基于遗传注意力机制的DLSTM电力系统混沌预测

doi:10.19596/j.cnki.1001-246x.8400

摘要/Abstract

摘要：

提出一种基于遗传算法优化注意力机制的深度长短期记忆网络(DLSTM)方法，用于电力系统的混沌预测。通过传递共享参数，将遗传算法优化的注意力机制加入DLSTM模型中，可以挖掘时间序列中潜在特征，同时避免陷入局部优化。该方法是一种受进化计算方法启发的寻优方法，可以很好地学习注意力层中的参数。电力系统混沌预测实验表明所提模型比其他参考模型具有更高的预测精度和长期预测能力。

关键词: 电力系统, 深度学习, 深度长短期记忆网络, 混沌预测, 注意力机制

Abstract:

A deep learning algorithm using deep long short-term memory and genetic attention mechanism (DLSTM-GA) is proposed for the prediction of chaotic behavior of power system. With shared parameters, attention mechanism is added to optimize DLSTM model based on genetic algorithm. One can find potential characteristics in time sequence and avoid the local optimization. Inspired by evolutionary computation of optimization method, the method is a good way to learn parameters in the attention layer. It shows that the trained DLSTM-GA network not only has higher prediction accuracy than the reference model, but also has long-term prediction ability.

Key words: power system, deep learning, deep long short-term memory, chaos prediction, attention mechanism

卢英东, 韦笃取. 基于遗传注意力机制的DLSTM电力系统混沌预测[J]. 计算物理, 2022, 39(3): 371-378.

Yingdong LU, Duqu WEI. Power System Chaos Prediction Based on DLSTM with Genetic Attention Mechanism[J]. Chinese Journal of Computational Physics, 2022, 39(3): 371-378.

图/表 8

图1 遗传算法优化具有注意力机制的DLSTM流程图

Fig.1 Illustration of DLSTM using genetic algorithm with optimize attention mechanism

表1 励磁限制电力系统参数的符号和取值

Table 1 Symbols and values of excitation-limited power system parameters

符号	参数	数值	符号	参数	数值
x_d	发电机d轴同步电抗	1	E_{fd_max}	励磁控制器输出上限	5
x′_d	发电机d轴瞬变电抗	0.4	E_{fd_min}	励磁控制器输出下限	0
T′_d0	发电机d轴励磁绕组时间常数	10	E_fd0	限制器输入参考电压	2
M	发电机转动惯量	10	T_A	励磁控制器时间常数	1

图2 考虑励磁限制的电力系统测试集ω(t)的时间序列

Fig.2 Time series of test set ω(t) of a power system considering excitation constraints

图3 遗传算法训练DLSTM-GA的过程

Fig.3 The training process of DLSTM-GA with genetic algorithm

图4 注意力分布的热图

Fig.4 Heat maps of the distribution of attention

图5 DLSTM-GA的测试数据与预测数据

Fig.5 Test data and forecast data of DLSTM-GA

图6 DLSTM-GA长期预测性能

Fig.6 The performance of long-term prediction of DLSTM-GA

表2 DLSTM-GA、DLSTM-Attention、DLSTM、Multi-RNN、DGRU、DA-RNN、LSTM-Attention、LSTM的最佳结果

Table 2 Best results of DLSTM-GA, DLSTM-Attention, DLSTM, Multi-RNN, DGRU, DA-RNN, LSTM-Attention and LSTM

模型	RMSE	MAE
DLSTM-GA	0.006 765	0.005 720
DLSTM-Attention	0.010 023	0.008 612
DLSTM	0.019 374	0.0126 98
Multi-RNN	0.028 503	0.018 181
DGRU	0.063 732	0.039 145
DA-RNN	0.052 671	0.032 445
LSTM-Attention	0.108 452	0.082 951
LSTM	0.186 159	0.151 927

参考文献 27

1	SHI J P , YANG L T , LIU D . Synchronization and parameter identification of chaotic systems based on quantum-behaved particle swarm optimization[J]. Chinese Journal of Computational Physics, 2019, 36 (2): 236- 244.
2	HAES ALHELOU H , HAMEDANI-GOLSHAN M E , NJENDA T C , et al. A survey on power system blackout and cascading events: Research motivations and challenges[J]. Energies, 2019, 12 (4): 682. DOI
3	ESCRIVÁ-ESCRIVÁ G , ÁLVAREZ-BEL C , ROLDÁN-BLAY C , et al. New artificial neural network prediction method for electrical consumption forecasting based on building end-uses[J]. Energy and Buildings, 2011, 43 (11): 3112- 3119. DOI
4	ZHENG L , LIU Z , SHEN J , et al. Very short-term maximum Lyapunov exponent forecasting tool for distributed photovoltaic output[J]. Applied Energy, 2018, 229 (4): 1128- 1139.
5	LIU C , LIU C , SHANG Y , et al. An adaptive prediction approach based on workload pattern discrimination in the cloud[J]. Journal of Network and Computer Applications, 2017, 80 (6): 35- 44.
6	WANG J , HU J . A robust combination approach for short-term wind speed forecasting and analysis-Combination of the ARIMA (Autoregressive Integrated Moving Average), ELM (Extreme Learning Machine), SVM (Support Vector Machine) and LSSVM (Least Square SVM) forecasts using a GPR (Gaussian Process Regression) model[J]. Energy, 2015, 93 (12): 41- 56.
7	SINGH R , SRIVASTAVA S . Stock prediction using deep learning[J]. Multimedia Tools and Applications, 2017, 76 (18): 18569- 18584. DOI
8	SAGHEER A , KOTB M . Time series forecasting of petroleum production using deep LSTM recurrent networks[J]. Neurocomputing, 2019, 323 (1): 203- 213.
9	QING X , NIU Y . Hourly day-ahead solar irradiance prediction using weather forecasts by LSTM[J]. Energy, 2018, 148, 461- 468. DOI
10	LAHMIRI S , BEKIROS S . Cryptocurrency forecasting with deep learning chaotic neural networks[J]. Chaos, Solitons & Fractals, 2019, 118 (1): 35- 40.
11	ZHOU Y H , LIU H Q , QI P , et al. Forecast of oil production in fractured-vuggy reservoir by using recurrent neural networks[J]. Chinese Journal of Computational Physics, 2018, 35 (6): 668- 674.
12	SHI J P , LI P S , LIU G P . Parameter estimation of chaotic system based on hybrid swarm intelligence optimization algorithm[J]. Chinese Journal of Computational Physics, 2019, 36 (5): 621- 630.
13	HAO S , LEE D H , ZHAO D . Sequence to sequence learning with attention mechanism for short-term passenger flow prediction in large-scale metro system[J]. Transportation Research Part C: Emerging Technologies, 2019, 107, 287- 300. DOI
14	ZHANG X , WANG X , TANG X , et al. Description generation for remote sensing images using attribute attention mechanism[J]. Remote Sensing, 2019, 11 (6): 612. DOI
15	WANG Y, HUANG M, ZHU X, et al. Attention-based LSTM for aspect-level sentiment classification [C]// Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2016(1): 606-615.
16	LI S L , CUI G M , CHEN J X , et al. An improved particle swarm optimization based on diversity monitor and real-time updating strategy[J]. Chinese Journal of Computational Physics, 2017, 34 (3): 344- 354.
17	MUJEEB S , JAVAID N , ILAHI M , et al. Deep long short-term memory: A new price and load forecasting scheme for big data in smart cities[J]. Sustainability, 2019, 11 (4): 987. DOI
18	SAGHEER A , KOTB M . Time series forecasting of petroleum production using deep LSTM recurrent networks[J]. Neurocomputing, 2019, 323 (1): 203- 213.
19	BOUKTIF S , FIAZ A , OUNI A , et al. Optimal deep learning LSTM model for electric load forecasting using feature selection and genetic algorithm: Comparison with machine learning approaches[J]. Energies, 2018, 11 (7): 1636. DOI
20	CHUNG H , SHIN K . Genetic algorithm-optimized long short-term memory network for stock market prediction[J]. Sustainability, 2018, 10 (10): 3765. DOI
21	CHAKRABORTY B. Genetic algorithm with fuzzy fitness function for feature selection[C]//Proceedings of the 2002 IEEE International Symposium on IEEE, 2002, 1(1): 315-319.
22	BIAN S , LU X , XIANG H , et al. Modeling and simulation of AC excited VSCF in wind power systems[J]. Proceedings of the Chinese Society of Electrical Engineering, 2005, 25 (16): 57.
23	BANSAL R C . Three-phase self-excited induction generators: An overview[J]. IEEE Transactions on Energy Conversion, 2005, 20 (2): 292- 299. DOI
24	SAGHEER A , KOTB M . Time series forecasting of petroleum production using deep LSTM recurrent networks[J]. Neurocomputing, 2019, 323 (1): 203- 213.
25	WANG J , YAN J , LI C , et al. Deep heterogeneous GRU model for predictive analytics in smart manufacturing: Application to tool wear prediction[J]. Computers in Industry, 2019, 111 (5): 1- 14.
26	QIN Y, SONG D, CHEN H, et al. A dual-stage attention-based recurrent neural network for time series prediction[C]//Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17), 2017(1): 2627-2633.
27	YANG T , LI B , XUN Q . LSTM-attention-embedding model-based day-ahead prediction of photovoltaic power output using Bayesian optimization[J]. IEEE Access, 2019, 7 (1): 171471- 171484.

[1]	刘凯歌, 韦笃取. 基于WOA-ESN的电机系统混沌振荡预测[J]. 计算物理, 2022, 39(4): 498-504.
[2]	张庆福, 姚军, 黄朝琴, 李阳, 王月英. 裂缝性介质多尺度深度学习模型[J]. 计算物理, 2019, 36(6): 665-672.