Household Power Disaggregation Using HMM and External Factors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power consumption monitoring systems face challenges in accurately identifying individual device energy consumption due to high costs associated with smart plugs and limitations in providing detailed user behavior information with low-resolution data, while high-resolution data requires additional installations and varied collection formats, and Intrusive Load Monitoring has inaccuracies in device identification.
Innovation Solution
A power consumption disaggregation system using a Hidden Markov Model (HMM) that considers external factors, which extracts characteristic variables from high-resolution data to a reference resolution, converting data formats, and applies these variables to disaggregate total household power consumption into individual states using a database, extraction, and analysis units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If smart plugs are installed to monitor each energy consuming device, then device-level power consumption information can be obtained, but the cost increases significantly
Solution Approach 1:
The patent extracts device-level power consumption information from aggregate power data by analyzing transient responses and characteristic variables. Instead of installing smart plugs on each device, the system extracts individual device signatures from the total power consumption signal at the service entrance, thereby obtaining device-level information without the associated hardware cost.
Solution Approach 2:
The patent introduces an intermediary processing system that acts as a mediator between the aggregate power data and device-level information. The extraction unit and analysis unit serve as intermediaries that process the total power consumption signal to reveal individual device characteristics, avoiding the need for direct device-level measurement hardware.
2Adaptability or versatility
If low-resolution power data (15-minute unit) is used for NILM, then scalability is improved, but detailed user behavior information cannot be provided
Solution Approach 1:
The patent applies preliminary action by pre-processing power data at multiple resolutions and extracting characteristic variables before analysis. The system prepares data at different time resolutions in advance, allowing it to maintain scalability while preserving the ability to extract detailed behavioral patterns when needed.
Solution Approach 2:
The patent adds another dimension by processing power data at multiple time resolutions simultaneously. Instead of choosing between single-resolution approaches, the system analyzes data at both 15-minute intervals and higher resolutions, extracting characteristic variables from each dimension to provide both scalability and detailed behavior information.
3Measurement precision
If high-resolution power data is used to provide detailed user behavior information, then measurement precision is improved, but additional measurement devices and varied collection formats are required
Solution Approach 1:
The patent applies universality by designing a processing system that can handle multiple data resolutions and formats through a unified approach. The extraction unit and HMM-based analysis system are configured to process both low-resolution (15-minute) and high-resolution data using the same methodology, eliminating the need for separate processing pipelines and reducing installation complexity.
Solution Approach 2:
The patent utilizes parameter changes by adjusting the time resolution parameter in the data processing pipeline. The system can switch between analyzing data at 15-minute intervals or higher resolutions by changing the temporal sampling parameter, allowing detailed behavior analysis without requiring permanent high-resolution infrastructure installation.
4Measurement precision
If Intrusive Load Monitoring is used for device identification, then device-level monitoring is achieved, but identification accuracy is poor
Solution Approach 1:
The patent replaces the mechanical approach of Intrusive Load Monitoring (which requires physical connection to devices) with a signal processing approach. Instead of mechanically interfacing with each device, the system uses mathematical extraction of transient responses and characteristic variables from aggregate power data to identify devices, achieving better accuracy without intrusive installation.
Data Source
AI summary
A power consumption disaggregation system according to an embodiment of the present invention comprises: a DB for receiving power data from a smart meter installed in a house, receiving data according to external factors affecting the power data from an external server, and storing the data in time series; an extraction unit for extracting a characteristic variable of a reference resolution time interval from the power data and the data according to external factors stored in time series; and an analysis unit for disaggregating total power consumption in the house by inferring a hidden power consumption state by inputting the characteristic variable into a power consumption disaggregation model based on a Hidden Markov Model (HMM) considering dependency of external factors.