Fine non-invasive load monitoring based on CEEMD and GWO-LSTM

doi:10.3969/j.issn.2097-0706.2023.11.005

Abstract

Abstract:

Aiming at the low accuracy of short-term thermal load prediction caused by the lag of central heating system and other factors， a fine non-invasive load monitoring method based on Complementary Ensemble Empirical Mode Decomposition （CEEMD） data processing technology and Gray Wolf Optimization and Long Short-Term Memory （GWO-LSTM） network is proposed. The collected raw data about heat load are decomposed into several stationary intrinsic mode functions （IMFs） by CEEMD， and each IMF is modelled and predicted separately. Then，the final prediction result is obtained by superimposing the predicted values of each IMF. To improve the prediction accuracy， GWO is used to optimize the number of neurons in the hidden layer， the number of training times， and the initial learning rate of the LSTM， then the CEEMD-GWO-LSTM prediction model for short-term heat load is established. Simulation experiments were conducted on the simple LSTM model and CEEMD-LSTM model. The experimental results showed that the root mean square error（RMSE）， mean absolute error（MAE）， and mean absolute percentage error （MAPE）of the prediction made by CEEMD-GWO-LSTM model were 0.591 5 MW， 0.460 2 MW， and 8.083 8%， respectively， which were significantly lower than that of other models.

Key words: central heating system, short-term thermal load prediction, non-invasive load monitoring, long and short-term memory network, Grey Wolf Optimization algorithm, complementary ensemble empirical mode decomposition

CLC Number:

TK11:TP399

LU Weiwei, YIN Linfei. Fine non-invasive load monitoring based on CEEMD and GWO-LSTM[J]. Integrated Intelligent Energy, 2023, 45(11): 36-44.

Figures/Tables 21

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Table 1

Fig.8

Fig.9

Fig.10

Fig.11

Fig.12

Table 2

Table 3

Table 4

Fig.13

Fig.14

Fig.15

Fig.16

Table 5

References 17

[1]	GHALEHKHONDABI I, ARDJMAND E, WECKMAN G R, et al. An overview of energy demand forecasting methods published in 2005—2015[J]. Energy Systems, 2016, 8(2):1-37. doi: 10.1007/s12667-017-0233-0
[2]	CARMO C M RDO, CHRISTENSEN T H. Cluster analysis of residential heat load profiles and the role of technical and household characteristics[J]. Energy and Buildings, 2016, 125: 171-180. doi: 10.1016/j.enbuild.2016.04.079
[3]	PAPALEXOPOULOS A D, HESTERBERG T C. A regression-based approach to short-term system load forecasting[J]. IEEE Transactions on Power Systems, 1990, 5(4): 1535-1547. doi: 10.1109/59.99410
[4]	WANG X Y, QI W G. Research on the chaotic characteristics for heating load time series[C]// 6th IEEE Conference on Industrial Electronics and Applications(ICIEA). IEEE, 2011:1323-1327.
[5]	SUN H, BI Z. Hybrid forecasting model of electric heating load based on CART decision tree regression[C]// 2021 IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA). IEEE, 2021:204-208.
[6]	ZHANG N, LI Z, ZOU X, et al. Comparison of three short-term load forecast models in Southern California[J]. Energy, 2019, 189:116358. doi: 10.1016/j.energy.2019.116358
[7]	CHENG D Z. On logic-based intelligent systems[C]// Proceedings of 5th International Conference on Control and Automation. IEEE, 2005: 71-76.
[8]	彭维珂, 聂椿明, 陈衡, 等. 基于智能算法的空冷火电机组负荷预测研究[J]. 华电技术, 2021, 43(3): 57-64.
	PENG Weike, NIE Chunming, CHEN Heng, et al. Study on load forecasting for air cooling thermal power units based on intelligent algorithm[J]. Huadian Technology, 2021, 43(3): 57-64.
[9]	GUO B, CHENG L, XU J, et al. Prediction of the heat load in central heating systems using GA-BP algorithm[C]// 2017 International Conference on Computer Network, Electronic and Automation (ICCNEA).IEEE, 2017:441-445.
[10]	曹力, 潘巧波, 王明宇, 等. 基于混合核函数支持向量机的风电机组发电机温度预警方法[J]. 华电技术, 2020, 42(5): 43-49.
	CAO Li, PAN Qiaobo, WANG Mingyu, et al. Early warning method for wind turbine generator temperature based on HK-SVM[J]. Huadian Technology, 2020, 42(5): 43-49.
[11]	DU Y, CUI N X, LI H X, et al. The vehicle's velocity prediction methods based on RNN and LSTM neural network[C]// 2020 Chinese Control and Decision Conference (CCDC). IEEE, 2020:99-102.
[12]	廖康, 吴益平, 李麟玮, 等. 基于时间序列与GWO-ELM模型的滑坡位移预测[J]. 中南大学学报(自然科学版), 2019, 50(3):619-626.
	LIAO Kang, WU Yiping, LI Linwei, et al. Displacement prediction model of landslide based on time series and GWO-ELM[J]. Journal of Central South University (Science and Technology), 2019, 50(3):619-626.
[13]	高明, 郝妍. 基于BiLSTM网络与误差修正的超短期负荷预测[J]. 综合智慧能源, 2023, 45(1): 31-40. doi: 10.3969/j.issn.2097-0706.2023.01.004
	GAO Ming, HAO Yan. Ultra-short-term load forecasting based on BiLSTM network and error correction[J]. Integrated Intelligent Energy, 2023, 45(1): 31-40. doi: 10.3969/j.issn.2097-0706.2023.01.004
[14]	李彬, 胡纯瑾, 王婧. 基于EEMD-BiLSTM的可调节负荷预测方法[J]. 综合智慧能源, 2022, 44(9): 33-39. doi: 10.3969/j.issn.2097-0706.2022.09.005
	LI Bin, HU Chunjin, WANG Jing. Prediction method for adjustable load based on EEMD-BiLSTM[J]. Integrated Intelligent Energy, 2022, 44(9): 33-39. doi: 10.3969/j.issn.2097-0706.2022.09.005
[15]	黄文聪, 张宇, 杨远程, 等. 模糊信息粒化和GWO-SVM算法结合的短期风速范围预测[J]. 华侨大学学报(自然科学版), 2020, 41(5):674-682.
	HUANG Wencong, ZHANG Yu, YANG Yuancheng, et al. Short-term wind speed range prediction based on fuzzy information granulation and GWO-SVM algorithm[J]. Journal of Huaqiao University (Natural Science), 2020, 41(5):674-682.
[16]	魏骜, 茅大钧, 韩万里, 等. 基于EMD和长短期记忆网络的短期电力负荷预测研究[J]. 热能动力工程, 2020, 35(4):203-209.
	WEI Ao, MAO Dajun, HAN Wanli, et al. Short-term load forecasting based on EMD and long short-term memory neural networks[J]. Journal of Engineering for Thermal Energy and Power, 2020, 35(4):203-209.
[17]	LI J, JIANG X, SHAO L, et al. Energy consumption data prediction analysis based on EEMD-ARMA model[C]// 2020 IEEE International Conference on Mechatronics and Automation (ICMA). IEEE, 2020:1338-1342.

时间		平均热负荷	时间		平均热负荷
2016-11-01	09:00	5.7	2016-12-12	04:00	3.8
	10:00	5.8		05:00	3.9
	11:00	5.5		06:00	4.1
	12:00	5.9		07:00	4.4
	13:00	6.3		08:00	4.3
	14:00	8.9		09:00	4.5
	15:00	10		10:00	4.3
	16:00	9.8		11:00	5.4
	17:00	10.8		12:00	5.2
	18:00	8.7		13:00	7.1
	19:00	8.1		14:00	5.8
	20:00	9.9		15:00	5.3
	21:00	8.5		16:00	6.9
	22:00	6.3		17:00	6.7
	23:00	5.3		18:00	6.5
…		…		19:00	5.2
2016-12-12	00:00	6.0		20:00	5.5
	01:00	5.9		21:00	4.9
	02:00	4.3		22:00	4.0
	03:00	4.1		23:00	3.5

参数	数值
灰狼种群个数	20
最大迭代次数	20
优化维度	3
LSTM隐藏层神经元数目i	[1, 200]
训练次数j	[1, 200]
LSTM初始学习率μ	[0.001,0.010]

参数	数值
输入特征维数	1
隐藏层神经元数目	100
全连接层数	1
初始学习率	0.003
学习率更新周期	125
学习率更新因子	0.2
求解器	Adam

分量	隐藏层节点数	训练次数	初始学习率
I₁	1.336 5	200.000 0	0.009 9
I₂	93.398 8	168.932 7	0.006 8
I₃	200.000 0	146.320 2	0.009 0
I₄	92.562 4	200.000 0	0.009 5
I₅	133.801 8	200.000 0	0.010 0
I₆	33.181 6	172.798 0	0.009 5
I₇	14.607 7	129.445 4	0.004 3
I₈	8.866 5	186.174 9	0.001 6
I₉	175.654 9	124.115 3	0.009 5
R	49.579 2	33.243 0	0.008 0

模型	δ_RMSE/MW	δ_MAE/MW	δ_MAPE/%
LSTM	1.198 5	0.958 8	16.317 8
CEEMD-LSTM	1.161 0	0.767 6	13.033 6
CEEMD-GWO-LSTM	0.591 5	0.460 2	8.083 8