摘要
提出了一种基于GA-XGBoost模型的水下螺旋盘管换热器换热量预测方法。首先,搭建了基于PLC和Wincc上位机的用于模拟水下螺旋盘管换热器和地表水源热泵联合运行过程的实验测试平台,由此获取到不同工况下的实验数据集。提出了一种基于遗传算法和XGBoost的复合模型对换热量进行预测,该模型采用遗传算法对XGBoost模型进行优化,同时引入一种新的复合适应度函数指导遗传算法的参数寻优过程;最后基于所采集的实验数据集进行验证,以换热器多个特征作为模型输入,以换热量作为输出,得到平均绝对百分比误差为3.21%,均方根误差为0.476,决定系数达到0.974,结果表明该方法在数据集上良好的预测性能和泛化能力,为水下换热器的设计和性能优化提供了参考依据。
A heat exchange prediction method for underwater spiral coil heat exchanger based on the GA-XGBoost model is proposed.Firstly,an experimental test platform based on PLC and Wincc host computer is built for simulating the joint operation process of underwater spiral coil heat exchanger and surface water heat pump,from which the experimental datasets under different working conditions are obtained;then,a composite model based on genetic algorithm and XG-Boost is proposed for predicting heat exchanger capacity,which adopts genetic algorithm to opti-mize the XGBoost model,and introduces a new composite model to predict heat exchanger capaci-ty,and a new composite fitness function is introduced to guide the parameter optimization process of the genetic algorithm;finally,a validation is carried out based on the collected experimental dataset,with multiple features of the heat exchanger as the model inputs and heat exchanger ca-pacity as the outputs,and an average absolute percentage error of 3.21%is obtained,the root-mean-square error is 0.476,and the coefficient of determination reaches 0.974,and the results show that the method has a good prediction performance and generalization ability on the dataset,which provides a reference basis for the design and performance optimization of underwater heat exchangers and has high value for engineering applications.
作者
陈于飞
蔡文剑
蔡慧
黄瑶瑶
Chen Yufei;Cai Wenjian;Cai Hui;Huang Yaoyao(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China;School of Information Science and Engineering,Ningbo Tech University,Ningbo 315100,China)
出处
《低温工程》
CAS
CSCD
北大核心
2024年第5期98-103,110,共7页
Cryogenics
基金
浙江省教育厅一般科研项目(Y202353705)
宁波市建设科研计划项目(No.202303040)。
