مدلی ترکیبی در پیش‌بینی داده محور مصرف انرژی در صنعت فولاد: رویکردی بر پایه LSTM و بهینه‌سازی هایپرپارامترها با الگوریتم Optuna

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانشجوی دکتری مهندسی صنایع، گروه مهندسی صنایع، دانشگاه یزد، یزد، ایران

2 دانشیار، گروه مهندسی صنایع، دانشگاه یزد، یزد، ایران

3 استاد، گروه مهندسی صنایع، دانشگاه یزد، یزد، ایران

10.22034/abmir.2025.23569.1157

چکیده

در دنیای مدرن، انرژی الکتریکی یکی از اساسی‌ترین نیازهای جوامع، نقش حیاتی در بخش‌های صنعتی، اقتصادی و اجتماعی ایفا می‌کند. با توجه به عدم امکان ذخیره‌سازی در مقیاس بزرگ و نوسانات عرضه و تقاضا، پیش‌بینی دقیق مصرف بار الکتریکی ضروری است. مدل‌های سنتی سری‌های زمانی، اغلب در مواجهه با الگوهای غیرخطی و پیچیده داده‌های صنعتی دچار چالش می‌شوند. این پژوهش یک چارچوب ترکیبی نوین برای پیش‌بینی دقیق سری زمانی مصرف انرژی الکتریکی ارائه می‌دهد. رویکرد پیشنهادی بر پایه شبکه عصبی حافظه بلند-کوتاه مدت (LSTM) است که برای به حداکثر رساندن عملکرد مدل، هایپرپارامترهای شبکه LSTM با استفاده از الگوریتم بهینه‌سازی خودکار Optuna  بهینه‌سازی شده‌اند. داده‌های مورد استفاده شامل سری‌های زمانی مصرف برق کارخانه فولاد بافت به همراه متغیرهای برون‌زا مانند میزان تولید فولاد، مدت زمان توقف تولید، شرایط آب و هوایی و تقویمی هستند. نتایج ارزیابی نشان می‌دهند که مدل ترکیبی LSTM-Optuna  با ضریب تعیین ((R2) 89/0و درصد خطای مطلق میانگین (MAPE) 46/18٪، نه تنها از مدل پایه با ضریب تعیین 85/0 و درصد خطای مطلق میانگین با مقدار 44/20٪ عملکرد بهتری نشان می‌دهد، بلکه از سایر روش‌های یادگیری ماشین مانند شبکه‌های عصبی بازگشتی (RNN) و ماشین بردار پشتیبان ((SVM.

کلیدواژه‌ها

عنوان مقاله [English]

A Hybrid Model for Data-Driven Energy Consumption Forecasting in the Steel Industry: An Approach Based on LSTM and Hyperparameter Optimization with the Optuna Algorithm

نویسندگان [English]

  • Leila Zolqadr 1
  • Majid Shakhsi-Niaei 2
  • Yahia Zare Mehrjerdid 3
  • Mohammad Mehdi Lotfi 3
1 Ph.D. Student in Industrial Engineering, Department of Industrial Engineering, Yazd University, Yazd, Iran
2 Associate Professor, Department of Industrial Engineering, Yazd University, Yazd, Iran
3 Professor, Department of Industrial Engineering, Yazd University, Yazd, Iran
چکیده [English]

In the modern world, electrical energy, as one of the most fundamental needs of societies, plays a vital role in the industrial, economic, and social sectors. Given the impossibility of large-scale storage and fluctuations in supply and demand, accurate forecasting of electrical load consumption is essential. Traditional time-series models often struggle to handle the nonlinear and complex patterns of industrial data. This research presents a novel hybrid framework for accurately forecasting electrical energy consumption time series. The proposed approach is based on a Long Short-Term Memory (LSTM) neural network; to maximize the model's performance, the LSTM network's hyperparameters have been optimized using the Optuna automatic optimization algorithm. The data used includes time series of electricity consumption from the Baft steel factory, along with exogenous variables such as steel production volume, production downtime, and weather and calendar conditions. The evaluation results indicate that the hybrid LSTM-Optuna model, with a coefficient of determination (R²) of 0.89 and a Mean Absolute Percentage Error (MAPE) of 18.46%, not only demonstrates better performance than the baseline model (with an R² of 0.85 and a MAPE of 20.44%) but also outperforms other machine learning methods such as Recurrent Neural Networks (RNN) and both simple and optimized Support Vector Machines (SVM).

کلیدواژه‌ها [English]

  • Data-Driven Modeling
  • STM-Optuna
  • Walk-Forward Validation
  • Electrical Energy Consumption
  • Steel Industry
  • RNN
  • SVM

مراجع

[1]     S. Ghasaee and R. Ravanmehr, “Short-Term Prediction of Electrical Load Consumption Using Deep Neural Networks, CNN, and LSTM,” J. Qual. Product. Iran Electr. Ind., vol. 10, no. 1, pp. 35–51, 2021 [In Persian].
[2]     G. Zhang and M. Qi, “Neural Network Forecasting for Seasonal and Trend Time Series,” Eur. J. Oper. Res., vol. 160, no. 2, pp. 501-514, 2005, doi: 10.1016/j.ejor.2003.08.037.
[3]     H. Pashaei and M. Dehkharghani, “Predictive Modeling of Energy Consumption in the Steel Industry Using CatBoost Regression: A Data-Driven Approach for Sustainable Energy Management,” J. Clean. Prod., vol. 401, p. 136706, 2023, doi: 10.1016/j.jclepro.2023.136706.
[4]     A.Sherstinsky, “Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network,” Physica D: Nonlinear Phenomena, vol. 404, p. 132306, 2020, doi: 10.1016/j.physd.2020.132306.
[5]     S. Arslan, “A Hybrid Forecasting Model Using LSTM and Prophet for Energy Consumption with Decomposition of Time Series Data,” PeerJ Comput. Sci., vol. 8, p. e1001, 2022, doi: 10.7717/peerj-cs.1001.
[6]     R. Zhu, N. Li, Y. Duan, G. Li, G. Liu, F. Qu, et al., “Well-Production Forecasting Using Machine Learning with Feature Selection and Automatic Hyperparameter Optimization,” Energies, vol. 18, no. 1, p. 99, dic. 2024, doi: 10.3390/en18010099.
[7]     F. He, J. Zhou, Z.-k. Feng, G. Liu, and Y. Yang, “A Hybrid Short-Term Load Forecasting Model Based on Variational Mode Decomposition and Long Short-Term Memory Networks, Considering Relevant Factors with a Bayesian Optimization Algorithm,” Appl. Energy, vol. 237, pp. 108–117, Mar. 2019, doi: 10.1016/j.apenergy.2019.01.055.
[8]     T. Gao, Y. Sun, and C. Li, “LSTM Network Hyperparameter Optimization for Stock Price Prediction Using the Optuna  Framework,” J. Phys.: Conf. Ser., vol. 1785, no. 1, p. 012061, 2021, doi: 10.1088/1742-6596/1785/1/012061.
[9]     D. Bunn and E. Farmer, “Review of Short-Term Forecasting Methods in the Electric Power Industry,” in Comparative Models for Electrical Load Forecasting, D. Bunn and E. Farmer, Eds. Berlin: Springer, 1985, pp. 13-30.
[10] Moghram and S. Rahman, “Analysis and Evaluation of Five Short-Term Load Forecasting Techniques,” IEEE Trans. Power Syst., vol. 4, no. 4, pp. 1484-1491, Nov. 1989.
[11] K. Amasyali and N. M. El-Gohary, “A Review of Data-Driven Building Energy Consumption Prediction Studies,” Renew. Sustain. Energy Rev., vol. 81, pp. 1192-1205, Jan. 2018.
[12] M. H. L. Lee, Y. C. Ser, G. Selvachandran, P. H. Thong, L. Cuong, L. H. Son, et al., “A Comparative Study of Forecasting Electricity Consumption Using Machine Learning Models,” Mathematics, vol. 10, no. 1329, pp. 1-23, 2022.
[13] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction. New York: Springer, 2017.
[14] S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Comput., vol. 9, no. 8, pp. 1735-1780, Nov. 1997.
[15] Y. Zheng, J. Zhang, and Y. Wang, “Short-Term Load Forecasting Based on a Long Short-Term Memory Neural Network,” J. Mod. Power Syst. Clean Energy, vol. 5, no. 2, pp. 296-302, Mar. 2017.
[16] S. Pei, H. Qin, L. Yao, Y. Liu, C. Wang, and J. Zhou, “Multi-Step Ahead Short-Term Load Forecasting Using Hybrid Feature Selection and Improved Long Short-Term Memory Network,” Energies, vol. 13, no. 16, p. 4121, Aug. 2020, doi: 10.3390/en13164121.
[17] M. Zohaki Raht and H. Sadeghi Saghadol, “Modeling and Forecasting of Iran's Short-Term Electricity Consumption Using Neural Network and TPE Algorithm,” Journal of Energy Economics Studies, vol. 20, no. 83, pp. 159-182, 2024 [In Persian].
[18] J. Zhu, Y. Wang, C. Lai, and X. Zhou, “Deep Learning-Based Energy Consumption Prediction Model for Green Industrial Parks,” Appl. Artif. Intell., vol. 39, no. 1, p. 2462375, 2021.
[19] L. Shen, W. Chen, and J. Kwok, “Deep Learning for Short-Term Energy Consumption Forecasting of Smart Buildings Considering Microgrid and External Factors,” Mathematics, vol. 10, no. 1329, pp. 1-23, 2022.
[20] S. Sayadinejad, A. Esmailzadeh Maghari, and M. R. Rostami, “Presenting a Bitcoin Return Prediction Model Using a Hybrid Deep Learning Signal Decomposition Algorithm (CEEMD-DL),” Quarterly Journal of Financial Economics, vol. 17, no. 62, pp. 217-238, 2022 [In Persian].
[21] J. Kim, H. Kim, H. Kim, D. Lee, and S. Yoon, “A Comprehensive Survey of Deep Learning for Time Series Forecasting: Architectural Diversity and Open Challenges,” Artif. Intell. Rev., vol. 58, p. 216, Apr. 2025.
[22] N. E. Huang, Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, et al., “The Empirical Mode Decomposition and the Hilbert Spectrum for Nonlinear and Non-Stationary Time Series Analysis,” Proc. R. Soc. Lond. A, vol. 454, no. 1971, pp. 903-995, Mar. 1998.
[23] K. Dragomiretskiy and D. Zosso, “Variational Mode Decomposition,” IEEE Trans. Signal Process., vol. 62, no. 3, pp. 531-544, Feb. 2014, doi: 10.1109/TSP.2013.2288675.
[24] Y. Ruan, G. Wang, H. Meng, and F. Qian, “A Hybrid Model for Power Consumption Forecasting Using VMD-Based Long Short-Term Memory Neural Network,” Front. Energy Res., vol. 9, p. 772508, Feb. 2022, doi: 10.3389/fenrg.2021.772508.
[25] I. O. Ekundayo, “Optuna  optimization-based CNN-LSTM model for predicting electric power energy consumption,” M.S. thesis, School of Comput., Nat. College of Ireland, Dublin, 2020.
[26] C. Gharai, S. K. Mohapatra, S. Parida, C. Dora, R. K. Mohanta, and S. Chakravarty, “Optimized Deep Learning Model for Power Consumption Prediction Using Hyperparameter Tuning Techniques,” in Int. Conf. Intell. Cloud Comput. (ICoICC), Bhubaneswar, India, 2025, pp. 1-6, doi: 10.1109/ICoICC64033.2025.11052111.
[27] R. Uddin, M. R. Hazari, S. Ahmad, C. A. Hossain, M. S. Rahman Zishan, and A. Ahmed, “Short-Term Load Forecasting Using Deep Learning Algorithms with Hyperparameter Optimization,” in 4th Int. Conf. Robot., Electr. Signal Process. Techn. (ICREST), Dhaka, Bangladesh, 2025, pp. 366-371, doi: 10.1109/ICREST63960.2025.10914446.
[28] S. Bouktif, R. F'Haim, and M. Lazaar, “A Novel Hybrid Model for Short-Term Load Forecasting Using LSTM and Metaheuristic Optimization Algorithms,” Energy, vol. 196, p. 117042, Apr. 2020.
[29] J. Brownlee, Deep Learning for Time Series Forecasting: Theory to Practice. San Juan, PR: Machine Learning Mastery, 2018.
[30] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA: MIT Press, 2016.
[31] J. Zhao, G. Nie, and Y. Wen, “Monthly Precipitation Prediction in Luoyang City Based on EEMD-LSTM-ARIMA Model,” Water Sci. Technol., vol. 87, no. 1, pp. 318–335, Jan. 2023, doi: 10.2166/wst.2022.425.
[32] Y. Ren, Z. Yan, D. Liu, J. Hou, and W. Zhou, “Optimized EWT-Seq2Seq-LSTM with Attention Mechanism for Insulator Fault Prediction,” Energies, vol. 15, no. 2, p. 528, Jan. 2022, doi: 10.3390/en15020528.
[33] A. Alghamdi and A. Desuqi, “Predicting Electricity Consumption Using Machine Learning Techniques,” Int. J. Adv. Comput. Sci. Appl., vol. 11, no. 11, pp. 384-391, 2020.
[34] M. Alimohammadi Ardakani, M. H. Karimi-Zarchi, and D. Shishebori, “A Hybrid Forecasting Model for Accurate Prediction of Building Energy Consumption Based on Multi-Stage Decomposition and Optimized Deep Learning,” Energy, vol. 305, p. 126955, 2024.
[35] R. J. Hyndman and G. Athanasopoulos. (2021). Forecasting: Principles and Practice (3rd ed.) [Online]. Available: https://otexts.com/fpp3/
[36] B. G. M. and G. N. Pillai, “Hyperparameter Optimization of Long Short Term Memory Models for Interpretable Electrical Fault Classification,” IEEE Access, vol. 11, pp. 123688-123704, Nov. 2023, doi: 10.1109/ACCESS.2023.3330056.
پژوهش های نظری و کاربردی هوش ماشینی
دوره 3، شماره 2
پاییز و زمستان 1404
آذر 1404
صفحه 89-105

فایل ها

هم رسانی

ارجاع به این مقاله

آمار