为解决生物质大比例直燃耦合燃煤机组中煤-生物质动态前馈微分时间与增益难以精量化整定的难点,提出了一种基于模糊补偿的锅炉主控前馈预测模型(CF-ICM)。首先,结合锅炉主控前馈,设计相关建模变量,并利用决策树与LASSO回归构建的自适应特征选择方法对混沌分析的煤量时间序列高维特征及最大信息系数筛选后的时延建模变量进行特征筛选；然后,采用引力搜索算法优化的ELMAN网络构建炉主控动态前馈预测模型；最后,建立了模糊误差补偿控制器,达到误差修正的目的。基于660MW生产数据的实验结果表明所提模型的预测精度小于0.25。

A feed-forward prediction model (CF-ICM) based on fuzzy compensation is proposed to solve the difficulty of accurately tuning the differential time and gain of boiler master control for dynamic feed-forward in large-scale direct coupled biomass on coal-fired units. Firstly, combined with the relevant design variables for boiler master control feed-forward, the modeling variables are determined by using an adaptive feature selection method constructed with CART decision tree and LASSO regression, which is developed to perform feature screening on the high-dimensional characteristics of the coal quantity time series analyzed for chaos,as well as the time-delay modeling variables filtered by the maximum information coefficient; Then, the ELMAN network optimized by the gravity search algorithm is used to establish the dynamic feed-forward prediction model for the boiler master control; Finally, a fuzzy error compensation controller is established to achieve the purpose of correcting prediction errors. The experimental results based on 660MW operational data illustrates that the proposed model prediction accuracy less than 0.25.