預測週期性家用電量對電力供應者除了可以更準確的了解家戶電能使用，更可以進行電能的事前規劃。本研究提出一特徵工程之訓練框架，在進行資料探索性分析後，增加時間序列特徵並使用Lasso演算法進行特徵選擇，以提升模型訓練的準確度。並選用長短期記憶神經網路（Long Short-Term Memory，LSTM）、卷積+長短期記憶神經網路(Convolutional neural network Long - Short-Term Memory，CNN-LSTM) 與門控循環神經網路 (Gated Recurrent Unit，GRU)三種常見的時間序列及衍生模型進行實驗。本研究利用一具週期性規律之家庭用電量公開資料集進行模型訓練及驗證。實驗結果發現使用本研究所提出之特徵選擇功能，LSTM準確度可提升約14.88%，模型訓練時間減少9%；CNN-LSTM準確度可提升約8.21%，模型訓練時間減少27.93%；GRU準確度可提升約11.17%，模型訓練時間減少21.47%。利用成對統計檢定(t-test)可發現本研究所提出之特徵選擇方法在LSTM模型及GRU模型的準確度上具有顯著的提升。對所有的方法而言，成對統計檢定，確認加入特徵選擇方法可讓模型訓練時間顯著的減少。

Predicting the periodic household electricity can help the electricity supplier to understand the household electricity usage more accurately, and also to plan the electricity usage in advance. In this study, a feature engineering training framework is proposed. After exploratory data analysis, time series features are added and Lasso algorithm is used for feature selection to improve the accuracy of model training. The Long Short-Term Memory (LSTM), Convolutional neural network Long - Short-Term Memory (CNN-LSTM), and Gated Recurrent Unit (GRU) Three common time series and derivative models are experimented. In this study, a public dataset of household electricity consumption with a periodic pattern is used for model training and validation. The experimental results show that the accuracy of LSTM can be improved by about 14.88% and the model training time can be reduced by 9%; the accuracy of CNN-LSTM can be improved by about 8.21% and the model training time can be reduced by 27.93%; the accuracy of GRU can be improved by about 11.17% and the model training time can be reduced by 21.47% using the feature selection function proposed in this study. Using pairwise statistical testing (t-test), it is found that the proposed feature selection method has a significant improvement in the accuracy of LSTM and GRU models. For all the methods, the Paired-t test statistical testing confirms that adding the feature selection method can significantly reduce the model training time.

[1] A. Baratta, A. Cimino, F. Longo, V. Solina, and S. Verteramo, "The Impact of ESG Practices in Industry with a Focus on Carbon Emissions: Insights and Future Perspectives," vol. 15, no. 8, p. 6685, 2023.
[2] D. Mhlanga, "Artificial Intelligence and Machine Learning for Energy Consumption and Production in Emerging Markets: A Review," vol. 16, no. 2, p. 745, 2023.
[3] M. Smith, K. C. Hargroves, P. Stasinopoulos, R. Stephens, C. Desha, and S. Hargroves, "Energy Transformed: Sustainable energy solutions for climate change mitigation," 2007.
[4] W. Henisz, T. Koller, and R. Nuttall, "Five ways that ESG creates value," 2019.
[5] I. Zumente, J. J. J. o. O. I. T. Bistrova, Market,, and Complexity, "ESG importance for long-term shareholder value creation: Literature vs. practice," vol. 7, no. 2, p. 127, 2021.
[6] D. Daugaard and A. J. S. Ding, "Global drivers for ESG performance: The body of knowledge," vol. 14, no. 4, p. 2322, 2022.
[7] M. J. V. a. W. S. I. Eckhart and E. Reporting, "Financial regulations and ESG investing: Looking back and forward," pp. 211-228, 2020.
[8] F. Moslein and K. E. J. E. C. L. Sorensen, "The Commission's Action Plan for Financing Sustainable Growth and Its Corporate Governance Implications," vol. 15, p. 221, 2018.
[9] S. Bell, The Rise of the People's Bank of China. Harvard University Press, 2013.
[10] B. J. T. J. o. W. M. Cornell, "ESG investing: Conceptual issues," vol. 23, no. 3, pp. 61-69, 2020.
[11] Y. Cho, S. Kim, J. You, H. Moon, H. J. E. J. o. T. Sung, and Development, "Application of ESG measures for gender diversity and equality at the organizational level in a Korean context," vol. 45, no. 4/5, pp. 346-365, 2021.
[12] B. C. Olsen, K. Awuah-Offei, and D. J. R. P. Bumblauskas, "Setting materiality thresholds for ESG disclosures: A case study of US mine safety disclosures," vol. 70, p. 101914, 2021.
[13] M. Singhania, N. J. J. o. S. F. Saini, and Investment, "Institutional framework of ESG disclosures: comparative analysis of developed and developing countries," vol. 13, no. 1, pp. 516-559, 2023.
[14] J. Wang, Y. Ma, L. Zhang, R. X. Gao, and D. J. J. o. m. s. Wu, "Deep learning for smart manufacturing: Methods and applications," vol. 48, pp. 144-156, 2018.
[15] T. Ince, S. Kiranyaz, L. Eren, M. Askar, and M. J. I. T. o. I. E. Gabbouj, "Real-time motor fault detection by 1-D convolutional neural networks," vol. 63, no. 11, pp. 7067-7075, 2016.
[16] H. Salehinejad, S. Sankar, J. Barfett, E. Colak, and S. J. a. p. a. Valaee, "Recent advances in recurrent neural networks," 2017.
[17] S. Hochreiter and J. J. A. i. n. i. p. s. Schmidhuber, "LSTM can solve hard long time lag problems," vol. 9, 1996.
[18] J. Chung, C. Gulcehre, K. Cho, and Y. J. a. p. a. Bengio, "Empirical evaluation of gated recurrent neural networks on sequence modeling," 2014.
[19] A. Azadeh, S. Ghaderi, S. J. E. C. Sohrabkhani, and management, "Annual electricity consumption forecasting by neural network in high energy consuming industrial sectors," vol. 49, no. 8, pp. 2272-2278, 2008.
[20] R. F. Berriel, A. T. Lopes, A. Rodrigues, F. M. Varejao, and T. Oliveira-Santos, "Monthly energy consumption forecast: A deep learning approach," in 2017 International Joint Conference on Neural Networks (IJCNN), 2017, pp. 4283-4290: IEEE.
[21] T.-Y. Kim and S.-B. Cho, "Predicting the household power consumption using CNN-LSTM hybrid networks," in Intelligent Data Engineering and Automated Learning–IDEAL 2018: 19th International Conference, Madrid, Spain, November 21–23, 2018, Proceedings, Part I 19, 2018, pp. 481-490: Springer.
[22] J. Hyeon, H. Lee, B. Ko, and H. J. J. T. J. o. E. Choi, "Deep learning‐based household electric energy consumption forecasting," vol. 2020, no. 13, pp. 639-642, 2020.
[23] T. Le, M. T. Vo, B. Vo, E. Hwang, S. Rho, and S. W. J. A. S. Baik, "Improving electric energy consumption prediction using CNN and Bi-LSTM," vol. 9, no. 20, p. 4237, 2019.
[24] K. Venkat, T. Gautam, M. Yadav, and M. Singh, "An xgboost ensemble model for residential load forecasting," in Proceedings of International Conference on Intelligent Computing, Information and Control Systems: ICICCS 2020, 2021, pp. 321-334: Springer.
[25] J. J. E. R. Qin, "Experimental and analysis on household electronic power consumption," vol. 8, pp. 705-709, 2022.
[26] L. Cascone, S. Sadiq, S. Ullah, S. Mirjalili, H. U. R. Siddiqui, and M. J. B. D. R. Umer, "Predicting Household Electric Power Consumption Using Multi-step Time Series with Convolutional LSTM," vol. 31, p. 100360, 2023.
[27] S. Kumar, L. Hussain, S. Banarjee, and M. Reza, "Energy load forecasting using deep learning approach-LSTM and GRU in spark cluster," in 2018 fifth international conference on emerging applications of information technology (EAIT), 2018, pp. 1-4: IEEE.
[28] X. Tang, Y. Dai, T. Wang, Y. J. I. G. Chen, Transmission, and Distribution, "Short‐term power load forecasting based on multi‐layer bidirectional recurrent neural network," vol. 13, no. 17, pp. 3847-3854, 2019.
[29] L. Suganthi, A. A. J. R. Samuel, and s. e. reviews, "Energy models for demand forecasting—A review," vol. 16, no. 2, pp. 1223-1240, 2012.
[30] R. Jain, T. Damoulas, and C. Kontokosta, "Towards data-driven energy consumption forecasting of multi-family residential buildings: feature selection via the lasso," in Computing in Civil and Building Engineering (2014), 2014, pp. 1675-1682.
[31] Y. He, Y. Qin, S. Wang, X. Wang, and C. J. A. e. Wang, "Electricity consumption probability density forecasting method based on LASSO-Quantile Regression Neural Network," vol. 233, pp. 565-575, 2019.
[32] F. Lazzari et al., "User behaviour models to forecast electricity consumption of residential customers based on smart metering data," vol. 8, pp. 3680-3691, 2022.
[33] A. Loginov, M. I. Heywood, and G. Wilson, "Benchmarking a coevolutionary streaming classifier under the individual household electric power consumption dataset," in 2016 International Joint Conference on Neural Networks (IJCNN), 2016, pp. 2834-2841: IEEE.