由於電力擁有無法長時間儲存的特性，實時供應與消耗的平衡對於電力公司和輸電網路至關重要。在這些情境中，短期負載預測在決策過程中扮演著關鍵角色，涉及經濟調度、購買和銷售策略等，並且與電網的穩定性和供電成本密切相關。此外，短期負載預測在工業公司、太陽能發電和智慧建築等領域也廣泛應用。因此，開發準確的電力負載預測模型成為一個極為重要的需求。
有許多方法可以用於預測電力負載，包含數學模型、相似日。然而，由於電力系統的非線性之特性，近年來，隨著科技的進步，機器學習技術逐漸被用於解決負載預測問題，從而得到更準確的預測。因此，提供給機器學習模型的特徵成為影響模型準確性的關鍵因素。本研究提出了一種新的方法，利用股市中短期預測的技術分析指標作為短期負載預測的特徵。這種方法的概念是基於股市和電力系統都是非線性系統，因此技術分析指標可能包含了有用的負載資訊，可以提高負載預測的性能。此外，雖然將國定假日和普通假日作為特徵在其他研究中已經得到了不錯的性能，但尚缺乏研究解釋這些特徵所帶來的改善。因此，本研究進行了一系列比較，評估了包括節假日作為特徵在內的預測性能。再者，隨著時間的推移，電力負載預測模型的資料以及超參數需要不斷更新。為了解決這個問題，本研究提出了一種使用Tree-Structured Parzen Estimator (TPE)演算法的方法，來自動化超參數調整的過程，從而減少手動調整參數的繁瑣工作。這種方法可以更有效地優化模型的性能，並隨著時間推移保持其預測能力。
在本研究中，目標是增強短期負載預測模型，並且不限制其應用範圍。最終，我們在兩個數據集（國家級和城市級電力負載資料集）上使用基於樹的方法，如隨機森林和XGBoost，評估了原始方法與本研究所提出的方法。實驗結果顯示，在這兩個資料集上分別達到了大約96%與84%的R-Squared，證明了本研究所取得的改進。結果表明，我們的方法在短期負載預測方面具有很高的預測準確性，並且能夠應用於不同的電力負載預測場景中。

Due to the inherent inability of electricity to be stored for extended periods, maintaining a real-time balance between supply and consumption is crucial for power companies and transmission networks. In these contexts, short-term load forecasting plays a pivotal role in decision-making processes, involving economic dispatch, buying and selling strategies, and is closely related to grid stability and electricity costs. Moreover, short-term load forecasting is also widely applied in industries, solar power generation, and smart buildings. Thus, developing an accurate electric load forecasting model has become an imperative need.
While various methods, including mathematical models and similar days, can be employed for power load prediction, the nonlinear nature of the power system has led to the increasing use of machine learning techniques in recent years to address the forecasting problem, resulting in more accurate predictions. Consequently, the features provided to the machine learning model become a critical factor influencing its accuracy. This study proposes a novel approach by employing technical analysis indicators from short-term stock market forecasting as features for short-term load prediction. The reason behind this approach is that both stock markets and power systems are nonlinear systems; therefore, technical analysis indicators might encapsulate valuable load information, enhancing forecasting performance. Moreover, although incorporating national and regular holidays as features has demonstrated decent performance in other studies, there's a lack of exploration into the improvements these features bring. Hence, this study conducted a series of comparisons, assessing the predictive performance inclusive of holiday features. Furthermore, as time progresses, the data and hyperparameters of the power load forecasting model need constant updating. To address this issue, this study introduces a method using the Tree-Structured Parzen Estimator (TPE) algorithm to automate the hyperparameter tuning process, thereby reducing the tedious task of manual adjustments. This method can optimize the model's performance more efficiently and maintain its predictive capability over time.
In this study, the aim is to enhance short-term load forecasting models without limiting their scope of application. Ultimately, we employed tree-based models, such as Random Forests and XGBoost, to evaluate the original methods and those introduced in this study on two datasets (national and city-level power load datasets). Experimental results show that approximately 96% and 84% R-Squared values were achieved on these two datasets respectively, highlighting the improvements accomplished by our research. The results demonstrate that our method possesses high predictive accuracy in short-term load forecasting and can be applied across various power load forecasting scenarios.

[1] I. K. Nti, M. Teimeh, O. Nyarko-Boateng, and A. F. Adekoya, “Electricity load forecasting: a systematic review,” Journal of Electrical Systems and Information Technology, vol. 7, no. 1, Sep. 2020, doi: 10.1186/s43067- 020-00021-8.
[2] J. Morley, K. Widdicks, and M. Hazas, “Digitalisation, energy and data demand: The impact of Internet traffic on overall and peak electricity consumption,” Energy Research & Social Science, vol. 38, pp. 128–137, Apr. 2018, doi: 10.1016/j.erss.2018.01.018.
[3] “2020 AWS 5G x AIoT決勝未來智慧創新論壇,” Amazon Web Services, Inc. https://aws.amazon.com/tw/events/aiot2020/
[4] “PJM Learning Center - Load Forecasting,” PJM Learning Center - Load Forecasting. https://learn.pjm.com/three-priorities/planning-for-the-future/load-forecasting
[5] ESCN, http://escn.com.cn/news/show-693678.html
[6] “Enhanced Load Forecasting - ENTSO-E,” Enhanced Load Forecasting - ENTSO-E. https://www.entsoe.eu/Technopedia/techsheets/enhanced-load-forecasting
[7] “Electricity interconnection targets,” Energy. https://energy.ec.europa.eu/topics/infrastructure/electricity-interconnection-targets_en
[8] H. Zheng, J. Yuan, and L. Chen, "Short-Term Load Forecasting Using EMD-LSTM Neural Networks with a Xgboost Algorithm for Feature Importance Evaluation," Energies, vol. 10, no. 8, article 1168, Aug. 2017. doi: 10.3390/en10081168.
[9] S. H. Rafi, Nahid-Al-Masood, S. R. Deeba and E. Hossain, "A Short-Term Load Forecasting Method Using Integrated CNN and LSTM Network," in IEEE Access, vol. 9, pp. 32436-32448, 2021, doi: 10.1109/ACCESS.2021.3060654.
[10] S. S. Subbiah and J. Chinnappan, “Short-Term Load Forecasting Using Random Forest with Entropy-Based Feature Selection,” Artificial Intelligence and Technologies, pp. 73–80, Dec. 2021, doi: 10.1007/978-981-16-6448-9_8.
[11] W. Kong, Z. Y. Dong, D. J. Hill, F. Luo and Y. Xu, "Short-Term Residential Load Forecasting Based on Resident Behaviour Learning," in IEEE Transactions on Power Systems, vol. 33, no. 1, pp. 1087-1088, Jan. 2018, doi: 10.1109/TPWRS.2017.2688178.
[12] A. A. Muzumdar, C. N. Modi, M. G. M and C. Vyjayanthi, "Designing a Robust and Accurate Model for Consumer-Centric Short-Term Load Forecasting in Microgrid Environment," in IEEE Systems Journal, vol. 16, no. 2, pp. 2448-2459, June 2022, doi: 10.1109/JSYST.2021.3073493.
[13] J Bergstra, R Bardenet, Y Bengio et al., "Algorithms for hyper-parameter optimization[J]", Advances in neural information processing systems, vol. 24, 2011.
[14] Y. Xie, Y. Ueda, and M. Sugiyama, "A Two-Stage Short-Term Load Forecasting Method Using Long Short-Term Memory and Multilayer Perceptron," Energies, vol. 14, no. 18, article 5873, Sep. 2021. doi: 10.3390/en14185873.
[15] P. Singh and P. Dwivedi, "Integration of new evolutionary approach with artificial neural network for solving short term load forecast problem," Applied Energy, vol. 217, pp. 537-549, May 2018. doi: 10.1016/j.apenergy.2018.02.131.
[16] J. Li et al., "A Novel Hybrid Short-Term Load Forecasting Method of Smart Grid Using MLR and LSTM Neural Network," in IEEE Transactions on Industrial Informatics, vol. 17, no. 4, pp. 2443-2452, April 2021, doi: 10.1109/TII.2020.3000184.
[17] M. Lee et al., “A Comparative Study of Forecasting Electricity Consumption Using Machine Learning Models,” Mathematics, vol. 10, no. 8, p. 1329, Apr. 2022, doi: 10.3390/math10081329.
[18] R. Trivedi, S. Patra and S. Khadem, "A Data-Driven Short-Term PV Generation and Load Forecasting Approach for Microgrid Applications," in IEEE Journal of Emerging and Selected Topics in Industrial Electronics, vol. 3, no. 4, pp. 911-919, Oct. 2022, doi: 10.1109/JESTIE.2022.3179961.
[19] M. Jayashankara, P. Shah, A. Sharma, P. Chanak and S. K. Singh, "A Novel Approach for Short-Term Energy Forecasting in Smart Buildings," in IEEE Sensors Journal, vol. 23, no. 5, pp. 5307-5314, 1 March1, 2023, doi: 10.1109/JSEN.2023.3237876.
[20] T. F. Megahed, S. M. Abdelkader, and A. Zakaria, “Energy Management in Zero-Energy Building Using Neural Network Predictive Control,” IEEE Internet of Things Journal, vol. 6, no. 3, pp. 5336–5344, Jun. 2019, doi: 10.1109/jiot.2019.2900558.
[21] G. Gross and F. D. Galiana, "Short-term load forecasting," in Proceedings of the IEEE, vol. 75, no. 12, pp. 1558-1573, Dec. 1987, doi: 10.1109/PROC.1987.13927.
[22] H. J. Sadaei, P. C. de Lima e Silva, F. G. Guimarães, and M. H. Lee, “Short-term load forecasting by using a combined method of convolutional neural networks and fuzzy time series,” Energy, vol. 175, pp. 365–377, May 2019, doi: 10.1016/j.energy.2019.03.081.
[23] M. C. Ruiz-Abellón, L. A. Fernández-Jiménez, A. Guillamón, A. Falces, A. García-Garre, and A. Gabaldón, "Integration of Demand Response and Short-Term Forecasting for the Management of Prosumers’ Demand and Generation," Energies, vol. 13, no. 1, article 11, Dec. 2019. doi: 10.3390/en13010011.
[24] H. Zheng, J. Yuan, and L. Chen, "Short-Term Load Forecasting Using EMD-LSTM Neural Networks with a Xgboost Algorithm for Feature Importance Evaluation," Energies, vol. 10, no. 8, article 1168, Aug. 2017. doi: 10.3390/en10081168.
[25] S. H. Rafi, Nahid-Al-Masood, S. R. Deeba and E. Hossain, "A Short-Term Load Forecasting Method Using Integrated CNN and LSTM Network," in IEEE Access, vol. 9, pp. 32436-32448, 2021, doi: 10.1109/ACCESS.2021.3060654.
[26] G. Dudek, "Short-Term Load Forecasting Using Random Forests," in Advances in Intelligent Systems and Computing, vol. 323, 2014.
[27] G. Dudek, “A Comprehensive Study of Random Forest for Short-Term Load Forecasting,” MDPI, Oct. 13, 2022. https://www.mdpi.com/1996-1073/15/20/7547
[28] S. S. Subbiah and J. Chinnappan, “Short-Term Load Forecasting Using Random Forest with Entropy-Based Feature Selection,” Artificial Intelligence and Technologies, pp. 73–80, Dec. 2021, doi: 10.1007/978-981-16-6448-9_8.
[29] I. K. Nti, M. Teimeh, O. Nyarko-Boateng, and A. F. Adekoya, "Electricity load forecasting: a systematic review," Journal of Electrical Systems and Information Technology, vol. 7, Article no. 13, Sep. 2020.
[30] I. Nti, A.-A. Samuel, and A. Michael, "Predicting Monthly Electricity Demand Using Soft-Computing Technique," 2019.
[31] eeeguide, “Types of Forecasting Methods - EEEGUIDE.COM,” EEEGUIDE.COM, Jan. 10, 2017. https://www.eeeguide.com/types-of-forecasting-methods/
[32] S. Bouktif, A. Fiaz, A. Ouni, and M. A. Serhani, “Optimal Deep Learning LSTM Model for Electric Load Forecasting using Feature Selection and Genetic Algorithm: Comparison with Machine Learning Approaches †,” MDPI, Jun. 22, 2018.
[33] P. Stavast, "Prediction of energy consumption using historical data and twitter," The University of Groningen, Groningen, 2014. [Online]. Available: https://www.cs.rug.nl/~aiellom/tesi/stavast.pdf
[34] “PJM Learning Center - Load Forecasting,” PJM Learning Center - Load Forecasting. https://learn.pjm.com/three-priorities/planning-for-the-future/load-forecasting
[35] J. R. Quinlan, “Induction of decision trees,” Machine Learning, vol. 1, no. 1, pp. 81–106, Mar. 1986, doi: 10.1007/bf00116251.
[36] L. Breiman, "Random Forests," Machine Learning, vol. 45, no. 1, pp. 5-32, Oct. 2001.
[37] J. Ali, R. Khan, N. Ahmad, and I. Maqsood, "Random Forests and Decision Trees," International Journal of Computer Science Issues (IJCSI), vol. 9, Sep. 2012.
[38] T. Chen and C. Guestrin, "XGBoost: A Scalable Tree Boosting System," in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016.
[39] J. Friedman, "Greedy function approximation: A gradient boosting machine," Annals of Statistics, pp. 1189-1232, 1999.
[40] hyperopt, “GitHub - hyperopt/hyperopt-sklearn: Hyper-parameter optimization for sklearn,” GitHub, Dec. 15, 2022. https://github.com/hyperopt/hyperopt-sklearn
[41] B. Komer, J. Bergstra, and C. Eliasmith, "Hyperopt-Sklearn: Automatic Hyperparameter Configuration for Scikit-Learn," in Proc. of the 13th Python in Science Conf. (SciPy), 2014.
[42] "Specific National Considerations," ENTSO-E. [Online]. Available: https://eepublicdownloads.entsoe.eu/clean-documents/Publications/Statistics/Specific_national_considerations.pdf. Accessed on: July 11, 2023.
[43] “Daily Electricity Demand Forecast-Machine Learning,” Daily Electricity Demand Forecast-Machine Learning | Kaggle. https://www.kaggle.com/code/manualrg/daily-electricity-demand-forecast-machine-learning
[44] A. D. Roy and A. Yeafi, "Implementation of Encoder-Decoder based Long Short-Term Memory Network for Short-Term Electrical Load Forecasting," 2022 4th International Conference on Sustainable Technologies for Industry 4.0 (STI), Dhaka, Bangladesh, 2022, pp. 1-6, doi: 10.1109/STI56238.2022.10103285.
[45] M. Lekshmi and K. N. A. Subramanya, "Short-Term Load Forecasting of 400kV Grid Substation Using R-Tool and Study of Influence of Ambient Temperature on the Forecasted Load," 2019 Second International Conference on Advanced Computational and Communication Paradigms (ICACCP), Gangtok, India, 2019, pp. 1-5, doi: 10.1109/ICACCP.2019.8883005.
[46] ”holidays - PyPI." [Online]. Available: https://pypi.org/project/holidays/. Accessed on: June 5, 2023.
[47] S. Pattnaik, "Hourly Load Data," Kaggle, 2021. [Online]. Available: https://www.kaggle.com/datasets/pattnaiksatyajit/hourly-load-data. Accessed on: June 5, 2023.
[48] MathsGee, “What is the difference between the sliding window and expanding window approaches to forecasting?,” MathsGee AI Prompt Directory, Oct. 19, 2020. https://mathsgee.com/21209/difference-between-sliding-expanding-approaches-forecasting
[49] K. Chen, K. Chen, Q. Wang, Z. He, J. Hu and J. He, "Short-Term Load Forecasting With Deep Residual Networks," in IEEE Transactions on Smart Grid, vol. 10, no. 4, pp. 3943-3952, July 2019, doi: 10.1109/TSG.2018.2844307.
[50] T. Kang, D. Y. Lim, H. Tayara, and K. T. Chong, "Forecasting of Power Demands Using Deep Learning," Applied Sciences, vol. 10, no. 20, p. 7241, Oct. 2020, doi: 10.3390/app10207241.