为提高机组热耗率在线计算的精度与鲁棒性，提出多种群果蝇优化算法(Multipopulation fruit fly optimization algorithm,MFOA)和广义回归神经网络(Generalized regression neural network,GRNN)相结合的汽轮机热耗率预测模型。以影响机组热耗率的主要运行参数为输入参数，建立基于GRNN的机组热耗率计算模型，并进一步采用改进的多种群果蝇优化算法优化GRNN模型中的光滑因子。将所建MFOAGRNN热耗率预测模型应用到某1 000 MW机组中，结果表明该模型具有很好的计算精度，在测量数据发生方差增大、定值偏移等异常情况时该模型也能给出可靠的计算结果，具有较强的泛化能力和鲁棒性，满足实际工程需要。

In order to improve the accuracy and robustness of online calculation of unit heat rate,a prediction model of steam turbine heat rate based on multipopulation fruit fly optimization algorithm (MFOA) and generalized regression neural network (GRNN) is proposed.Taking the main operating parameters affecting the unit heat rate as input parameters,the calculation model of unit heat rate based on GRNN is established,and the smoothing factor in GRNN model is further optimized by the improved multipopulation fruit fly optimization algorithm.The built MFOAGRNN heat rate prediction model is applied to a 1 000 MW unit,and the results show that the model has good calculation accuracy.The model can also generate reliable calculation results when the variance of measured data increases or the fixed value migration occurs.It shows that the proposed model has strong generalization ability and robustness,which can meet the actual engineering needs.

TK261

国家自然科学基金(51106099)；气动噪声控制重点实验室开放基金(ANCL201602)；2014上海市军民结合专项项目（No.26）