Electricity consumption portrait-based old person living safety alarm threshold customization method and system

Abstract

The present application relates to a method and system for setting alarm thresholds for the daily life safety of the elderly based on power consumption portraits, the method comprising: collecting real-time power consumption data of a target residence; identifying appliance start-stop events and determining corresponding appliance types through a non-intrusive load identification method, and generating a sequence of behavior events; extracting multi-dimensional portrait features representing daily life rules from the sequence of behavior events; generating an alarm rule instance for the target elderly person according to an alarm rule template; and mapping the multi-dimensional portrait features to personalized threshold parameters of the alarm rule template according to a predefined mapping strategy; writing the personalized threshold parameters into the alarm rule instance for the target elderly person, and making an alarm determination based on real-time behavior events and the alarm rule instance. This method does not require the use of cameras or smart wearable devices, is privacy-friendly, has low cost, low false alarm rate, and high timeliness.

Description

Technical Field

[0001] This invention relates to the field of smart elderly care and electricity data analysis technology, and in particular to a method and system for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles. Background Technology

[0002] With the increasing trend of social aging, home safety monitoring for elderly people living alone or in empty nests has become a focus of social attention in recent years.

[0003] How to achieve real-time and effective monitoring of the daily life of the elderly while respecting their privacy, and promptly detect safety risks such as sudden illness or accidental falls that lead to prolonged periods of inactivity, is a technical challenge that urgently needs to be solved in the field of smart elderly care.

[0004] Currently, home-based elderly safety monitoring technologies mainly fall into three categories: 1. Wearable device-based monitoring technologies. These monitor physiological indicators such as heart rate, steps, and sleep through devices like smart bracelets and watches. However, this type of technology relies on the elderly actively wearing and charging the devices, leading to issues such as forgetting to wear them, device discomfort, and insufficient battery life, making long-term monitoring difficult. 2. Environmental sensor-based monitoring technologies. These deploy infrared human body sensors, door magnets, vibration sensors, etc., within the home to detect the elderly's activity trajectory. While this technology achieves some degree of non-intrusive monitoring, the deployment of multiple sensors is complex and costly, and it is prone to false alarms due to environmental interference (such as pet activity). 3. Video / audio-based monitoring technologies. These collect audio and video information of the elderly through cameras or microphones. Although this technology provides rich information, it poses serious privacy intrusion problems, resulting in low acceptance by the elderly and their families. Furthermore, the high cost of data processing and storage makes it unsuitable for widespread home-based elderly care scenarios.

[0005] In summary, existing technologies suffer from privacy intrusion, high deployment costs, and reliance on the elderly's active cooperation, and struggle to simultaneously meet the requirements of privacy, convenience, economy, and monitoring reliability. In recent years, electricity consumption data analysis technologies have gained attention due to their non-invasiveness, privacy-friendly nature, and the elimination of the need for additional sensor deployment; however, no existing technologies have been found to apply them to monitoring and alarm systems for the safety of elderly residents' daily lives. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles, which is timely, low-cost, has a low false alarm rate, and is privacy-friendly.

[0007] In a first aspect, the technical solution adopted by the present invention is a method for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles, comprising the following steps: S1. Collect real-time electricity consumption data of the target residence through an intelligent monitoring terminal, and preprocess the electricity consumption data to obtain preprocessed electricity consumption data; S2. Based on the preprocessed electricity consumption data, identify appliance start-stop events using a non-intrusive load identification method, determine the corresponding appliance type, and generate a sequence of behavioral events. S3. Using a rolling time window of the most recent N days, extract multi-dimensional profile features representing daily routines from the sequence of behavioral events. The multi-dimensional profile features include wake-up time features, bedtime features, distribution features of continuous non-active behavior duration, and usage patterns of key daily appliances. S4. Load the alarm rule template, generate alarm rule instances for the target elderly based on the alarm rule template, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule template according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters. S5. Write the personalized threshold parameter into the alarm rule instance of the target elderly person, and make an alarm determination based on the real-time behavioral event and the alarm rule instance.

[0008] The beneficial effects of this invention are as follows: The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles, as described above, collects household electricity consumption data through a smart monitoring terminal, eliminating the need for cameras or smart wearable devices, thus protecting privacy and reducing costs. The household electricity consumption data is processed through non-intrusive load identification to generate behavioral event sequences, extracting multi-dimensional living profile features, and mapping them to personalized time windows and behavioral interval thresholds. These are then written into alarm rule instances for judgment. This method customizes thresholds based on individual profiles, avoiding false alarms caused by uniform thresholds, resulting in a low false alarm rate. Furthermore, the method uses a rolling window to dynamically update the profile, adapting to seasonal and habit changes, ensuring high timeliness.

[0009] Preferably, step S2 includes the following steps: S21. Extract the active power sequence based on the preprocessed electricity consumption data and detect the power change. S22. When the absolute value of the power change exceeds a preset power threshold and the duration exceeds a preset time threshold, it is identified as an appliance start-stop event. S23. Extract the electrical appliance features corresponding to the electrical appliance start-stop event, match them with a preset electrical appliance feature library, and determine the electrical appliance type corresponding to the electrical appliance start-stop event; the electrical appliance features include power features, harmonic features, and transient waveform features; S24. Based on the identified appliance start / stop events and their corresponding appliance types, generate a sequence of behavioral events containing timestamps, appliance types, and action types.

[0010] Preferably, step S23 includes the following steps: The steady-state active power change, reactive power change, and total harmonic distortion rate of the current are extracted from the first preset time before the electrical appliance start-stop event to the second preset time after the event, as power features; the 3rd, 5th, and 7th harmonic components in the current waveform at the time of the electrical appliance start-stop event are extracted as harmonic features. The rise time, fall time, peak overshoot, and transient duration of the power waveform at the time of the electrical appliance start-stop event are extracted as transient waveform features. The extracted power features, harmonic features, and transient waveform features are combined to form an event feature vector; Calculate the cosine similarity between the event feature vector and the feature template vectors corresponding to various types of electrical appliances in the preset electrical appliance feature library; The appliance type with the highest cosine similarity and greater than a preset similarity threshold is selected as the identification result of the appliance start / stop event; if all similarities are less than or equal to the preset similarity threshold, they are marked as unidentified events.

[0011] Preferably, step S3 includes the following steps: S31. Set the window length of the scrolling time window to N days, take the current date as the end date of the window, and extract the sequence of behavioral events from the most recent N days; S32. Statistically determine the occurrence time of the first active electricity consumption event with a duration exceeding a first preset duration each day, form a wake-up time sample set, and calculate the statistical distribution of the wake-up time sample set as a wake-up time feature; wherein, the first active electricity consumption event with a duration exceeding the first preset duration refers to: starting from the early morning of the day, detecting the first behavioral event segment with a duration greater than or equal to 15 minutes and containing at least one electrical appliance running continuously for more than 15 minutes. S33. Statistically determine the start time of entering the low-power state after the last electricity consumption event of the day, form a bedtime sample set, and calculate the statistical distribution of the bedtime sample set as the bedtime period feature; wherein, the start time of entering the low-power state refers to the starting time when the total active power is lower than the preset base power threshold for more than 2 consecutive hours after the last electricity consumption event of the day. S34. Calculate the time interval between any two adjacent electricity consumption events that are actively triggered by humans each day, forming an interval duration sample set, and calculate the statistical distribution of the interval duration sample set as the distribution feature of continuous non-active behavior duration; wherein, the electricity consumption events that are actively triggered by humans refer to the events remaining after excluding the start-stop events of refrigerators, routers and set-top boxes. S35. Statistically analyze the usage frequency and time of each use of key electrical appliances within a rolling time window, and extract the concentration characteristics of the usage period as the usage pattern characteristics of key daily electrical appliances. The concentration characteristics include the center value and dispersion of the usage period.

[0012] Preferably, in step S4, the specific process of loading the alarm rule template and generating an alarm rule instance for the target elderly person based on the alarm rule template includes the following steps: S401. Load a predefined alarm rule template. The alarm rule template contains several alarm rules. Each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. S402. Obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; S403. If a corresponding historical alarm rule instance exists, then load the historical alarm rule instance; if it does not exist, then copy the alarm rule template, name it with the unique identifier, and generate a new alarm rule instance, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. S404. Store the generated alarm rule instance in the rule instance library and establish an association with the target elderly person.

[0013] Preferably, in step S4, the specific process of mapping the multi-dimensional profile features to the personalized threshold parameters of the alarm rule template according to a predefined mapping strategy includes the following steps: S411. Obtain the typical wake-up time feature in the multi-dimensional profile feature, and use the statistical quantile mapping method to map the Pth percentile and (100-P)th percentile of the wake-up time sample set in the feature to the lower and upper bounds of the wake-up time window parameter in the alarm rule template, respectively. S412. Obtain the typical bedtime features in the multi-dimensional profile features, and use the statistical quantile mapping method to map the Q percentile and (100-Q) percentile of the bedtime sample set in the feature to the lower and upper bounds of the bedtime window parameters in the alarm rule template, respectively. S413. Obtain the distribution features of continuous non-active behavior duration in the multi-dimensional profile features, use the median absolute deviation method to suppress outliers in the interval duration sample set, take the R percentile of the suppressed sample, and map it to the behavior interval threshold parameter in the alarm rule template.

[0014] Preferably, step S5 includes the following steps: S51. Obtain the personalized threshold parameters generated in step S4; S52. Write the personalized threshold parameter into the alarm rule instance of the target elderly person, overwriting the threshold parameter that already exists in the alarm rule instance; S53. Collect and process electricity consumption data in real time, and generate real-time behavioral events according to step S2; S54. Compare the real-time behavioral event with each alarm rule in the alarm rule instance. If the current time exceeds the upper limit of the wake-up time window in the alarm rule instance and no wake-up characteristic event has been detected that day, trigger a wake-up too late alarm. If the current time exceeds the upper limit of the bedtime time window in the alarm rule instance and no bedtime characteristic event has been detected that day, trigger a bedtime abnormal alarm. If the interval between consecutive periods of no active behavior in the real-time behavioral event exceeds the behavior interval threshold parameter in the alarm rule instance, trigger a long period of no active behavior alarm. Before triggering the alarm, it is necessary to determine whether the current time is within the outing registration period. If it is within the outing registration period, suppress the alarm or delay the alarm. S55. Push the triggered alarm information to the preset guardian terminal or cloud management platform.

[0015] Secondly, the technical solution adopted by this invention is a customized alarm threshold system for elderly people's daily living safety based on electricity consumption profiles, the system comprising: The data acquisition module is used to collect real-time electricity consumption data of the target residence through an intelligent monitoring terminal, and to preprocess the electricity consumption data to obtain preprocessed electricity consumption data. The profile building module is used to identify appliance start-stop events and determine the corresponding appliance types based on preprocessed electricity consumption data using a non-intrusive load identification method, and generate a behavioral event sequence; and to extract multi-dimensional profile features representing daily routines from the behavioral event sequence using a rolling time window of the most recent N days; the multi-dimensional profile features include wake-up time features, bedtime features, continuous non-active behavior duration distribution features, and key daily appliance usage pattern features. The rule generation and mapping module is used to load alarm rule templates, generate alarm rule instances for the target elderly based on the alarm rule templates, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule templates according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters. The alarm execution module is used to write the personalized threshold parameters into the alarm rule instance of the target elderly person, and to make an alarm determination based on real-time behavioral events and the alarm rule instance. The alarm execution module is also used to trigger the profile construction module and the rule generation and mapping module with a daily update cycle, to re-extract multi-dimensional profile features and update personalized threshold parameters based on the latest recent N-day rolling time window. In addition, information on the target elderly person's outing status is obtained, and during the confirmed outing period, alarm rules requiring home status are suppressed, downgraded, or delayed.

[0016] Preferably, the portrait construction module specifically includes: An appliance start-stop event identification unit is used to detect power change based on the active power sequence in the preprocessed power consumption data. When the absolute value of the power change exceeds a preset power threshold and the duration exceeds a preset time threshold, it is identified as an appliance start-stop event. The appliance type matching unit is used to extract the power characteristics, harmonic characteristics and transient waveform characteristics corresponding to the appliance start-stop event, and match them with a preset appliance feature library to determine the appliance type corresponding to the appliance start-stop event. The event sequence generation unit is used to generate a sequence of behavioral events containing timestamps, appliance types, and action types based on the identified appliance start / stop events and their corresponding appliance types.

[0017] Preferably, the rule generation and mapping module specifically includes: The template loading unit is used to load a predefined alarm rule template. The alarm rule template contains several alarm rules, and each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. The instance generation unit is used to obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; if it exists, the historical alarm rule instance is loaded; if it does not exist, the alarm rule template is copied, named with the unique identifier, and a new alarm rule instance is generated, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. The parameter mapping unit is used to map the multi-dimensional profile features into personalized threshold parameters according to a predefined mapping strategy. This includes mapping the wake-up time features and bedtime features into time window parameters using the statistical quantile mapping method, and mapping the continuous non-active behavior duration distribution features into behavior interval threshold parameters using the median absolute deviation method. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles, as described in this invention. Figure 2This is a schematic diagram of the elderly living safety alarm threshold customization system based on electricity consumption profiles according to the present invention. Detailed Implementation

[0019] The invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can implement it based on the description. The scope of protection of the invention is not limited to these specific embodiments.

[0020] Example 1 This embodiment provides a method for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles. Taking an elderly person living alone as the target user, a smart meter is deployed at the electricity meter in their residence to collect real-time electricity consumption data.

[0021] The system relied upon in this embodiment includes an intelligent monitoring terminal (i.e., a smart meter), a data acquisition module, a profile building module, a rule generation and mapping module, an alarm execution module, and a guardian terminal (such as a guardian's mobile APP).

[0022] like Figure 1 As shown, the method includes the following steps: Step 1: Obtain preprocessed electricity data by using real-time electricity consumption data from the target residence via a smart meter.

[0023] The real-time electricity consumption data specifically includes: voltage (V), current (A), active power (W), and reactive power (Var).

[0024] Specifically, the preprocessing process for the collected electricity consumption data is as follows: linear interpolation is used to fill in missing data caused by communication interruptions, etc.; the median absolute deviation method is used to detect outliers that exceed the preset fluctuation range, mark them and replace them with the average of adjacent valid values; the data after completion and replacement are aligned according to the collection timestamp and uniformly converted into structured data in JSON format to obtain the preprocessed electricity consumption data.

[0025] Step 2: Based on the preprocessed electricity consumption data, identify appliance start-stop events using a non-intrusive load identification method, determine the corresponding appliance type, and generate a sequence of behavioral events.

[0026] Specifically, step two includes the following steps: S21. Extract the active power sequence based on the preprocessed electricity consumption data and detect the power change at adjacent time points. ; S22. Set the power threshold to 20W and the time threshold to 3 seconds. When |ΔP| ≥ 20W and the duration of this power change is ≥ 3 seconds, it is identified as an appliance start / stop event; S23. Extract the electrical appliance features corresponding to the electrical appliance start-stop event, match them with a preset electrical appliance feature library, and determine the electrical appliance type corresponding to the electrical appliance start-stop event; the electrical appliance features include power features, harmonic features, and transient waveform features; S24. Based on the identified appliance start / stop events and their corresponding appliance types, generate a sequence of behavioral events containing timestamps, appliance types, and action types.

[0027] The specific process of step S23 is as follows: The steady-state active power change, reactive power change, and total harmonic distortion rate of the current are extracted from 0.5 seconds before the start-stop event to 1 second after the event as power features; the 3rd, 5th, and 7th harmonic components in the current waveform at the time of the start-stop event are extracted as harmonic features. The rise time, fall time, peak overshoot, and transient duration of the power waveform at the time of the electrical appliance start-stop event are extracted as transient waveform features. The extracted power features, harmonic features, and transient waveform features are combined to form an event feature vector; Calculate the cosine similarity between the event feature vector and the feature template vectors corresponding to various types of electrical appliances in the preset electrical appliance feature library; wherein, various types of electrical appliances may be rice cookers, kettles, microwave ovens, televisions, refrigerators, routers, air conditioners, etc. The appliance type with the highest cosine similarity and greater than a preset similarity threshold is selected as the identification result of the appliance start / stop event; if all similarities are less than or equal to the preset similarity threshold, they are marked as unidentified events.

[0028] Step 3: Using a rolling time window of the most recent N days, extract multi-dimensional profile features representing daily routines from the sequence of behavioral events; the multi-dimensional profile features include wake-up time features, bedtime features, distribution features of continuous non-active behavior duration, and key daily appliance usage patterns.

[0029] Specifically, step three includes the following steps: S31. Set the window length of the scrolling time window to 14 days, with the current date as the end date of the window, and extract the sequence of behavioral events for the most recent 14 days; S32. Extract wake-up time characteristics: Statistically determine the occurrence time of the first behavioral event segment of each day that lasts longer than 15 minutes and includes at least one appliance running continuously for more than 15 minutes. For example, if an elderly person habitually turns on the kettle around 6:30 AM and then uses the rice cooker, and this activity lasts longer than 15 minutes, then the wake-up time for that day is recorded as 6:30 AM. Form a wake-up time sample set from 14 days of samples, and calculate its 5th percentile (P5) and 95th percentile (P95), which will be used as the lower and upper bounds of the wake-up time window, respectively. S33. Extract bedtime characteristics: Analyze the starting point when the total active power remains below a preset baseline power threshold (e.g., 50W) for more than two consecutive hours after the last electricity usage event of the day. For example, if an elderly person turns off the TV at 10:00 PM and the total power remains below 50W for the next two hours, then their bedtime for that day is recorded as 10:00 PM. Compile the 14 days of bedtime samples into a bedtime sample set, and calculate its 5th and 95th percentiles as the lower and upper bounds of the bedtime window. S34. Extract the distribution characteristics of non-active behavior duration: Calculate the time interval between any two adjacent electricity consumption events triggered by humans each day, excluding start-stop events of continuously running or timed appliances such as refrigerators, routers, and set-top boxes. Form an interval duration sample set from all valid interval durations within 14 days, and calculate its 95th percentile (Q95) and standard deviation σ; S35. Extract usage patterns of key everyday appliances: Statistically analyze the usage frequency and time of each of the preset key appliances (e.g., kettles, rice cookers) over a 14-day period, and use the mean clustering algorithm to extract the cluster centers and standard deviations for each usage period. For example, kettle usage is concentrated between 7:00-7:30 and 19:00-19:30, with center values ​​of 7:15 and 19:15, respectively.

[0030] Step 4: Load the alarm rule template, generate alarm rule instances for the target elderly person based on the alarm rule template, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule template according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters.

[0031] Specifically, the process of loading the alarm rule template and generating an alarm rule instance for the target elderly person based on the alarm rule template includes the following steps: S401. Load a predefined alarm rule template. The alarm rule template contains several alarm rules. Each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. S402. Obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; S403. If a corresponding historical alarm rule instance exists, then load the historical alarm rule instance; if it does not exist, then copy the alarm rule template, name it with the unique identifier, and generate a new alarm rule instance, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. S404. Store the generated alarm rule instance in the rule instance library and establish an association with the target elderly person.

[0032] The predefined alarm rule templates include: Rule 1 (Wake up too late): The parameters are the lower and upper bounds of the wake-up time window, with a default value of [6:00, 9:00]. Rule 2 (Abnormal Bedtime): The parameters are the lower and upper bounds of the bedtime window, with a default value of [21:00, 23:00]; Rule 3 (Long period without proactive behavior): The parameter is the behavior interval threshold, with a default value of 3 hours.

[0033] Specifically, the process of mapping the multi-dimensional profile features to personalized threshold parameters of the alarm rule template according to a predefined mapping strategy includes the following steps: S411. Obtain the typical wake-up time features in the multi-dimensional profile features, and use the statistical quantile mapping method to map the Pth percentile and (100-P)th percentile of the wake-up time sample set in the feature to the lower and upper bounds of the wake-up time window parameters in the alarm rule template, respectively; for example, map the wake-up time features 6:15 and 8:15 to the lower and upper bounds of the wake-up time window in rule one, that is, the wake-up time window is [6:15, 8:15]; S412. Obtain the typical bedtime features in the multi-dimensional profile features, and use the statistical quantile mapping method to map the Q percentile and (100-Q) percentile of the bedtime sample set in the feature to the lower and upper bounds of the bedtime window parameters in the alarm rule template, respectively; map the bedtime features 21:30 and 23:00 to the lower and upper bounds of the bedtime window in rule two, respectively, that is, the bedtime window is [21:30, 23:00]; S413. Obtain the distribution features of continuous non-active behavior duration in the multi-dimensional profile features, use the median absolute deviation method to suppress outliers in the interval duration sample set, take the R percentile of the suppressed sample, and map it to the behavior interval threshold parameter in the alarm rule template.

[0034] Step 5: Write the personalized threshold parameters into the alarm rule instance of the target elderly person, and make an alarm determination based on the real-time behavioral events and the alarm rule instance.

[0035] Specifically, step five includes the following steps: S51. Obtain the personalized threshold parameters generated in step S4; for example, wake-up time window [6:15, 8:15], bedtime window [21:30, 23:00], and behavior interval threshold of 4.5 hours. S52. Write the personalized threshold parameter into the alarm rule instance of the target elderly person, overwriting the threshold parameter that already exists in the alarm rule instance; S53. Collect and process electricity consumption data in real time, and generate real-time behavioral events according to the method in step S2; S54. Compare the real-time behavioral event with each alarm rule in the alarm rule instance. If the current time exceeds the upper limit of the wake-up time window in the alarm rule instance and no wake-up characteristic event has been detected that day, trigger a late wake-up alarm. If the current time exceeds the upper limit of the bedtime time window in the alarm rule instance and no bedtime characteristic event has been detected that day, trigger an abnormal bedtime alarm. If the interval between consecutive periods of no active behavior in the real-time behavioral event exceeds the behavior interval threshold parameter in the alarm rule instance, trigger a long period of no active behavior. Alarms; before triggering an alarm, it is necessary to determine whether the current time is within the outing registration period. If it is, the alarm will be suppressed or delayed. For example: if the current time is past 8:15 (the upper limit of the wake-up time window) and no wake-up characteristic event has been detected that day, a wake-up too late alarm will be triggered; if the current time is past 23:00 (the upper limit of the bedtime time window) and no bedtime characteristic event has been detected that day, an abnormal bedtime alarm will be triggered; if the interval between consecutive periods of no active behavior in real-time behavioral events exceeds 4.5 hours, a long period of no active behavior alarm will be triggered. S55. Once an alarm is triggered, the system will push the alarm information (including user ID, alarm type, trigger time, and real-time behavioral event description) to the preset guardian terminal or cloud management platform.

[0036] like Figure 2 As shown, a system for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles is described. The system includes: The data acquisition module is used to collect real-time electricity consumption data of the target residence through an intelligent monitoring terminal, and to preprocess the electricity consumption data to obtain preprocessed electricity consumption data. The profile building module is used to identify appliance start-stop events and determine the corresponding appliance types based on preprocessed electricity consumption data using a non-intrusive load identification method, and generate a behavioral event sequence; and to extract multi-dimensional profile features representing daily routines from the behavioral event sequence using a rolling time window of the most recent N days; the multi-dimensional profile features include wake-up time features, bedtime features, continuous non-active behavior duration distribution features, and key daily appliance usage pattern features. The rule generation and mapping module is used to load alarm rule templates, generate alarm rule instances for the target elderly based on the alarm rule templates, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule templates according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters. The alarm execution module is used to write the personalized threshold parameters into the alarm rule instance of the target elderly person, and to make an alarm determination based on real-time behavioral events and the alarm rule instance. The alarm execution module is also used to trigger the profile construction module and the rule generation and mapping module with a daily update cycle, to re-extract multi-dimensional profile features and update personalized threshold parameters based on the latest recent N-day rolling time window. In addition, information on the target elderly person's outing status is obtained, and during the confirmed outing period, alarm rules requiring home status are suppressed, downgraded, or delayed.

[0037] The portrait construction module specifically includes: An appliance start-stop event identification unit is used to detect power change based on the active power sequence in the preprocessed power consumption data. When the absolute value of the power change exceeds a preset power threshold and the duration exceeds a preset time threshold, it is identified as an appliance start-stop event. The appliance type matching unit is used to extract the power characteristics, harmonic characteristics and transient waveform characteristics corresponding to the appliance start-stop event, and match them with a preset appliance feature library to determine the appliance type corresponding to the appliance start-stop event. The event sequence generation unit is used to generate a sequence of behavioral events containing timestamps, appliance types, and action types based on the identified appliance start / stop events and their corresponding appliance types.

[0038] The rule generation and mapping module specifically includes: The template loading unit is used to load a predefined alarm rule template. The alarm rule template contains several alarm rules, and each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. The instance generation unit is used to obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; if it exists, the historical alarm rule instance is loaded; if it does not exist, the alarm rule template is copied, named with the unique identifier, and a new alarm rule instance is generated, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. The parameter mapping unit is used to map the multi-dimensional profile features into personalized threshold parameters according to a predefined mapping strategy. This includes mapping the wake-up time features and bedtime features into time window parameters using the statistical quantile mapping method, and mapping the continuous non-active behavior duration distribution features into behavior interval threshold parameters using the median absolute deviation method.

[0039] The system updates daily, automatically repeating steps three through five every morning to re-extract profile features and update personalized threshold parameters based on the latest 14-day rolling window. For example, since the elderly habitually wake up early after the start of summer, the wake-up time window will be automatically adjusted to [5:45, 7:45].

Claims

1. A method for customizing alarm thresholds for elderly people's daily living safety based on electricity consumption profiles, characterized in that: Includes the following steps: S1. Collect real-time electricity consumption data of the target residence through an intelligent monitoring terminal, and preprocess the electricity consumption data to obtain preprocessed electricity consumption data; S2. Based on the preprocessed electricity consumption data, identify appliance start-stop events using a non-intrusive load identification method, determine the corresponding appliance type, and generate a sequence of behavioral events. S3. Using a rolling time window of the most recent N days, extract multi-dimensional profile features representing daily routines from the sequence of behavioral events. The multi-dimensional profile features include wake-up time features, bedtime features, distribution features of continuous non-active behavior duration, and usage patterns of key daily appliances. S4. Load the alarm rule template, generate alarm rule instances for the target elderly based on the alarm rule template, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule template according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters. S5. Write the personalized threshold parameter into the alarm rule instance of the target elderly person, and make an alarm determination based on the real-time behavioral event and the alarm rule instance.

2. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 1, characterized in that: Step S2 includes the following steps: S21. Extract the active power sequence based on the preprocessed electricity consumption data and detect the power change. S22. When the absolute value of the power change exceeds a preset power threshold and the duration exceeds a preset time threshold, it is identified as an appliance start-stop event. S23. Extract the electrical appliance features corresponding to the electrical appliance start-stop event, match them with a preset electrical appliance feature library, and determine the electrical appliance type corresponding to the electrical appliance start-stop event; the electrical appliance features include power features, harmonic features, and transient waveform features; S24. Based on the identified appliance start / stop events and their corresponding appliance types, generate a sequence of behavioral events containing timestamps, appliance types, and action types.

3. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 2, characterized in that: The specific process of step S23 includes the following steps: The steady-state active power change, reactive power change, and total harmonic distortion of current are extracted from the first preset time before the electrical appliance start-stop event to the second preset time after the event, and used as power characteristics. The 3rd, 5th, and 7th harmonic components in the current waveform when the electrical appliance starts or stops are extracted as harmonic features; The rise time, fall time, peak overshoot, and transient duration of the power waveform at the time of the electrical appliance start-stop event are extracted as transient waveform features. The extracted power features, harmonic features, and transient waveform features are combined to form an event feature vector; Calculate the cosine similarity between the event feature vector and the feature template vectors corresponding to various types of electrical appliances in the preset electrical appliance feature library; The appliance type with the highest cosine similarity and greater than a preset similarity threshold is selected as the identification result of the appliance start / stop event; if all similarities are less than or equal to the preset similarity threshold, they are marked as unidentified events.

4. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 2 or 3, characterized in that: Step S3 includes the following steps: S31. Set the window length of the scrolling time window to N days, take the current date as the end date of the window, and extract the sequence of behavioral events from the most recent N days; S32. Statistically determine the occurrence time of the first active electricity consumption event with a duration exceeding a first preset duration each day, form a wake-up time sample set, and calculate the statistical distribution of the wake-up time sample set as a wake-up time feature; wherein, the first active electricity consumption event with a duration exceeding the first preset duration refers to: starting from the early morning of the day, detecting the first behavioral event segment with a duration greater than or equal to 15 minutes and containing at least one electrical appliance running continuously for more than 15 minutes. S33. Statistically determine the start time of entering the low-power state after the last electricity consumption event of the day, form a bedtime sample set, and calculate the statistical distribution of the bedtime sample set as the bedtime period feature; wherein, the start time of entering the low-power state refers to the starting time when the total active power is lower than the preset base power threshold for more than 2 consecutive hours after the last electricity consumption event of the day. S34. Calculate the time interval between any two adjacent electricity consumption events that are actively triggered by humans each day, forming an interval duration sample set, and calculate the statistical distribution of the interval duration sample set as the distribution feature of continuous non-active behavior duration; wherein, the electricity consumption events that are actively triggered by humans refer to the events remaining after excluding the start-stop events of refrigerators, routers and set-top boxes. S35. Statistically analyze the usage frequency and time of each use of key electrical appliances within a rolling time window, and extract the concentration characteristics of the usage period as the usage pattern characteristics of key daily electrical appliances. The concentration characteristics include the center value and dispersion of the usage period.

5. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 4, characterized in that: In step S4, the specific process of loading the alarm rule template and generating an alarm rule instance for the target elderly person based on the alarm rule template includes the following steps: S401. Load a predefined alarm rule template. The alarm rule template contains several alarm rules. Each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. S402. Obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; S403. If a corresponding historical alarm rule instance exists, then load the historical alarm rule instance; if it does not exist, then copy the alarm rule template, name it with the unique identifier, and generate a new alarm rule instance, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. S404. Store the generated alarm rule instance in the rule instance library and establish an association with the target elderly person.

6. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 5, characterized in that: In step S4, the specific process of mapping the multi-dimensional profile features to the personalized threshold parameters of the alarm rule template according to the predefined mapping strategy includes the following steps: S411. Obtain the typical wake-up time feature in the multi-dimensional profile feature, and use the statistical quantile mapping method to map the Pth percentile and (100-P)th percentile of the wake-up time sample set in the feature to the lower and upper bounds of the wake-up time window parameter in the alarm rule template, respectively. S412. Obtain the typical bedtime features in the multi-dimensional profile features, and use the statistical quantile mapping method to map the Q percentile and (100-Q) percentile of the bedtime sample set in the feature to the lower and upper bounds of the bedtime window parameters in the alarm rule template, respectively. S413. Obtain the distribution features of continuous non-active behavior duration in the multi-dimensional profile features, use the median absolute deviation method to suppress outliers in the interval duration sample set, take the R percentile of the suppressed sample, and map it to the behavior interval threshold parameter in the alarm rule template.

7. The method for customizing alarm thresholds for elderly living safety based on electricity consumption profiles according to claim 6, characterized in that: Step S5 includes the following steps: S51. Obtain the personalized threshold parameters generated in step S4; S52. Write the personalized threshold parameter into the alarm rule instance of the target elderly person, overwriting the threshold parameter that already exists in the alarm rule instance; S53. Collect and process electricity consumption data in real time, and generate real-time behavioral events according to step S2; S54. Compare the real-time behavioral event with each alarm rule in the alarm rule instance. If the current time exceeds the upper limit of the wake-up time window in the alarm rule instance and no wake-up characteristic event has been detected that day, trigger a wake-up too late alarm. If the current time exceeds the upper limit of the bedtime time window in the alarm rule instance and no bedtime characteristic event has been detected that day, trigger a bedtime abnormal alarm. If the interval between consecutive periods of no active behavior in the real-time behavioral event exceeds the behavior interval threshold parameter in the alarm rule instance, trigger a long period of no active behavior alarm. Before triggering the alarm, it is necessary to determine whether the current time is within the outing registration period. If it is within the outing registration period, suppress the alarm or delay the alarm. S55. Push the triggered alarm information to the preset guardian terminal or cloud management platform.

8. A system for customizing alarm thresholds for elderly living safety based on electricity consumption profiles, used to implement the customization method as described in any one of claims 1 to 7, the system comprising: The data acquisition module is used to collect real-time electricity consumption data of the target residence through an intelligent monitoring terminal, and to preprocess the electricity consumption data to obtain preprocessed electricity consumption data. The profile building module is used to identify appliance start-stop events based on preprocessed electricity consumption data using a non-intrusive load identification method, determine the corresponding appliance type, and generate a sequence of behavioral events. And by using a rolling time window of the most recent N days, multi-dimensional profile features representing daily routines are extracted from the sequence of behavioral events; the multi-dimensional profile features include wake-up time features, bedtime features, distribution features of continuous non-active behavior duration, and key daily appliance usage patterns. The rule generation and mapping module is used to load alarm rule templates, generate alarm rule instances for the target elderly based on the alarm rule templates, and map the multi-dimensional profile features to personalized threshold parameters of the alarm rule templates according to a predefined mapping strategy. The personalized threshold parameters include time window parameters and behavior interval threshold parameters. The alarm execution module is used to write the personalized threshold parameters into the alarm rule instance of the target elderly person, and to make an alarm determination based on real-time behavioral events and the alarm rule instance. The alarm execution module is also used to trigger the profile construction module and the rule generation and mapping module with a daily update cycle, to re-extract multi-dimensional profile features and update personalized threshold parameters based on the latest recent N-day rolling time window. In addition, information on the target elderly person's outing status is obtained, and during the confirmed outing period, alarm rules requiring home status are suppressed, downgraded, or delayed.

9. The elderly living safety alarm threshold customization system based on electricity consumption profile as described in claim 8, characterized in that: The portrait construction module specifically includes: An appliance start-stop event identification unit is used to detect power change based on the active power sequence in the preprocessed power consumption data. When the absolute value of the power change exceeds a preset power threshold and the duration exceeds a preset time threshold, it is identified as an appliance start-stop event. The appliance type matching unit is used to extract the power characteristics, harmonic characteristics and transient waveform characteristics corresponding to the appliance start-stop event, and match them with a preset appliance feature library to determine the appliance type corresponding to the appliance start-stop event. The event sequence generation unit is used to generate a sequence of behavioral events containing timestamps, appliance types, and action types based on the identified appliance start / stop events and their corresponding appliance types.

10. The elderly living safety alarm threshold customization system based on electricity consumption profiles according to claim 9, characterized in that: The rule generation and mapping module specifically includes: The template loading unit is used to load a predefined alarm rule template. The alarm rule template contains several alarm rules, and each alarm rule has a rule type identifier, a default threshold parameter set, and a parameter mapping interface. The instance generation unit is used to obtain the unique identifier of the target elderly person, and query the rule instance library to see if there is a corresponding historical alarm rule instance based on the unique identifier; if it exists, the historical alarm rule instance is loaded; if it does not exist, the alarm rule template is copied, named with the unique identifier, and a new alarm rule instance is generated, wherein the parameters of each alarm rule are initialized to the default threshold parameters in the alarm rule template. The parameter mapping unit is used to map the multi-dimensional profile features into personalized threshold parameters according to a predefined mapping strategy. This includes mapping the wake-up time features and bedtime features into time window parameters using the statistical quantile mapping method, and mapping the continuous non-active behavior duration distribution features into behavior interval threshold parameters using the median absolute deviation method.

Images