1. Introduction
Smart meter applications have evolved significantly, offering multifaceted benefits beyond conventional energy monitoring. Initially designed to facilitate more efficient energy use, smart meters now play a pivotal role in integrating with Internet of Things (IoTs) devices and systems in many research projects [
1,
2,
3], thereby enabling innovative home automation solutions. Furthermore, smart meters serve as foundational components of the Internet of Things (IoTs) ecosystem, seamlessly integrating with other IoT devices and systems to enable innovative home automation solutions. By leveraging the data collected from smart meters, IoT devices can optimize energy usage, automate tasks, and enhance the overall comfort and convenience for homeowners [
1,
4,
5]. This integration opens up possibilities for intelligent energy management systems that dynamically respond to changing energy demands and environmental conditions, paving the way for a more sustainable and connected future [
6,
7]. With real-time data collection and analysis capabilities, smart meters empower users to make informed decisions about their energy consumption patterns, leading to improved energy efficiency and cost savings [
8,
9,
10].
Non-Intrusive Appliance Load Monitoring (NIALM), initially proposed in [
11], utilizes supervised learning algorithms to disaggregate total household electricity consumption into individual appliance usage patterns without requiring additional sensors or intrusive monitoring devices [
12,
13,
14]. While NIALM offers valuable insights into household energy consumption patterns by analyzing the distinctive electrical signatures of various appliances, it raises significant privacy concerns as it involves the detailed monitoring of individuals’ activities within their homes [
15,
16,
17,
18]. The high-resolution nature of NIALM can reveal sensitive information about users, prompting the need for robust privacy safeguards and data protection measures.
In Japan, smart meter-based monitoring systems for older adults have become integral components of home energy management strategies, prioritizing energy conservation, electrical safety, and the well-being of elderly individuals living independently. Japan has one of the fastest aging populations globally, making it an ideal context for studying innovative solutions to support the well-being of older adults living independently. The country faces unique challenges related to an aging society, such as increasing healthcare costs and a shrinking workforce, driving the need for effective monitoring systems to ensure the safety and health of older individuals. With the deregulation of the electricity industry in 2016, Japanese power companies have introduced tailored home care services to address market competition and societal aging. These systems, deployed by various electric power companies, utilize low-resolution smart meter data collected at a frequency of 30 min to analyze the changes in lifestyle patterns directly from electricity usage. Initiatives, such as Tokyo Electric Power Company’s (TEPCO), Tokyo, Japan “Anshin Plan for Long-Distance Families”, Kyushu Electric Power Company’s (Kyushu, Japan) “Kyuden Anshin Support” plan, Chubu Electric Power Company’s (Chubu, Japan) “Monitoring Assistance Service”, and Kansai Electric Power Company’s (Kansai, Japan) “Happy Protector” service, utilize smart meter data to monitor the changes in electricity consumption patterns, detect abnormal usage, establish daily life rhythms, and issue notifications for deviations in routine. Through unsupervised learning techniques, they establish indicators, such as electricity thresholds, routine models, and upper and lower limits, based on electricity usage over the last 7, 14, and 30 days. Caregivers receive notifications when daily electricity usage indicators deviate from the norm, enhancing their ability to respond to older people’s routine changes, which might be associated with a possible decline in health [
17]. However, this approach relies entirely on power consumption data and does not map to appliance usage, raising privacy concerns. The increasing integration of smart meter technology in home energy management systems represents a significant advancement in how we can support older adults living independently. As Japan faces a rapidly aging population, there is a pressing need to develop non-intrusive, scalable solutions that ensure the safety and well-being of elderly people, while maximizing their ability to live independently. Smart meters, which primarily collect low-resolution power consumption data, offer a unique opportunity to monitor daily routines and detect any abnormal activities without using more invasive methods. This study aims to explore the effectiveness of these systems in real-world settings, evaluating their potential to assist caregivers by providing timely alerts about significant changes in electricity usage patterns. This data-driven approach not only promises to enhance quality of care, but also ensures timely interventions, ultimately supporting the health and safety of older adults. By focusing on the capabilities of smart meter technology, we address critical challenges, such as increasing healthcare costs and the shrinking workforce, associated with societal aging. Detecting the deviations in electricity usage patterns among older adults living independently is highly important for caregivers and family members. Changes in energy consumption can signal shifts in routine, lifestyle, or health status. For instance, sudden increases or decreases might suggest alterations in daily activities, mobility issues, or the onset of health problems [
19]. By monitoring these patterns, caregivers can intervene and address long-term health concerns. Therefore, leveraging smart meter technology to track electricity usage offers a proactive and non-intrusive approach to safeguarding the well-being of older adults, supporting the need for efficient remote monitoring and caregiving in aging populations [
20].
This study leverages smart meter data from the Taiwan Power Company (TPC), Taipei, Taiwan. The TPC Smart Meter Initiative entails the deployment of advanced metering infrastructure across 1.2 million households, offering a comprehensive platform for detailed energy consumption analysis. With a data frequency of 15 min, the smart meters collect and transmit electricity usage data, facilitating the monitoring and analysis of energy consumption patterns. This large-scale initiative represents a concerted effort to enhance lifestyle monitoring and energy management, leveraging the capabilities of smart meter technology to promote efficiency and sustainability in households across Taiwan.
This study applies Taiwan Power Company’s (TPC) smart electricity meter data to the daily routine analysis of older adults. Like the service models of various power companies in Japan, this research aims to develop a lifestyle monitoring system using low-resolution smart meter data for older adults living independently at home. Power consumption data are acquired every 15 min using the TPC smart meter. The data are mapped to appliance usage (0/1), which often relates to older adult’s instrumental activities of daily living (IADL). The indices of “activity” and “regularity” of appliance usage based on comparison with the pattern of the last 28 days are calculated. A prototype dashboard is developed to display the users’ indices of activity and regularity.
This paper is structured as follows:
Section 1 provides an introduction to the research.
Section 2 details the methodology, including smart meter data extraction and routine pattern evaluation.
Section 3 presents the results.
Section 4 presents the discussion, and
Section 5 provides the conclusions and future work.
4. Discussion
This study investigates how smart meter usage data can be leveraged to monitor the lifestyles of older adults living independently. This study converts energy consumption data into a low-resolution usage/non-usage format to ensure residents’ privacy, while still providing valuable insights into their daily routines and activities. Furthermore, this study explores methods to send these data to caregivers and family members by developing a prototype dashboard, enabling them to monitor the well-being and activity levels of older adults remotely. The dashboard displays the activity and regularity of the user, which are used as measures of the older adults’ well-being. This approach facilitates intervention and support when necessary, contributing to enhanced care and support for independent living [
28,
29]. Caregivers can benefit from the information derived from smart meter usage data gaining insights into daily activities and routines related to electricity usage. This enables caregivers to detect any deviations or irregularities in behavior that may indicate potential health issues or changes in well-being [
30]. Moreover, by having access to objective and quantitative data on the well-being of older adults, caregivers can make more informed decisions and tailor their support strategies accordingly, leading to improved overall care and quality of life [
31].
The active scores derived from smart meter data serve as quantitative measures of the household activity level. Higher active scores indicate days with increased energy usage, likely corresponding to the time when occupants are engaged in various tasks or activities. Conversely, low active scores suggest days with reduced energy consumption, potentially indicating more periods of inactivity, which might be a sign of poor health or decreased mobility. By analyzing these scores over time, caregivers and healthcare providers can gain insights into the daily routines of elderly individuals [
32,
33] and identify any deviations from their usual patterns. Regularity assessments based on smart meter data help identify the consistency and predictability of daily routines. Consistent energy usage patterns indicate stable lifestyles with predictable activity cycles, which can be reassuring indicators of overall well-being. On the other hand, irregularities or fluctuations in energy usage may signal changes in health status, behavioral patterns, or environmental factors. By detecting deviations from the established routines, caregivers can proactively intervene to address emerging issues or provide necessary support [
34,
35,
36]. The interpretation of activity levels and regularity assessments enables the detection of subtle changes that may signify alterations in health or behavior among elderly individuals. For instance, a sudden decrease in activity levels or disruptions in regular patterns could be early indicators of declining health, mobility issues, or cognitive impairment [
37]. By continuously monitoring these metrics, caregivers can promptly identify potential long-term health concerns in older adults, which may otherwise lead to institutionalization. This enables them to initiate appropriate interventions, such as medical evaluations or lifestyle adjustments. Armed with this information, caregivers can take proactive measures to address these concerns. They can initiate timely interventions, such as scheduling medical evaluations, adjusting medication regimens, or implementing lifestyle modifications, to support the overall health and well-being of older adults. Additionally, caregivers can use the data to tailor support services and assistance according to the specific needs and preferences of the individuals they care for, thereby enhancing the quality of care provided. Furthermore, the ability to remotely monitor energy usage and activity levels through smart meter data offers caregivers greater flexibility and peace of mind. They can track the well-being of older adults from a distance, ensuring that they remain safe and supported even when the caregivers are not physically present. This remote monitoring capability fosters a sense of security for both caregivers and older adults, promoting independent living while ensuring that assistance is readily available when needed.
Household power usage patterns are expected to have small fluctuations during the week, signaling consistent usage behavior by older adults; however, typically, changes are expected during weekends and holidays.
Figure 10 illustrates the active score and correlation coefficient (CC) of a user from Monday to Sunday, revealing a notable increase in power usage during the weekend, leading to higher active scores and lower CC values. This trend can be attributed to several factors commonly seen in the lifestyles of older adults. During the week, older adults may adhere to more structured routines, with fewer activities or outings, resulting in lower power usage and higher correlation coefficients as their daily patterns remain consistent. However, they may engage in more leisure activities and social interactions on weekends, including visiting family members, leading to increased power usage and deviations from their regular routines, reflected in higher active scores and lower correlation coefficients. These fluctuations in power usage underscore the importance of comprehensively monitoring older adults’ lifestyles, considering both weekdays and weekends, to capture variations and ensure timely interventions or support when needed. By understanding these patterns, caregivers and healthcare providers can gain insights into older adults’ activities and well-being, allowing tailored interventions and improved support for independent living.
In comparing the accuracy of identifying appliance usage with the background power threshold achieved in this study (94.8%) to those reported in related literature, it is evident that the results are quite competitive. For example, the study on “Effective Load Pattern Classification by Processing the Smart Meter Data” [
38] does not specify the exact accuracy rates, but emphasizes the effectiveness of using smart meter data for pattern recognition, which is fundamental in establishing the base load. Similarly, research presented in [
39] highlights a methodology that supports high accuracy in load monitoring and appliance identification. This study achieved a notable precision, but did not quantify it in the same direct percentage terms as the current study. Therefore, while direct numerical comparisons are challenging due to the variability in reporting metrics, the approach used in the current study, yielding a 94.8% accuracy rate, stands as highly effective when viewed against the backdrop of methodologies and outcomes discussed in these authoritative sources. This suggests that low-resolution data and the specific methods for threshold calculation in this study are valid and robust for practical applications in lifestyle monitoring based on smart meter data.
While our study has shown promising results, several limitations must be acknowledged. Firstly, our methodology relies solely on electricity consumption data, which may not comprehensively capture all aspects of an individual’s lifestyle. Our approach to establishing a background power threshold and converting data into a low-resolution format may introduce inaccuracies that overlook subtle changes in activity patterns. Additionally, it is important to acknowledge the potential impact of seasonal variations on electricity consumption patterns and their implications for our monitoring system’s effectiveness. The current methodology may not adequately capture how seasonal change influence individuals’ behavior, energy usage habits, and overall lifestyle. Moreover, while the system provides valuable insights into the long-term trends and gradual changes in lifestyle patterns, it may not be suitable for addressing acute health emergencies. Instead, caregivers should view the system as a supplemental tool for proactive health management, enabling them to detect subtle changes in behavior and routine that may signal underlying health concerns. It is essential to emphasize the importance of integrating our monitoring system with the existing health monitoring solutions designed for the real-time detection of acute health events, such as falls or medical emergencies. By complementing each other, these systems can provide a comprehensive approach to caring for older adults living independently, addressing both the long-term trends and immediate health concerns.
The health status dashboard prototype is designed to integrate data from three types of technologies commonly used in home monitoring services: wearable devices, ambient sensors, and household meters. Integrating the data from these sources serves several crucial purposes, but also has limitations. Wearables provide the continuous monitoring of vital signs, offering the advantage of real-time data collection for the immediate detection of abnormalities to provide a timely intervention. However, the drawbacks include potential discomfort or irritation from prolonged wear, the need for regular charging, and limited accuracy compared to that of medical-grade devices. Ambient sensors, such as smart mattress sensors, offer the passive monitoring of sleep patterns, providing valuable insights into sleep quality and duration without requiring active user engagement. They are non-intrusive and can seamlessly integrate into the home environment, but require additional installation. Smart meters track energy consumption patterns and the usage of household appliances, offering indirect insights into daily routines and activities. One advantage is their ability to capture overall energy usage trends and identify deviations from the normal patterns, which may indicate changes in lifestyle or health status. Smart meters are also cost-effective and readily available in many homes, but may lack granularity in capturing specific activities or behaviors. By integrating the data from wearables, ambient sensors, and smart meters, the health status dashboard prototype offers a comprehensive and multi-dimensional approach to monitoring the well-being of older adults, enabling the early detection of health issues, personalized care interventions, and proactive support for independent living.
Sleep data are sourced from the user’s home using the WhizPad system [
40], while vital sign data are obtained from dedicated monitoring devices transmitted to a gateway through BLE and stored in a database, with access to the data facilitated through APIs. Data collected from smart meters yield two indices: activity and regularity. Sleep data encompass five indices, the total time in bed, sleep continuity, sleep efficiency, prolonged pressure time, and the total sleep time, which are compared to those of the two-week norm [
40]. A vital sign system also provides indices for the heart rate, respiratory rate, blood pressure, and blood oxygen levels.
The health status dashboard includes a crucial section displaying the overall usage status. Specifically, it incorporates a smart meter component that utilizes two key indices. These indices are derived from comparing the previous day’s active score with the last 28-day norm and the correlation coefficient of the preceding day’s appliance usage data with the appliance usage norm of the previous 28-day data. These indices offer insights into the users’ activity and regularity levels, enhancing the understanding of their lifestyles and daily routines. The activity levels are categorized based on the comparison with norms, including a high activity level (an active score higher than high norm), a normal activity level (an active score between norm and high norm), a low activity level (an active score between norm and low norm), and an abnormally low level (an active score lower than low norm). Similarly, the regularity levels are determined based on the correlation coefficient, categorized as high regularity (CC ≥ 0.7), normal regularity (0.5 ≤ CC < 0.7), low regularity (0.03 ≤ CC < 0.5), and irregular (CC < 0.3). Users’ subjective evaluations conveyed through emojis enrich the dashboard’s assessment capabilities. A score from one to four indicates the user’s feelings, with face emojis representing the score; 1 depicts happiness, and 4 indicates sadness. The dashboard employs a color-coded system for interpretation, with colors representing the level of indices and the final assessment of users’ well-being. Green signifies a positive status, blue indicates normal, yellow denotes that attention is required, and red signals abnormality. Gray is used when the data are unavailable. For instance, in
Figure 9, the user’s active score for the day falls between the norm score and the high norm, suggesting a normal activity level depicted by the blue color in
Figure 9. Additionally, the CC exceeds 0.70, indicating a high regularity level, confirmed by the green color. Vital sign (VS) and sleep status (SS) data are currently unavailable at this phase, as indicated by the gray color. The overall user status shown in the dashboard is normal, aligning with the indices at the highest level (activity).
5. Conclusions
In this study, we utilized electricity consumption data collected from smart meter sensors installed in the homes of independently living older adults to track their lifestyle patterns. Processing these data at 15-min intervals provides valuable insights into their daily routines and energy usage behaviors. To safeguard privacy, we employed processing techniques to establish a base load threshold and convert the data into a low-resolution format (0/1), allowing us to focus on activity and regularity assessments. Our methodology involved calculating the active scores, norms, and correlation coefficients to evaluate activity and regularity, providing a comprehensive understanding of older adults’ daily habits and routines. This approach enables us to identify deviations or anomalies that may indicate changes in health or well-being. To facilitate visualization and interpretation, we developed a dashboard prototype that presents the results intuitively and in a user-friendly way.
Moreover, the dashboard is designed to integrate sleep information and vital signs data into our monitoring system, aiming for a holistic approach to elderly care. By incorporating these additional sources of information, we aimed to provide caregivers and family members with a comprehensive overview of older adults’ health status and well-being. While our study has shown promising results in monitoring the lifestyle patterns of older adults using smart meter data, several limitations must be acknowledged. Our methodology relied solely on electricity consumption data, which may not capture all the aspects of an individual’s lifestyle comprehensively. Additionally, it is important to acknowledge the potential impact of seasonal variations in electricity consumption patterns and their implications for our monitoring system’s effectiveness. Changes in seasons can influence individuals’ behavior, energy usage habits, and overall lifestyle, which may not be adequately captured by our current methodology. While the system provides valuable insights into the long-term trends and gradual changes in lifestyle patterns, it may not be suitable for addressing acute health emergencies. Instead, caregivers should view the system as a supplemental tool for proactive health management, enabling them to detect subtle changes in behavior and routine that may signal underlying health concerns.
Looking ahead, our primary objective is to further enhance the monitoring system by integrating vital signs and sleep data, improving the base load detection algorithms, and exploring advanced data analysis techniques. Additionally, real-world system validation and deployment will be crucial steps to ensure the effectiveness and reliability of our monitoring solution in supporting caregivers and keeping family members informed about the health status of older adults living independently.