Key Laboratory of Guangxi Electric Power System Optimization and Energy-saving Technology,Guangxi University
Innovation Project of Guangxi Graduate Education
在火力发电过程中，蒸汽量的准确测量，对于汽轮机机组的经济稳定运行具有重要的意义。针对传统蒸汽量测量方法精度低的问题，提出了一种基于宽度学习系统(broad learning system, BLS)和Lasso (least absolute shrinkage and selection operator)回归模型的组合预测模型。首先利用One-class SVM (one-class support vector machines)算法对样本进行异常值检测，将检测得到的异常值剔除。然后，采用最大信息系数(maximal informationcoefficient, MIC)对特征变量和蒸汽量进行非线性关联性分析，确定宽度学习系统和Lasso回归模型的输入变量，通过训练得出各自的预测结果。最后，通过最优加权组合法确定两单一模型的权重系数，将它们所得的预测结果线性组合，得到最终的预测结果。实例表明，所建立的组合模型有效地缓解了单一模型在变化剧烈的峰值和谷值预测偏差大的问题，能够准确地预测蒸汽量。
In thermal power generation, the accurate measurement of steam volume is of great significance for the economic and stable operation of steam turbine units. Aiming at the problem of low accuracy of traditional steam volume measurement methods, a combined prediction model based on broad learning system (BLS) and Lasso (least absolute shrinkage and selection operator) regression model is proposed. Firstly, the One-class SVM (one-class support vector machines) algorithm is used to detect the outliers of the samples data and eliminate the detected outliers. Then, the maximum information coefficient (MIC) is used to analyze the nonlinear correlation between characteristic variables and steam volume to determine the input variables of the BLS and Lasso regression model and obtain their prediction results through training, respectively. Finally, the weight coefficients of the two single models are determined by the optimal weighted combination method. The final prediction result is obtained by linearly combining the result of two single models. The example shows that the proposed combined model can effectively alleviate the problem that a large prediction deviation in the peak and valley values with sharp changes using the single model prediction. Moreover, it can accurately predict the steam volume.
封之聪,祝云,高枫. 基于BLS-Lasso组合模型的火电厂蒸汽量预测[J]. 科学技术与工程, , ():复制