KIPS Transactions on Computer and Communication Systems (정보처리학회논문지:컴퓨터 및 통신 시스템)

Korea Information Processing Society (한국정보처리학회)
권호 표지
DOI QR코드

DOI QR Code

Machine Learning Approach for Pattern Analysis of Energy Consumption in Factory

머신러닝 기법을 활용한 공장 에너지 사용량 데이터 분석

  • Received : 2019.02.07
  • Accepted : 2019.03.13
  • Published : 2019.04.30
Download PDF
⟨ Previous Next ⟩

Abstract

This paper describes the pattern analysis for data of the factory energy consumption by using machine learning method. While usual statistical methods or approaches require specific equations to represent the physical characteristics of the plant, machine learning based approach uses historical data and calculate the result effectively. Although rule-based approach calculates energy usage with the physical equations, it is hard to identify the exact equations that represent the factory's characteristics and hidden variables affecting the results. Whereas the machine learning approach is relatively useful to find the relations quickly between the data. The factory has several components directly affecting to the electricity consumption which are machines, light, computers and indoor systems like HVAC (heating, ventilation and air conditioning). The energy loads from those components are generated in real-time and these data can be shown in time-series. The various sensors were installed in the factory to construct the database by collecting the energy usage data from the components. After preliminary statistical analysis for data mining, time-series clustering techniques are applied to extract the energy load pattern. This research can attributes to develop Factory Energy Management System (FEMS).

본 연구에서는 머신 러닝 기법을 활용하여 공장에서 발생하는 에너지 사용량에 대한 데이터 분석 및 패턴 추출에 대해 다룬다. 통계학이나 기존의 방법들은 몇 가지 물리적 특성을 반영하는 수학적 모델을 구축하는 반면, 머신 러닝을 통한 접근방법은 데이터 학습을 통하여 모델의 계수들을 결정하게 된다. 기존의 방법들은 특정한 구조를 갖는 수학적 모델을 구축해야 한다는 어려움이 있으며 과연 데이터의 특징들을 잘 반영하는지에 대한 의문이 존재했다. 그러나 머신 러닝을 통한 방법은 사람이 구축하기 어려운 작업들을 용이하게 구축한다는 장점을 가지고 있기 때문에 데이터 간의 관계를 파악하기에 더 효율적이라는 장점을 가지고 있다. 공장의 에너지 소비에 직접적으로 영향을 끼치는 요소들이 존재하며 이러한 전력 소비는 시간에 따른 데이터로 나타나게 된다. 각 요소들로부터 발생하는 소비 전력을 계측하고 데이터 베이스를 구축하기 위해 각 요소에 센서를 장착하였다. 취득된 데이터에 대해 전처리 과정 및 통계적인 분석을 거친 뒤, 머신 러닝을 통해 패턴을 분석하는 과정을 거쳤다. 이를 통해 공장에서 발생하는 소비 전력 데이터에 대한 패턴 분석을 진행하였다.

Keywords

JBCRIN_2019_v8n4_87_f0001.png 이미지

Fig. 1. Other Electricity Consumption Components

JBCRIN_2019_v8n4_87_f0002.png 이미지

Fig. 2. Energy Consumption Data Example

JBCRIN_2019_v8n4_87_f0003.png 이미지

Fig. 3. Similarity Computation Using DTW[11]

JBCRIN_2019_v8n4_87_f0004.png 이미지

Fig. 4. Power Consumption According to Time

JBCRIN_2019_v8n4_87_f0005.png 이미지

Fig. 5. BIC Score for K-means Clustering

JBCRIN_2019_v8n4_87_f0006.png 이미지

Fig. 6. Clustered Result When k=6

JBCRIN_2019_v8n4_87_f0007.png 이미지

Fig. 7. Normalized Data of All Elements

JBCRIN_2019_v8n4_87_f0008.png 이미지

Fig. 8. Normalized Data of Total Power Consumption

JBCRIN_2019_v8n4_87_f0009.png 이미지

Fig. 9. Weekly Normalized Data of Total Power Consumption

JBCRIN_2019_v8n4_87_f0010.png 이미지

Fig. 10. Correlation Matrix Between Components

Table 1. Power Consuming Components

JBCRIN_2019_v8n4_87_t0001.png 이미지

References

  1. Il-Su Seol, Sun-Woong Kim, and Dong-You Choi, "A Study of the Current Status of Domestic Building Energy Management System and the Correct way for Improvement," Proceedings of KIIT Summer Conference, 2015.
  2. Ji-Young Eum, Soo-Hwan Choi, Si-Sam Park, and Yong-Ki Kim, "Development of Mathmatical Model for the Energy Demand Pattern Analysis of City Buildings," The Korean Institute of Electical Engineers, 2015.
  3. Ki-Ho Kim and Bae Kim, "Sustainable Developement for the City : City Design Initiatives through Greenways," Asia Design Journal, Vol.5, pp.136-165, 2010.
  4. L.P. Lombard, J. Ortiz, and C. Pout, "A review on buildings energy consumption information," Energy and Buildings, Vol.40, pp.394-398, 2008. https://doi.org/10.1016/j.enbuild.2007.03.007
  5. K. Amarasingle, D. Wijayasekara, H. Carey, M. Manic, D. He, and W. Chen, "Artificial Neural Networks based Thermal Energy Storage Control for Buildings," Proc. 41st Annual Conference of the IEEE Industrial Electronics Society, IEEE IECON 2015, Yokohama, Japan, Nov.09-12, 2015.
  6. D. Wijayasekara and M. Manic, "Data-Fusion for Increasing Temporal Resoultion of Building Energy Management System Data," Proc. 41st Annual Conference of the IEEE Industrial Electronics Society, IEEE IECON 2015, Yokohama, Japan, Nov. 09-12, 2015.
  7. Tae-Won Lee, Yong-Ki Kim, and Jae-Sang Seo, "The Major Functions and Using Method of the Building & Energy Management and Information System(BEMIS)," The Society of Air-Conditioning and Refrigerating Engineers of Korea, 2012.
  8. Hyeun Jun Moon, Sung Kwon Jung, and Seung Ho Ryu, "Building Cooling and Heating Energy Consumption Pattern Analysis Based on Building Energy Management System (BEMS) Data Using Machine Learning Techniques," The Society of Air-Conditioning And Refrigerating Engineers of Korea, 2015.
  9. M. Steinbach, G. Karypis, and V. Kumar, "A comparison of document clustering techniques," KDD workshop on text mining, 2000.
  10. Xiaoyue Wang, Abdullah Mueen, Hui Ding, Goce Trajcevski, Peter Scheuermann, and Eamonn Keogh, "Experimental comparison of representation methods and distance measures for time series data," Data in, Knowl, 2013.
  11. J Paparrizos and L Gravano, "Fast and accurate time-series clustering," ACMTranscations on Database System, vol. 42, 2017.