Accurate prediction of energy consumption is helpful for further value mining and data fusion. In order to achieve this purpose, this paper proposes an energy consumption prediction method based on SARIMAX (seasonal autoregressive integrated moving average with exogenous) and LSTM (Long Short-Term Memory) hybrid model. First, we import energy consumption data and weather environment factors such as temperature, moderateness, wind speed, etc., compare the corresponding relationships through data visualization, and use K-means to construct weather clusters as an exogenous variable. Then we import holiday information, and construct a holiday indicator as the second exogenous variables. Furtherly, we adjust the model according to the seasonal trend and use grid search to select the optimal parameters of the SARIMAX model. Finally, we mix the SARIMAX model with LSTM to optimize the prediction model to make predictions and compare the results. Experimental analysis shows that the SARIMAX-LSTM hybrid model can integrate weather factors, holiday factors, and seasonal factors to make high-precision predictions of energy consumption.

Yang Zhou. Regional energy consumption prediction based on SARIMAX-LSTM model. Academic Journal of Computing & Information Science (2021), Vol. 4, Issue 3: 41-51. https://doi.org/10.25236/AJCIS.2021.040307.

[1] Fu Zhixin, Li Xiaoyi, Yuan Yue. Discussion on Key Technologies of Ubiquitous Power Internet of Things [J]. Electric Power Construction, 2019, 40(5): 1-12.

[2] Tang Hu, Chen Ailun, Cui Hao et al. Overview of Urban Energy Internet Research [J]. Hydropower and Pumped Storage, 2020, 6(1): 13-16.

[3] Ju Ping, Zhou Xiaoxin, Chen Weijiang et al. Overview of "Smart Grid+" Research [J]. Electric Power Automation Equipment, 2018, 38(5): 2-11.

[4] Gong Jianya, Zhang Xiang, Xiang Longgang et al. Progress and application of integrated perception and intelligent decision-making in smart cities [J]. Journal of Surveying and Mapping, 2019, 48(12): 1482-1497.

[5] Fan Lijun. Building energy consumption prediction and building energy efficiency analysis based on multiple linear regression model [J]. Natural Science Journal of Xiangtan University, 2016, 38(1): 123-126.

[6] Lu Chengju, Li Meng, Liao Baoliang et al. Energy consumption prediction of diesel passenger cars based on ARIMA model [J]. Shandong Transportation Science and Technology, 2019, (2): 120-123.

[7] Guokun Lai1;Wei-Cheng Chang1;Yiming Yang1;Hanxiao Liu1. Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks [arXiv] [J]. arXiv, 2017, (0)

[8] Ji Xinge, Li Hui, Liu Sijia et al. Short-term photovoltaic power prediction based on MIE-LSTM [J]. Power System Protection and Control, 2020, 48(7): 50-57.

[9] XUE Mengfan, HAN Lei, PENG Dongliang. A combined algorithm of K-means and MTRL for multi-class classification [J]. Journal of Systems Engineering and Electronics, 2019, (5): 875-885.

[10] Zheng Yingxin. Application of cluster analysis based on the elbow rule in data mining in the optimization design of primary and middle school students’ travel routes [J]. Electronics World, 2017, (9): 146.

[11] You Pin. Research on the Distribution of DF (ADF) Test Parameters OLS Estimator[J]. Statistics and Decision, 2009, (20): 20-22.

[12] Chen Jingjie, Meng Qi. Prediction of unsaturated airport energy consumption based on two-step decomposition method and SARIMA [J]. Computer Applications and Software, 2019, 36(4): 46-50, 78.

[13] Zhang Yan. The Application of AIC Criterion of Multiple Multiple Linear Regression Model [J]. Modern Trade Industry, 2016, 37(11): 227-228.