摘要
针对PSO-BP(Particle Swarm Optimization-Back Propagation)神经网络预测模型在冰蓄冷空调冷负荷预测中存在输入输出数据关联度低和预测模型存在误差的情况,提出了一种基于JMP数据处理软件、PSO-BP神经网络和马尔可夫链的组合预测方法。利用JMP处理输入数据,剔除耦合度低的样本,进行PSO-BP神经网络训练,得到冷负荷预测结果,利用马尔可夫链消除系统产生的随机误差得到最终预测结果。结果表明:该组合预测方法对比传统PSO-BP算法其预测精度更高,预测结果符合商场冷负荷的变化规律,满足实际的应用需求。
Aiming at the low correlation between input and output data and the error of prediction model in PSO-BP neural network prediction model, a combined prediction method based on JMP, PSO-BP neural network and Markov chain is proposed. The method first uses JMP data processing software to process the input data and eliminating the low coupling degree samples, then conducts PSO-BP neural network training to obtain the cold load prediction results, and finally uses markov chain to eliminate the random errors generated by the system to obtain the final prediction results. The results show that the combined prediction method has higher prediction accuracy, and the prediction result conforms to the change rule of the shopping mall load, and meets the actual application requirements.
作者
于军琪
井文强
赵安军
任延欢
周梦
黄馨乐
杨雪
Yu Junqi;Jing Wenqiang;Zhao Anjun;Ren Yanhuan;Zhou Meng;Huang Xinle;Yang Xue(School of Construction Equipment and Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China;Xi'an University of Architecture and Technology Engineering Co.,Ltd.,Xi'an 710055,China;Datang Mobile Communications Equipment Co.,Ltd.,Xi'an 710055,China)
出处
《系统仿真学报》
CAS
CSCD
北大核心
2021年第1期54-61,共8页
Journal of System Simulation
基金
陕西省重点研发计划(2017zdcxl-sf-03-02)
陕西省教育厅产业化培育项目(17JF016)
陕西省科技厅专项科研项目(2017JM6106)
校基础研究基金(JC1706)。
