基于ARIMA模型的离散制造业能源消耗预测

Energy Consumption Prediction Based on ARIMA Model in the Discrete Manufacturing Industry

摘要: 离散制造业作为国民经济的重要支柱，其能源消耗问题关乎企业成本、环保与可持续发展。针对该行业生产多样化、定制化导致的能耗不稳定性问题，本研究旨在构建合适的精准能源预测模型，助力企业优化资源配置、提升效率与竞争力。基于北京地区典型离散制造企业2023年4月至2024年11月共20个月的月度真实运行数据，经严格数据清洗与特征工程，筛选出关键产品装配数量与工作天数作为核心输入特征。采用时间序列分析方法，构建ARIMA (2, 1, 0)能耗预测模型。实验结果表明，该模型在测试集上预测性能良好，验证了产能特征与工作天数特征组合在离散制造业能耗预测中的有效性及应用价值。研究表明ARIMA模型能够有效支持离散制造企业能耗的精准预测，为节能降耗与可持续发展决策提供量化依据。未来研究将融合大数据与人工智能技术，探索具备更高精度与泛化能力的先进预测模型。

Abstract: Discrete manufacturing industry acts as an important support for national economy and its energy consumption matters enterprise cost, environmental protection, and sustainable development. Aiming at the problem of energy consumption instability caused by the industrial production diversification and customization, a proper and precise energy prediction model was established in this paper to help enterprises optimize their resource allocation and enhance their efficiency and competitiveness. Based on the monthly practical operating data from a typical discrete manufacturing enterprise in Beijing from April 2023 to November 2024, the critical production assemblies and working days were screened out as the core input features after the strict data cleansing and featurization. Then based on the time series analysis method, an ARIMA (2, 1, 0) energy consumption prediction model was built. The experimental results revealed that the model performed well in prediction in terms of the test set, and verified the effectiveness and application value of the feature combination of capacity characteristics and working days in predicting the energy consumption of discrete manufacturing industry. The research proved that the ARIMA model worked efficiently in the precise prediction of energy consumption in the discrete manufacturing enterprises. It would provide a quantitative basis for any decision made on energy saving, cost reducing, and sustainable development. In the future, big data and AI technology will be fused in the research to explore the advanced prediction model with higher precision and generalization capability.