- Research & Development
- Energy Systems
FOR IMMEDIATE RELEASE No. 3245
TOKYO, January 29, 2019 - Mitsubishi Electric Corporation (TOKYO: 6503) announced today that it has developed new technology that allows the estimated power consumption of individual home appliances to be extrapolated from the overall power consumption of each household. The new solution, the result of joint research with Tohoku Electric Power Co., Inc., makes use of Mitsubishi Electric's pioneering AI technology Maisart®* to estimate power consumption to a high degree of accuracy without the need to install new measuring instruments.
- * Mitsubishi Electric's AI creates the State-of-the- ART in Technology
- ** See Tohoku Electric Power Co., Inc.'s news release at http://www.tohoku-epco.co.jp/news/normal/1197475_1049.html
Outline of visualization of power consumption details
Smart meters that measure household electricity consumption at frequent regular intervals are becoming increasingly common. Existing smart meters only measure the overall power consumption of the whole house, but there is a growing need to know the power consumption of individual home appliances. Technically, the installation of a current sensor on the home's power distribution board would allow the power consumption of each appliance to be monitored, but the cost of installing such sensors can be prohibitive. Mitsubishi Electric Corporation has therefore developed this new "Technology to Visualize Power Consumption" solution, which deploys artificial intelligence to extrapolate the power consumption of individual home appliances from the power consumption of the whole house to a high degree of accuracy.
- 1)Leverages AI technology to estimate individual home appliance power consumption without the need for additional measuring devices
- Artificial intelligence is used to extrapolate the power consumption of each home appliance from the power consumption data of the whole house, as measured by a smart meter.
- No need to install new measuring instruments since existing smart meters are used.
- The amount of data collected and stored is just one percent of that required by conventional estimation methods.
Using AI, typical electricity usage patterns are extracted from data such as family composition and the attributes of the home appliances. The power consumption of each home appliance is then extrapolated from the power consumption data of the whole house measured by a smart meter. Traditional methods measure the power consumption of individual home appliances at intervals of 10 seconds or less using a current sensor or other measuring device. However, this new technology utilizes the existing data captured by smart meters, so there is no need to install new measuring instruments. As a result, the amount of stored data can be reduced to one percent or less of that required by existing methods, which in turn reduces the amount of calculation required to provide estimates.
- 2)Typical patterns allow power consumption for each home appliance to be estimated to a high degree of accuracy
- The AI functionality performs a three-stage clustering based on the power consumption of the whole house and each home appliance measured in advance within monitored houses, as well as information on the composition of the family and the home appliances they own.
- A typical pattern is created using AI, which corrects fluctuations caused by variations in daily activity time and groups together houses with similar power consumption characteristics.
- The AI functionality automatically selects the most similar typical pattern, reducing errors by applying actual values and achieving a high level of accuracy.
The AI functionality performs three-stage clustering based on the pre-measured power consumption of the whole house and each home appliance, and on attributes such as family composition and the number and type of home appliances. Houses with similar electricity usage are automatically grouped and representative values for each group are created as a typical pattern. In addition, by absorbing minor time fluctuations in activities that vary depending on the day and the family, such as waking, cooking and the time at which family members return home, the AI calculates the correlation between the typical pattern and the measured data to estimate power consumption more accurately. Since the AI function automatically selects the most similar typical pattern, the discrepancy with actual values is reduced, achieving a high level of estimation accuracy.
- Research & Development
May 28, 2020
Mar 26, 2020
Mar 25, 2020
- Energy Systems
Mar 16, 2020
Oct 10, 2019
Jul 08, 2019