摘要
为更快捷、方便地获取煤矸石和煤炭等燃料的发热量,文中提出一种将机器学习技术与参数优化算法相结合的煤矸石发热量预测模型。首先,对新疆主要矿区煤矸石开展工业分析和发热量的实验测量,构建新疆煤矸石工业分析指标和发热量数据库;然后,建立以煤矸石工业分析指标为输入,基于支持向量回归(SVR)、随机森林(RF)和多层感知(MLP)神经网络的煤矸石发热量非线性预测模型;同时,为提高模型预测精度,引入麻雀搜索算法(SSA)和黏菌算法(SMA)对模型关键参数进行优化;最后,对所构建的几种煤矸石发热量预测模型进行对比分析。结果表明:与SSA相比,采用SMA优化参数能够更好地提高三种预测模型的精度;SVR和MLP模型的预测性能优于RF模型,其中SMA-SVR模型的收敛速度最快且预测精度较高,适用于煤矸石等燃料发热量的预测研究。
In order to obtain the calorific value of coal-gangue,coal and other fuels more quickly and conveniently,a coal-gangue calorific value prediction model combing machine learning with parameter optimization algorithm is proposed.The industrial analysis and experimental measurement of calorific value of coal gangue from major mining areas in Xinjiang are conducted.The industrial analysis index and calorific value database of coal gangue from major mining areas in Xinjiang are constructed.The nonlinear prediction model of coal gangue calorific value is established based on support vector regression(SVR),random forest(RF)and multi-layer perceptive(MLP)neural network with the industrial analysis index of coal gangue as input.The key parameters of these prediction models are optimized by introducing the sparrow search algorithm(SSA)and slime mold algorithm(SMA)to improve the prediction accuracy.The predictive performance of these constructed coal-gangue calorific value prediction models are compared and analyzed.The results show that,in comparison with the SSA,parameters optimized by SMA can better improve the accuracy of the three prediction models.The prediction performance of SVR model and MLP model are better than that of RF model,among which SMA-SVR prediction model has the fastest convergence speed and higher accuracy,and is suitable for the prediction for the calorific value of coal-gangue and other fuels.
作者
高湘彬
贾博
李根
马小晶
GAO Xiangbing;JIA Bo;LI Gen;MA Xiaojing(Xinjiang Xinneng Group Company Limited,Urumqi Electric Power Construction and Commissioning Institute,Urumqi 830000,China;Electric Power Research Institute,State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830000,China;School of Electrical Power Engineering,South China University of Technology,Guangzhou 510641,China;School of Electrical Engineering,Xinjiang University,Urumqi 830017,China)
出处
《现代电子技术》
2023年第14期168-174,共7页
Modern Electronics Technique
基金
新疆新能集团有限责任公司乌鲁木齐电力建设调试所科技项目(SGXJXN00TSJS2200141)。
