가전제품 전력 사용 분류를 위한 장단기 메모리 기반 비침입 부하 모니터링 기법

경찬욱,선준호,선영규,김진영,Kyeong, Chanuk,Seon, Joonho,Sun, Young-Ghyu,Kim, Jin-Young 한국인터넷방송통신학회 2021 한국인터넷방송통신학회 논문지 Vol.21 No.4

본 논문은 분산자원 집합 거래시장의 활성화와 에너지 관리의 중요성이 증가되면서 에너지 관리 모니터링 기술로서 합산된 전체 전력으로부터 각각의 가전제품의 전력을 찾아내는 비 침입 부하 모니터 기법을 제안한다. 본 논문에서는 데이터 전처리를 통해 각 가전제품들의 power on-off상태가 나오도록 한다. 이러한 데이터를 LSTM을 모델로 사용하여 각 가전제품들의 power on-off 상태를 예측한다. 예측한 상태들을 데이터 후처리를 한 후, 실제 상태들과 비교하여 정확도를 측정한다. 본 논문에서는 전자제품의 개수, 데이터 후처리 방법과 Time step size를 다르게 하여 정확도를 측정하여 비교한다. 전자 제품의 개수가 6개이고, Round함수로 데이터 후처리 방법을 사용하고, Time step size는 6으로 설정하였을 때, 가장 높은 정확도가 나온 것으로 측정되었다. In this paper, we propose a non-intrusive load monitoring(NILM) system which can find the power of each home appliance from the aggregated total power as the activation in the trading market of the distributed resource and the increasing importance of energy management. We transform the amount of appliances' power into a power on-off state by preprocessing. We use LSTM as a model for predicting states based on these data. Accuracy is measured by comparing predicted states with real ones after postprocessing. In this paper, the accuracy is measured with the different number of electronic products, data postprocessing method, and Time step size. When the number of electronic products is 6, the data postprocessing method using the Round function is used, and Time step size is set to 6, the maximum accuracy can be obtained.

가정용 전력 모니터링을 위한 전력신호 분석 알고리즘 개발

박성욱(Sung-Wook Park),왕보현(Bo-Hyeun Wang) 한국지능시스템학회 2011 한국지능시스템학회논문지 Vol.21 No.6

본 연구에서는 가정 내 모든 기기가 연결된 하나의 전력선을 모니터링 하여 그 전력선에 연결된 기기 각각의 젼력 소비 상황을 모니터링 하는 NILM(Non-Intrusive Load Monitoring) 시스템 구축에 필요한 신호 분석 알고리즘에 대한 연구를 수행하였다. 본 연구에서 제안한 신호 분석 알고리즘은 전력선에서 관찰된 여러 기기의 전력 소비 패턴이 혼합된 혼합 전력 패턴을 복수개의 시간 구간으로 분리하고, 연속된 시간 구간 사이의 신호 차이를 구한 후, 이 차이 신호를 분석하여 어떤 장치가 현재 시간 구간에서 동작 중인지를 알아낸다. 이 때 시간 구간을 충분히 작게 하고, 신호 분석에 사용되는 특징들이 독립적이고 additivity특징을 가지도록 선정한다면, 이 차이 신호에는 한 장치의 특징만이 남아 있으므로, 동시에 동작할 수 있는 장치 조합의 수 2<SUP>N</SUP>개가 아닌 장치 N 개에 해당하는 특징만을 이용하여 혼합 신호를 분석할 수 있다. 이를 통하여 장치 개수가 증가하더라도 연산량 역시 산술적으로 증가하는 합리적인 확장성을 확보할 수 있다. 실제 가정에서 각 장치의 데이터 패턴을 채집한 후 이를 인위적으로 조합하여 만든 실험 데이터를 활용하여 제안한 방법을 검증하였다. 검증 결과 4개의 장치가 동시에 동작하고 그 장치의 특징들이 제안한 알고리즘에서 제시한 기준을 만족하는 경우, 비록 제한된 실험이었지만 완벽한 분류 성공률을 보였다. 제안된 알고리즘을 실제 사용하기 위해서는 장치의 수를 증가하고, 시간 구간을 조정하며, 신호 혼합 패턴을 다양하게 한 실증적인 연구가 더 필요하다. 하지만 이 경우 본 연구에서 제안한 기준을 만족하는 특징을 선택한다면, 그렇지 않은 경우에 비하여, 일정 정도 성능이 보장되는 NILM 시스템을 구축할 수 있을 것으로 기대된다. This paper presents an algorithm identifying devices that generate observed mixed signals that are collected at main power-supply line. The proposed algorithm, which is necessary for low cost electric power monitoring system at appliance-level, that is non-intrusive load monitoring system, divides incoming mixed signal into multiple time intervals, calculating difference-signals between consecutive time interval, and identifies which device is operating at the time interval by analysing the difference-signals. Since the features of one device can remain when the time interval is short enough and the features are independent and additive, well-known classification algorithms can be used to classify the difference-signals with features of N individual devices, otherwise 2<SUP>N</SUP> features might be necessary. The proposed algorithm was verified using data mixed in a laboratory with individual devices’s data collected from field. When maximum 4 devices operate or stop sequentially and when features satisfy the requirements of proposed algorithm, the proposed algorithm resulted nearly 100% success rate under the constrained test condition. In order to apply the proposed algorithm in real world, the number devices shall increase, the time interval shall be smaller and the pattern of mixture shall be more diverse. However we can expect, if features used follow guidelines of proposed algorithm, future system could have certain level of performance without the guideline.

Gramian angular field 기반 비간섭 부하 모니터링 환경에서의 다중 상태 가전기기 분류 기법

선준호,선영규,김수현,경찬욱,심이삭,이흥재,김진영,Seon, Joon-Ho,Sun, Young-Ghyu,Kim, Soo-Hyun,Kyeong, Chanuk,Sim, Issac,Lee, Heung-Jae,Kim, Jin-Young 한국인터넷방송통신학회 2021 한국인터넷방송통신학회 논문지 Vol.21 No.3

비간섭 부하 모니터링은 사용자 에너지 소비량의 실시간 모니터링을 통해 가전기기의 사용량 예측 및 분류를 하는 기술로, 최근 에너지 절약의 수단으로 관심이 증가하고 있다. 본 논문에서는 GAF(Gramian angular field) 기반 1차원 시계열 데이터를 2차원 행렬로 변환하는 기법과, 합성곱 신경망(convolutional neural networks)을 결합해 사용자 전력 사용량 데이터로부터 가전기기를 예측하는 시스템을 제안한다. 학습을 위해 공개 가정용 전력 데이터인 REDD(residential energy disaggregation dataset)를 사용하고, GASF(Gramian angular summation field), GADF(Gramian angular difference field)의 분류 정확도를 확인한다. 시뮬레이션 결과, 이중 상태(on/off)를 가지는 가전기기에서 두 모델 모두 97%의 정확도를 보였고, 다중 상태를 가지는 기기에서 GASF는 95%로 GADF보다 3% 높은 정확도를 보임을 확인하였다. 차후 데이터의 량을 증가시키고 모델을 최적화해 정확도와 속도를 개선할 예정이다. Non-intrusive load monitoring is a technology that can be used for predicting and classifying the type of appliances through real-time monitoring of user power consumption, and it has recently got interested as a means of energy-saving. In this paper, we propose a system for classifying appliances from user consumption data by combining GAF(Gramian angular field) technique that can be used for converting one-dimensional data to the two-dimensional matrix with convolutional neural networks. We use REDD(residential energy disaggregation dataset) that is the public appliances power data and confirm the classification accuracy of the GASF(Gramian angular summation field) and GADF(Gramian angular difference field). Simulation results show that both models showed 94% accuracy on appliances with binary-state(on/off) and that GASF showed 93.5% accuracy that is 3% higher than GADF on appliances with multi-state. In later studies, we plan to increase the dataset and optimize the model to improve accuracy and speed.

Load Prorofile Disaggregation Method for Home A Appliances Usinng Active Power Consumption

herie Park 대한전기학회 2013 Journal of Electrical Engineering & Technology Vol.8 No.3

Power metering and monitoring system is the basic element of Smart Grid technology. This paper proposes a new Non-Intrusive Load Monitoring (NILM) method for residential buildings sector using the measured total active power consumptions. Home electric appliances are classified by ON/OFF state model, Multi-state model and Composite model according to their operational characteristics observed by measurements. In order to disaggregate the operation and the power consumption of each model, an algorithm which includes switching function, truth table matrix and matching process is presented. Typical profiles of each appliances and disaggregation results are shown and classified. To improve the accuracy, Time Lagging (TL) algorithm and Permanent-On model (PO) algorithm are additionally proposed. The method is validated as comparing the simulation results to the experimental ones with high accuracy

Load Profile Disaggregation Method for Home Appliances Using Active Power Consumption

Park, Herie The Korean Institute of Electrical Engineers 2013 Journal of Electrical Engineering & Technology Vol.8 No.3

Power metering and monitoring system is a basic element of Smart Grid technology. This paper proposes a new Non-Intrusive Load Monitoring (NILM) method for a residential buildings sector using the measured total active power consumption. Home electrical appliances are classified by ON/OFF state models, Multi-state models, and Composite models according to their operational characteristics observed by experiments. In order to disaggregate the operation and the power consumption of each model, an algorithm which includes a switching function, a truth table matrix, and a matching process is presented. Typical profiles of each appliances and disaggregation results are shown and classified. To improve the accuracy, a Time Lagging (TL) algorithm and a Permanent-On model (PO) algorithm are additionally proposed. The method is validated as comparing the simulation results to the experimental ones with high accuracy.

Load Profile Disaggregation Method for Home Appliances Using Active Power Consumption

Herie Park 대한전기학회 2013 Journal of Electrical Engineering & Technology Vol.8 No.3

Power metering and monitoring system is a basic element of Smart Grid technology. This paper proposes a new Non-Intrusive Load Monitoring (NILM) method for a residential buildings sector using the measured total active power consumption. Home electrical appliances are classified by ON/OFF state models, Multi-state models, and Composite models according to their operational characteristics observed by experiments. In order to disaggregate the operation and the power consumption of each model, an algorithm which includes a switching function, a truth table matrix, and a matching process is presented. Typical profiles of each appliances and disaggregation results are shown and classified. To improve the accuracy, a Time Lagging (TL) algorithm and a Permanent-On model (PO) algorithm are additionally proposed. The method is validated as comparing the simulation results to the experimental ones with high accuracy.

Non-Intrusive Residential Load Monitoring System Using Appliance: Based Energy Disaggregation Models

Paramasivam Mohan Devie,Sundaram Kalyani 대한전기학회 2023 Journal of Electrical Engineering & Technology Vol.18 No.5

With large scale implementation of smart metering technology, the implications demand to account the energy consumption at every node of the system, but the device scalability is of great concern. Energy Disaggregation serves the purpose of finding the appliance level energy consumption from the aggregate energy, which helps to unlock the interactions between the devices through load characterization. This paper proposes a decision tree-based approach for identifying the device operations thereby effectively categorizing the load. A balanced data learning approach is adopted for data processing to eliminate the outliers during training and testing phase of classifier and also improve the classifier performance. The proposed model was evaluated using Reference Energy DisaggregationDataset (REDD) and Retrofit DecisionSupport Tools for UK Homes using Smart Home Technology (REFIT) Dataset. The performance metrics has been obtained for individual appliance using decision tree, naïve bayes and k-nearest neighbor classifiers andanalysed for validation. The proposed disaggregation approach has proven to give promising results in terms of better and accurate detection of appliance operation. The load monitoring system is developed to detect the appliance operation by sensing the voltage, current and power data at defined sampling rate of frequency. Even with large training dataset, the results obtained during testing phase with unseen dataset were viable for further allegations of proposed load model.

Power Load Disaggregation of Households with Solar Panels Based on an Improved Long Short-term Memory Network

JiaXuan Sun,JunNian Wang,Wenxin Yu,ZhenHeng Wang,YangHua Wang 대한전기학회 2020 Journal of Electrical Engineering & Technology Vol.15 No.5

With the increasing application of small distributed renewable energy systems in household power supplies, when a large number of distributed renewable energy power generation systems are connected to the power grid, the time-varying output power of small solar energy, wind turbines, etc. Disaggregation and analysis of regional household electricity and renewable energy power supply systems connected to household electricity will help grid companies to conduct power dispatch management. This paper employed a two-way two-layer Long Short-term Memory deep learning network with improved input form to perform non-intrusive load disaggregation on household power with solar panels, which can monitor the load status of household electrical appliances and the output power of solar power generation system in real time. The power situation provides a decision basis for optimizing the response value of household energy demand and improving the demand of the power system from the response management level. The combined dataset from UK-DALE and kaggle’solar panel power generation data is adopted to train and test the proposed improved Long Short-term Memory network. The test results show that the proposed algorithm is applied to the household electric load disaggregation with solar panels, with high accuracy and reliability.

개인 맞춤형 에너지 피드백 시스템을 위한 딥러닝 기반의 다수 가전기기 유효전력 분해

김임규,김현철,신상용 한국태양에너지학회 2022 한국태양에너지학회 논문집 Vol.42 No.1

Energy consumption feedback with an appliance-level feedback system can reduce consumption by a maximum of 12%. In this study, we proposed a data acquisition and training framework for configuring deep learning based on Non-Intrusive Load Monitoring (NILM) for a personalized and appliance-level energy consumption feedback system. To construct a training dataset, an aggregation of active power data from four types of home appliances (refrigerator, induction, TV, washing machine) was performed for approximately three weeks. LSTNet was applied to extract and recognize the features of active power data and the state of each home appliance. With an accuracy metric of more than 90% of the disaggregation result, the applicability of the appliance-level active power feedback system was verified.

NIALM 기반의 스마트 홈 안전관리시스템에 관한 연구

정한상(Han-Sang Jeong),성경상(Kyung-Sang Sung),오해석(Hae-Seok Oh) 한국산학기술학회 2017 한국산학기술학회논문지 Vol.18 No.8

기존의 전기 에너지 및 관리에 필요한 정보를 취득하기 위해 적용하였던 계측방식은 공간적인 문제와 시스템의 크기로 인해 신규 건축물 또는 교체가 가능한 지역에만 한정되었었다. 이러한 전기 부하관리 방법은 취약지구나 기존 기 구축되어 있는 가정 또는 사무실의 에너지 및 안전관리에 적용하기에 문제가 있다. 즉, 모든 분기마다 계측 모듈을 설치하는 문제로 인해, 그 시스템의 크기가 너무 크고, 계측모듈을 설치하더라도 부하의 종류를 인식하지 못해 효율적 관리가 이루어지지 않으며 많은 비용이 발생하고 있어서 보급에 어려움을 겪고 있다. 특히, 한국의 전통 재래시장 및 낙후된 시설 등에는 적용하기 매우 어려운 실정이다. 본 논문에서는 이러한 문제점을 개선하기 위해 NIALM 기술과 아크 감지기술을 적용하여, 정상적인 아크발생에 대한 NIALM의 적용가능성을 검증하고자 한다. 또한, 검증 결과를 기반으로 재래시장 및 기존 가정내 안전관리장치에 적용할 수 있는 효율적인 전기안전관리가 가능한 새로운 개념의 스마트 홈 안전관리 시스템을 제안한다. 제안된 시스템은 기존에 적용된 안전관리시스템과의 비교 성능 시험을 진행하였고, 기존 시스템 대비 40%의 공간내에서 기존 시스템에서는 불가능하였던 4가지 부하에 대한 부하인식을 95%이상 달성하였고, 또한 기존 시스템과 같은 아크 감지기능을 확인하였다. Due to spatial problems and system size,conventional measurement methods used to acquire the information needed for existing electrical energy and management have been limited to new buildings or areas where replacement is possible. This electric load management method is problematic whenapplying it to energy and safety management of vulnerable areas or existing homes or offices. The problem withinstalling ameasurement module in every branch is that the system is too large.Even if the measurement module is installed, the type of load is not recognized, and efficient management is not performed. In particular, it is very difficult to apply it to traditional markets and backward facilities in Korea. In this paper, we apply NIALM technology and arc detection technology to solve these problems and verify the feasibility of NIALM for normal arc generation. Also, based on the verification results, we propose a new smart home safety management system that can effectively manage electrical safety and that can be applied to conventional market and existing home safety management systems. The proposed system conductsa comparative performance test with anexisting safety management system. In addition, it achieves 95% or more load recognition for four loads, which is impossible in 40% of the existing systems, and the arc detection function was confirmed.