反馈精英鲸鱼优化算法优化LSSVM的热耗率软测量建模

提出一种基于反馈精英鲸鱼优化算法(FEWOA)和最小二乘支持向量机(LSSVM)的综合建模方法。首先,针对鲸鱼优化算法(WOA)寻优精度低的问题,提出了反馈精英WOA算法,通过精英策略对当前最优解进行变异操作以避免算法陷入局部最优解;同时,在鲸鱼位置更新后期增加反馈阶段,提高算法的全局搜索能力。数值仿真实验验证了FEWOA算法的优越性。在此基础上,提出了基于FEWOA优化LSSVM的热耗率软测量模型。最后采用某汽轮机组现场收集的运行数据建立汽轮机热耗率预测模型,将FEWOA-LSSVM模型预测结果与其它模型预测结果相比较,结果表明,FEWOA-LSSVM预测模型更能准确地预测汽轮机的热耗率。

Soft Sensor Model of Heat Rate Based on Optimized LSSVM by FEWOA

An integrated modeling method was proposed based on feedback elitist whale optimization algorithm (FEWOA) and least square support vector machine (LSSVM). Firstly, a feedback elitist WOA was proposed to solve the problem of low precision of the WOA. Based on the FEWOA, the current optimal solution was mutated to avoid the algorithm falling into local optimal solution by the elite strategy. Meanwhile, a feedback phase was introduced at the end of the position updating, so that the algorithm improves the precision. Numeric simulation results show that the proposed FEWOA exhibited superior performance over the other aforementioned algorithms. Then, a soft sensor model of heat rate based on optimized LSSVM by FEWOA was proposed. Finally, the turbine heat rate prediction model was established by using the operation data collected from a steam turbine generator set. The prediction results of FEWOA-LSSVM model were compared with the others, the result show that the FEWOA-LSSVM prediction model can predict the heat rate of steam turbine more accurately.