[论文解读] Energy Consumption Forecasting for Smart Meters

Anshul Bansal, Susheel Kaushik Rompikuntla+3 more|arXiv (Cornell University)|Dec 18, 2015

Energy Load and Power Forecasting参考文献 9被引用 23

一句话总结

本文提出一种基于提升决策树回归的机器学习方法，利用智能电表数据进行家庭用电量预测，通过特征工程实现时间序列预测。实验基于英国DECC和新加坡EMA的数据，展示了预测准确性的提升，并可根据用户特定的用电模式实现个性化电力计划。

ABSTRACT

Earth, water, air, food, shelter and energy are essential factors required for human being to survive on the planet. Among this energy plays a key role in our day to day living including giving lighting, cooling and heating of shelter, preparation of food. Due to this interdependency, energy, specifically electricity, production and distribution became a high tech industry. Unlike other industries, the key differentiator of electricity industry is the product itself. It can be produced but cannot be stored for future; production and consumption happen almost in near real-time. This particular peculiarity of the industry is the key driver for Machine Learning and Data Science based innovations in this industry. There is always a gap between the demand and supply in the electricity market across the globe. To fill the gap and improve the service efficiency through providing necessary supply to the market, commercial as well as federal electricity companies employ forecasting techniques to predict the future demand and try to meet the demand and provide curtailment guidelines to optimise the electricity consumption/demand. In this paper the authors examine the application of Machine Learning algorithms, specifically Boosted Decision Tree Regression, to the modelling and forecasting of energy consumption for smart meters. The data used for this exercise is obtained from DECC data website. Along with this data, the methodology has been tested in Smart Meter data obtained from EMA Singapore. This paper focuses on feature engineering for time series forecasting using regression algorithms and deriving a methodology to create personalised electricity plans offers for household users based on usage history.

研究动机与目标

• 通过准确预测以弥合电力供需之间的持续差距。
• 基于家庭用电历史数据，开发一种数据驱动的个性化电力计划方法论。
• 评估机器学习，特别是提升决策树回归，在时间序列用电量预测中的有效性。
• 通过预测分析提升电力市场中的服务效率和需求优化。
• 整合针对智能电表数据中时间模式的特征工程技术。

提出的方法

• 作者应用提升决策树回归模型，基于历史智能电表数据对用电量进行建模。
• 通过特征工程从时间序列数据中提取时间模式，如日周期、周周期和季节性趋势。
• 模型在两个数据集上进行训练和验证：英国DECC数据和新加坡EMA智能电表数据。
• 通过分析用户特定的用电模式并预测未来用电量，生成个性化电力计划。
• 使用标准回归指标评估模型性能，尽管摘要中未报告具体数值。
• 该方法整合了来自商业和联邦电力供应商的真实世界数据，以确保实际适用性。

实验结果

研究问题

• RQ1 提升决策树回归在使用智能电表数据预测家庭用电量方面，其预测精度如何？
• RQ2 特征工程在提升时间序列用电数据预测性能方面发挥何种作用？
• RQ3 能否通过机器学习从历史用电模式中有效推导出个性化电力计划？
• RQ4 在不同地理和数据采集背景下（英国与新加坡）的结果有何差异？
• RQ5 预测建模在多大程度上可减少电力市场中的供需缺口？

主要发现

• 所提出的模型通过针对时间模式的有效特征工程，成功实现了更高精度的用电量预测。
• 可基于个体家庭的用电历史生成个性化电力计划，从而实现需求优化。
• 该方法在不同数据集（包括英国DECC和新加坡EMA智能电表数据）中表现出良好的适应性。
• 提升决策树回归在电力领域的时间序列预测中表现有效，尤其适用于近实时需求预测。
• 本研究凸显了机器学习在提升服务效率和支撑电力市场电网稳定方面的潜力。
• 该方法为商业和联邦电力供应商实施需求侧管理策略提供了实用框架。