Research on load forecasting method of new cooling and heating system based on SA-GA-CNN-LSTM
CHEN Shiyu¹, LI Zhuangju¹, LIU Hao¹, CHEN Mengyi²
1. Beijing University of Civil Engineering and Architecture, Beijing 102616, China;
2. University of Queensland, Brisbane 4072, Australia
Abstract: In order to improve the prediction accuracy and provide accurate energy storage reference for the new phase change energy storage cooling and heating system, combined with the characteristics of the system, a new load prediction method is proposed. This method first performs fuzzy C-means clustering on the data, and then transmits the clustering results to the model combined by genetic algorithm (GA), self attention mechanism (SA) and convolutional long and short-term memory neural network (CNN-LSTM). Using the measured data of the phase change energy storage cooling and heating system of a substation in Changping, Beijing, the prediction model is trained and determined. Finally, the model is used for load prediction. The comparison between the predicted data and the measured data proves the effectiveness of the model. Compared with single neural network models CNN, LSTM and hybrid neural network models CNN-LSTM and GA-CNN-LSTM, the prediction accuracy of the proposed SA-GA-CNN-LSTM neural network model is the highest. Under the average absolute error (MAPE) index, it is 2.32% higher than the single neural network model LSTM and 1.49% higher than the hybrid neural network model CNN-LSTM.
Keywords: phase change energy storage;load forecasting;convolutional neural network;long and short term memory neural network;self attention mechanism
2023, 49(9):115-122  收稿日期: 2022-05-09;收到修改稿日期: 2022-07-13
基金项目:
作者简介: 陈石毓（1998-），男，北京市人，硕士研究生，专业方向为建筑节能、负荷预测研究
参考文献
[1] 师兵兵, 段哲民, 陆正俊. 神经网络预测中长期电力负荷对比研究[J]. 继电器, 2007(23): 43-45
[2] 刘银波, 吕越. RBF神经网络在船舶电力负荷预测中的应用[J]. 舰船科学技术, 2021, 43(18): 121-123
[3] 尹春杰, 肖发达, 李鹏飞, 等. 基于LSTM神经网络的区域微网短期负荷预测[J]. 计算机与现代化, 2022(4): 7-11
[4] 刘亚珲, 赵倩. 基于聚类经验模态分解的CNN-LSTM超短期电力负荷预测[J]. 电网技术, 2021, 45(11): 4444-4451
[5] 何津民, 张丽珍. 基于自注意力机制和CNN-LSTM深度学习的对虾投饵量预测模型[J]. 大连海洋大学学报, 2022, 37(2): 304-311
[6] 刘旭丽, 莫毓昌, 吴哲, 等. 基于DWT-CNN-LSTM的超短期光伏发电功率预测[J]. 郑州大学学报(理学版), 2022, 54(4): 86-94
[7] 贾睿, 杨国华, 郑豪丰, 等. 基于自适应权重的CNN-LSTM&GRU组合风电功率预测方法[J]. 中国电力, 2022, 55(5): 47-56
[8] 廖清阳, 王军, 胡凯强, 等. 基于深度并行CNN-BiLSTM的能源互联网电负荷和热负荷联合预测模型[J]. 中国测试, 2022, 48(4): 146-153
[9] 方明宽, 宁静, 陈春俊. 高速列车横向蛇行失稳的EEMD-CNN-LSTM预测方法[J]. 中国测试, 2021, 47(7): 79-83
[10] 滕金保, 孔韦韦, 田乔鑫, 等. 基于CNN和LSTM的多通道注意力机制文本分类模型[J]. 计算机工程与应用, 2021, 57(23): 154-162
[11] 王晨阳, 段倩倩, 周凯, 等. 基于遗传算法优化卷积长短记忆混合神经网络模型的光伏发电功率预测[J]. 物理学报, 2020, 69(10): 149-155
[12] 王晓玲, 李克, 张宗亮, 等. 耦合ALO-LSTM和特征注意力机制的土石坝渗压预测模型[J]. 水利学报, 2022, 53(4): 403-412
[13] 王琛, 王颖, 郑涛, 等. 基于ResNet-LSTM网络和注意力机制的综合能源系统多元负荷预测[J]. 电工技术学报, 2022(7): 187-197
[14] 刘智宇, 郝冬, 张妍懿, 等. 基于GA-ELM算法的燃料电池性能预测模型[J]. 电池, 2023, 53(3): 243-247.
[15] 李秉晨, 于惠钧, 丁华轩, 等. 基于CEEMD和LLSTM-ARIMA的短期风速预测[J]. 中国测试, 2022, 48(2): 163-168