基于XGBoost的热电联产供热需求预测方法

XGBoost-based Method for Prediction of Cogeneration Heat Demand

精准预测蒸汽需求量能为热电联产电厂优化生产调度提供关键数据支撑。为了充分利用蒸汽历史用量数据、检修数据、上下游数据、外部环境数据及节假日数据等热电联产电厂中关键数据预测蒸汽需求量,提出了一种基于XGBoost的供热需求预测方法。对上述数据进行特征提取后使用该方法进行模型训练和优化,用于预测供热需求量,最终预测结果的均方根误差为134,远低于ARIMA和随机森林等方法。提出的方法在预测供热需求量的工程实践中表现出良好的效果,可用于支撑实际工程工作。

Accurate prediction of steam demand could provide key data support for the optimization of production scheduling in cogeneration plants.In order to make full use of key data in cogeneration power plants such as historical steam consumption,maintenance data,upstream and downstream data,external environmental data and holiday data to predict steam demand,an XGBoost-based method was proposed for heat demand prediction.After feature extraction of above-mentioned data,model training and optimization were completed in the proposed approach for heat demand prediction.The root mean square error of the final prediction result was 134,much lower than that of ARIMA or random forest method.The proposed method produced good results in engineering practice for heat demand prediction and could be used to support actual engineering work.