TPE-Based Boosting Short-Term Load Forecasting Method

doi:10.16183/j.cnki.jsjtu.2022.483

Abstract

Abstract:

Short-term load forecasting is generally applied in power system real-time dispatching and day-ahead generation planning, which is of great significance for power system economic dispatching and safe operation of the system. Many researches on short-term load forecasting using smart models have been conducted at home and abroad. However, how to obtain the optimal structure and parameters accurately and quickly poses a challenge to short-term load forecasting, because the prediction performance of smart forecasting methods is more easily affected by the structure and parameters of the method, and the personality difference of the prediction object itself makes it difficult for the parameters to be reused. Aiming at this problem, a tree-structured Parzen estimator (TPE)-based boosting short-term load forecasting method is proposed. The results show that the proposed method can achieve rapid optimization of structure and parameters, which is verified in the application in short-term load forecasting of a southern province in China to improve the prediction accuracy.

Key words: short-term load forecasting, tree-structured Parzen estimator (TPE), ensemble learning, hyperparameter optimization

CLC Number:

TM712

LUO Min, YANG Jinfeng, YU Hui, LAI Yuchen, GUO Yangyun, ZHOU Shangli, XIANG Rui, TONG Xing, CHEN Xiao. TPE-Based Boosting Short-Term Load Forecasting Method[J]. Journal of Shanghai Jiao Tong University, 2024, 58(6): 819-825.

Figures/Tables 8

Fig.1

Fig.2

Tab.1

Tab.2

Fig.3

Tab.3

Fig.4

References 20

[1]	SAAD Z, NUR HAZIRAH A J, SUZIANA A, et al. Short-term load forecasting of 415 V, 11 kV and 33 kV electrical systems using MLP network[C]//2017 International Conference on Robotics, Automation and Sciences. Melaka, Malaysia: IEEE, 2017: 1-5.
[2]	HE T, DONG Z Y, MENG K, et al. Accelerating Multi-layer Perceptron based short term demand forecasting using Graphics Processing Units[C]//2009 Transmission & Distribution Conference & Exposition:Asia and Pacific. Seoul, Korea: IEEE, 2009: 1-4.
[3]	TSAKOUMIS A C, VLADOV S S, MLADENOV V M. Electric load forecasting with multilayer perceptron and Elman neural network[C]//6th Seminar on Neural Network Applications in Electrical Engineering. Belgrade, Yugoslavia: IEEE, 2002: 87-90.
[4]	DRAGOMIR O E, DRAGOMIR F, BREZEANU I, et al. MLP neural network as load forecasting tool on short-term horizon[C]//2011 19th Mediterranean Conference on Control & Automation. Corfu, Greece: IEEE, 2011: 1265-1270.
[5]	KONG W C, DONG Z Y, JIA Y W, et al. Short-term residential load forecasting based on LSTM recurrent neural network[J]. IEEE Transactions on Smart Grid, 2019, 10(1): 841-851.
[6]	RAFI S H, Al MASOOD N, DEEBA S R, et al. A short-term load forecasting method using integrated CNN and LSTM network[J]. IEEE Access, 2021, 9: 32436-32448.
[7]	TAN M, YUAN S P, LI S H, et al. Ultra-short-term industrial power demand forecasting using LSTM based hybrid ensemble learning[J]. IEEE Transactions on Power Systems, 2020, 35(4): 2937-2948.
[8]	LI C J, DONG Z Y, DING L, et al. Interpretable memristive LSTM network design for probabilistic residential load forecasting[J]. IEEE Transactions on Circuits and Systems I: Regular Papers, 2022, 69(6): 2297-2310.
[9]	SHANG C, GAO J W, LIU H B, et al. Short-term load forecasting based on PSO-KFCM daily load curve clustering and CNN-LSTM model[J]. IEEE Access, 2021, 9: 50344-50357.
[10]	AGENG D, HUANG C Y, CHENG R G. A short-term household load forecasting framework using LSTM and data preparation[J]. IEEE Access, 9: 167911-167919.
[11]	LIAO Z F, PAN H H, FAN X P, et al. Multiple wavelet convolutional neural network for short-term load forecasting[J]. IEEE Internet of Things Journal, 2021, 8(12): 9730-9739.
[12]	CHEN Y, LUH P B, GUAN C, et al. Short-term load forecasting: Similar day-based wavelet neural networks[J]. IEEE Transactions on Power Systems, 2010, 25(1): 322-330.
[13]	GUAN C, LUH P B, COOLBETH M A, et al. Very short-term load forecasting: Multilevel wavelet neural networks with data pre-filtering[C]//2009 IEEE Power & Energy Society General Meeting. Calgary, Canada: IEEE, 2009: 1-8.
[14]	BASHIR Z, EL-HAWARY M E. Short term load forecasting by using wavelet neural networks[C]//2000 Canadian Conference on Electrical and Computer Engineering. Conference Proceedings. Navigating to a New Era (Cat. No.00TH8492). Halifax, Canada: IEEE, 2000: 163-166.
[15]	CHEN X, DONG Z Y, MENG K, et al. Electricity price forecasting with extreme learning machine and bootstrapping[J]. IEEE Transactions on Power Systems, 2012, 27(4): 2055-2062.
[16]	RAFIEI M, NIKNAM T, AGHAEI J, et al. Probabilistic load forecasting using an improved wavelet neural network trained by generalized extreme learning machine[J]. IEEE Transactions on Smart Grid, 2018, 9(6): 6961-6971.
[17]	WANG L, ZHANG Z J, CHEN J Q. Short-term electricity price forecasting with stacked denoising autoencoders[J]. IEEE Transactions on Power Systems, 2017, 32(4): 2673-2681.
[18]	SUN W, ZHANG Y X, LI F T. The neural network model based on PSO for short-term load forecasting[C]//2006 International Conference on Machine Learning and Cybernetics. Dalian, China: IEEE, 2006: 3069-3072.
[19]	ZHANG C Q, LIN M, TANG M Y. BP neural network optimized with PSO algorithm for daily load forecasting[C]//2008 International Conference on Information Management, Innovation Management and Industrial Engineering. Taipei, China: IEEE, 2008: 82-85.
[20]	XU X B, LIU W X, ZHOU X, et al. Short-term load forecasting for the electric bus station based on GRA-DE-SVR[C]//2014 IEEE Innovative Smart Grid Technologies-Asia. Kuala Lumpur, Malaysia: IEEE, 2014: 388-393.

超参数	寻优空间
输入历史负荷数据	离散变量,取为集合{第(n-1)d的j时段实际负荷,第(n-1)、(n-7)d的j段实际负荷,第(n-1)、(n-7)、(n-14)d的j时段实际负荷}中的一种.
输入相关因素数据	确定值,取为第n d的j时段气象因素.
输出数据	确定值,取为第n d的j时段预测负荷.
学习率	连续变量,取值空间为[0.01,1].
学习器数量	离散整数变量,取为[10,200]之间的某一整数.

参数	取值
输入数据	第(n-1) d的j时段实际负荷第n d的j时段气象因素
输出数据	第n d的j时段预测负荷
学习率	0.174
学习器数量	166

日期	AdaBoost-TPE	BP-ANN	日期		BP-ANN
5月11日	98.84	97.38	5月22日	97.81	96.53
5月12日	98.21	97.30	5月23日	97.84	96.10
5月13日	98.50	94.26	5月24日	98.33	95.42
5月14日	98.76	98.45	5月25日	98.51	93.67
5月15日	98.37	97.77	5月26日	97.65	95.75
5月16日	98.29	96.97	5月27日	97.74	97.55
5月17日	97.93	97.79	5月28日	97.25	97.08
5月18日	98.45	97.52	5月29日	97.56	97.35
5月19日	98.51	97.06	5月30日	97.73	96.37
5月20日	98.73	96.14	5月31日	98.44	95.52
5月21日	98.43	96.07	平均值	98.18	96.57