二乘支持向量机(Least Squares Support Vector Machines,LSSVM)进行分别预测,为克服LSSVM算法模型中核函数宽度因子和正则化两参数的敏感性,提高算法的泛化能力,这里引入一种减法平均优化算法(subtraction average based optimizer, SABO)对两个参数进行寻优。最后将各模态分量进行重构,最终得出未来15s的电网频率预测值。仿真结果表明使用此种预测方法,频率预测的误差较小。在频率预测值的基础上,通过凝结水系统辅助机组调频的协同控制逻辑,达到提高机组一次调频支撑能力。

To solve the problem of load response lag during the primary frequency regulation of thermal power units assisted by condensed water thermal storage. This article first uses Variational Modal Decomposition (VMD) to perform modal decomposition on the power grid frequency, removing interference factors from the frequency data. We use Least Squares Support Vector Machines (LSSVM) to predict each modal variable separately. To overcome the sensitivity of kernel width factor and regularization parameters in the LSSVM algorithm model and improve the algorithm"s generalization ability, we introduce a subtraction average based optimizer (SABO) algorithm to optimize the two parameters. Then superimpose the various modal components to obtain the predicted power grid frequency for the next 15 seconds. The simulation results show that using this prediction method results in smaller frequency prediction errors. Based on the frequency prediction value, the coordinated control logic of the condensate system assisted unit frequency regulation can improve the safety of the unit and the primary frequency regulation support capability.

国家自然科学支持(项目名称,52376007)。This work is supported by National Natural Science Foundation of China (52376007)．第一作者李展(1990—),男,汉,山东菏泽,硕士,高级工程师。研究方向火力发电厂自动控制优化、火电蓄热深度利用、火-储联合调频。E-mail：huadianlizhan@163.com。