User power consumption behavior portraying method and system based on non-intrusive load monitoring

Abstract

The invention relates to a user power consumption behavior portraying method and system based on non-intrusive load monitoring, and belongs to the technical field of residential user power consumption behavior analysis and energy management. According to the technical scheme, a data acquisition module collects user intelligent electric meter data; the user load clustering analysis module adopts a clustering algorithm to classify user load curves; the equipment-level behavior analysis module is used for analyzing the energy consumption ratio, the single-equipment use period preference and the multi-equipment cooperation mode of the equipment type for the user; the user portrait construction module constructs a multi-dimensional user power consumption behavior portrait; and the system application module formulates a personalized energy efficiency optimization scheme. According to the method, user load curve clustering analysis and device-level power consumption data are fused, the types and start-stop time of power consumption devices are identified through the NILM technology, and the total load clustering result is combined, so that fine-grained modeling of user power consumption behaviors is realized, and multi-dimensional mining of the device type energy consumption ratio, the single-device use period preference and the multi-device cooperation mode is carried out.

Description

Technical Field

[0001] The present invention relates to a method and system for profiling user electricity consumption behavior based on non-invasive load monitoring, belonging to the technical field of residential user electricity consumption behavior analysis and energy management. Background Art

[0002] Existing power user profiling techniques primarily perform cluster analysis based on a user's 96-point daily load curve (i.e., 24-hour electricity consumption data collected every 15 minutes). Using algorithms such as K-means and DBSCAN, users are divided into several categories, generating typical electricity consumption patterns (e.g., "peak-valley" and "stable"). While these methods can initially reflect a user's overall electricity consumption patterns, they suffer from the following significant drawbacks in practical applications:

[0003] (1) Equipment type is invisible and energy consumption structure is fuzzy

[0004] Users with similar total load curves may have significantly different device combinations. For example: User A: 1 3-horsepower air conditioner (60% of the load in the cooling season) + basic lighting; User B: 5 low-power devices (computer + printer + water dispenser + LED light + sweeping robot, each with a power of less than 200W). The total load curves of the two may be similar, but the nature of their electricity consumption is completely different (A is "rigid cooling demand" and B is "home office mode"), and traditional clustering cannot distinguish them.

[0005] (2) Time period preferences are coarse-grained, and behavioral patterns are difficult to analyze

[0006] Traditional methods focus solely on "load levels," ignoring the precise time-of-day characteristics of devices. For example, a water heater's peak usage between 6:00 AM and 7:00 AM and 9:00 PM and 10:00 PM reflects users' morning and evening washing habits; a washing machine's concentrated weekend usage between 2:00 PM and 4:00 PM corresponds to a household cleaning schedule. These fine-grained behaviors require device-level data to be exploited, while the aggregated load curve only shows a flat top or peak.

[0007] (3) Lack of equipment coordination and lack of scenario-based analysis

[0008] Household electricity consumption often involves device-to-device collaboration. For example, when the air conditioner is on, the fan also runs simultaneously (reducing the air conditioner's power consumption); or kitchen appliances (range hood, gas stove, and microwave) are used intensively between 6 PM and 7 PM, corresponding to the dinnertime. Traditional methods are unable to identify these collaborative relationships, making them incapable of supporting scenario-based energy-saving recommendations.

[0009] Defining users' daily electricity usage behavior covers the following aspects:

[0010] (1) Overall power load characteristics

[0011] Load curve shape: reflects the changes in power consumption of users at different times of the day, including morning peak, evening peak, stable period and off-peak period, revealing the overall power consumption trend;

[0012] Energy consumption distribution: Statistics on energy consumption in each period reflect the user's electricity consumption and appliance usage density in different time periods;

[0013] (2) Single device power consumption behavior

[0014] Device identification: Analyzes the start and stop times, operating hours, and power characteristics of major household appliances through non-intrusive load monitoring (NILM) technology.

[0015] Device usage patterns: Based on the probability of device usage at different times of the day, identify patterns such as fixed-time usage, peak-time usage, sporadic random usage, and long-term continuous operation;

[0016] (3) Multi-device collaboration and association

[0017] Device collaboration: Analyzes the coordinated use of different appliances, such as kitchen appliances turning on simultaneously during dinner time or the simultaneous operation of air conditioners and electric fans.

[0018] Timing dependency: By analyzing the sequence and duration of device start-up and shutdown, we can build a model of device collaborative behavior in specific scenarios (such as dining, resting, and working). Summary of the Invention

[0019] The purpose of the present invention is to provide a user electricity consumption behavior profiling method and system based on non-invasive load monitoring. The type and start and stop time of the electrical equipment identified by NILM technology are combined with the total load clustering results to analyze the proportion of equipment electricity consumption, and construct user tags from three dimensions: equipment type, usage sequence, and collaborative mode. This breaks through the limitation of traditional methods that only rely on the total load curve and solves the above-mentioned problems existing in the background technology.

[0020] The technical solution of the present invention is:

[0021] A user electricity consumption behavior portrait system based on non-intrusive load monitoring includes a data acquisition module, a user load clustering analysis module, a device-level behavior analysis module and a user portrait construction module. The data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to parse out device-level electricity consumption information; the user load clustering analysis module uses a clustering algorithm to classify user load curves and determine the user load curve clustering results; the device-level behavior analysis module determines the device-level behavior characteristics based on the device-level electricity consumption information; the user portrait construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user electricity consumption behavior portrait.

[0022] The data acquisition module is responsible for collecting users' smart meter data and analyzing device-level electricity usage information through the non-intrusive load monitoring (NILM) method. The module's tasks include: collecting raw electricity usage data, including total load data; analyzing the operating information of various electrical appliances, including start and stop times, operating duration, and power characteristics; and performing data cleaning and preprocessing.

[0023] The data acquisition module collects raw electricity consumption data and obtains the total load data of the user's home through a smart meter or distributed energy consumption monitoring terminal. The smart meter records the user's electricity power, voltage, and current at a high frequency and stores them in a local or cloud database. The data is continuously recorded at a set sampling interval. When a sudden change in electricity consumption is detected, high-frequency acquisition is triggered to capture the start and stop events of electrical appliances. The data acquisition module analyzes the operating information of various electrical appliances and decomposes the total load data using the non-invasive load monitoring (NILM) method to identify the electricity consumption patterns of the main electrical appliances in the home. The power change characteristics of the appliances are used to identify different devices. The sudden change in power is monitored to determine the start and stop time of the equipment. The non-invasive load monitoring (NILM) method is used to analyze the load data and obtain device-level electricity consumption information. After identification, data including the start and stop time of the equipment and the operating time are extracted as the basis for analyzing the user's electricity consumption behavior. The data acquisition module cleans and preprocesses the data, stores the cleaned data in the database, and provides it to subsequent modules through an API or data interface.

[0024] The user load clustering analysis module classifies users' overall electricity consumption patterns to explore the typical electricity consumption behavior characteristics of different user groups and identify the overall electricity consumption patterns of different types of users for user grouping and labeling, energy-saving potential assessment, and demand response design. This module clusters user load curves through machine learning algorithms and combines them with time series feature analysis to identify the daily electricity consumption patterns of different types of users.

[0025] The load curve data of the user load cluster analysis module is stored in the form of time series, with 96 sampling points per day and a sampling interval of 15 minutes. In order to adapt to cluster analysis, the module converts the load data into a standard format:

[0026] For multiple days of load data of a single user i

[0027]

[0028] Where: P i,t (d) represents the power value of user i at time point t on day d, where D is the number of observation days for the user (e.g., 30 days).

[0029] The device-level behavior analysis module analyzes the energy consumption proportion of device types based on the non-intrusive load monitoring (NILM) method, classifies residential electrical equipment by scenario, and analyzes the proportion of each type of equipment in the user's overall electricity consumption; by counting the equipment operating time and power consumption in different scenarios, and combining it with the user's overall load data, the energy consumption contribution rate of each scenario is calculated.

[0030] The device-level behavior analysis module mines preferences for single-device usage periods by analyzing patterns in how users use specific devices in different time periods to identify regular device usage periods, peak usage periods, and abnormal usage behaviors; and by constructing a time series probability model, it captures the periodic characteristics and random fluctuations of device usage.

[0031] The device-level behavior analysis module mines multi-device collaborative patterns by analyzing the collaborative usage of household electrical devices to identify usage associations and temporal combination patterns between different devices. Based on an improved Apriori algorithm, this module incorporates time series features on top of traditional association rule mining to mine high-frequency collaborative usage patterns between devices and further analyze their temporal dependencies to reveal users' true electricity usage habits.

[0032] The user profile building module builds a multi-dimensional, fine-grained user electricity behavior profile by integrating the user load curve clustering results with device-level behavior characteristics. The generated user profile includes:

[0033] Overall load characteristics: user power consumption, load curve shape, peak and off-peak periods;

[0034] Equipment energy consumption structure: the contribution ratio of each equipment type to the total energy consumption;

[0035] Device usage preferences: the device's common startup time and usage probability during the day;

[0036] Device collaboration mode: Common multi-device linkage modes and time sequence dependencies;

[0037] Radar charts, pie charts and heat maps are used to intuitively display users' overall load, equipment energy consumption, time preferences and collaboration patterns.

[0038] The present invention also includes a system application module, which formulates a personalized energy efficiency optimization plan through a comprehensive analysis of the user's overall power consumption pattern, device-level behavior characteristics and collaborative pattern.

[0039] A user electricity consumption behavior profiling method based on non-intrusive load monitoring adopts the above-mentioned system and includes the following steps: a data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to parse device-level electricity consumption information; a user load clustering analysis module uses a clustering algorithm to classify user load curves and determine user load curve clustering results; a device-level behavior analysis module determines device-level behavior characteristics based on the device-level electricity consumption information; and a user profile construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user electricity consumption behavior profile.

[0040] A computer-readable storage medium stores a computer program for executing the above-mentioned method for profiling user electricity consumption behavior based on non-invasive load monitoring.

[0041] The present invention is applied to fields such as power demand response optimization, user energy-saving suggestion generation, and grid load forecasting, providing data support for energy management and user services.

[0042] The beneficial effects of this invention include: By integrating user load curve cluster analysis with device-level electricity usage data, fine-grained modeling of user electricity usage behavior is achieved, with a particular focus on multi-dimensional mining of device type energy usage proportions, single-device usage time preferences, and multi-device collaboration patterns. This invention can be applied to areas such as power demand response optimization, user energy-saving recommendation generation, and grid load forecasting, providing data support for energy management and user services. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described below through embodiments with reference to the accompanying drawings.

[0045] A user electricity consumption behavior profiling method based on non-intrusive load monitoring is proposed. The data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to parse device-level electricity consumption information. The user load clustering analysis module uses a clustering algorithm to classify user load curves and determine the user load curve clustering results. The device-level behavior analysis module determines the device-level behavior characteristics based on the device-level electricity consumption information. The user profile construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user electricity consumption behavior profile.

[0046] A computer-readable storage medium stores a computer program for executing the above-mentioned method for profiling user electricity consumption behavior based on non-invasive load monitoring.

[0047] A user electricity consumption behavior portrait system based on non-intrusive load monitoring includes a data acquisition module, a user load clustering analysis module, a device-level behavior analysis module and a user portrait construction module. The data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to parse out device-level electricity consumption information; the user load clustering analysis module uses a clustering algorithm to classify user load curves and determine the user load curve clustering results; the device-level behavior analysis module determines the device-level behavior characteristics based on the device-level electricity consumption information; the user portrait construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user electricity consumption behavior portrait.

[0048] The data acquisition module is responsible for collecting users' smart meter data and analyzing device-level electricity usage information through the non-intrusive load monitoring (NILM) method. The module's tasks include: collecting raw electricity usage data, including total load data; analyzing the operating information of various electrical appliances, including start and stop times, operating duration, and power characteristics; and performing data cleaning and preprocessing.

[0049] The data acquisition module collects raw electricity usage data and obtains the total load data of the user's household through smart meters or distributed energy consumption monitoring terminals. The smart meter records the user's power consumption, voltage, and current information at a high frequency (such as 1Hz or higher) and stores it in a local or cloud database. The data is continuously recorded at a set sampling interval to ensure data integrity. When a sudden change in electricity usage (such as a sudden increase or decrease in power) is detected, high-frequency data acquisition is triggered to capture the start and stop events of appliances.

[0050] The data acquisition module analyzes the operating information of various electrical appliances, decomposes the total load data through the non-invasive load monitoring (NILM) method, and identifies the power consumption patterns of the main electrical appliances in the home. This process can use existing pattern matching, power feature comparison or machine learning methods. Different devices are identified by using the power change characteristics of electrical appliances (such as peak power, steady-state power, etc.); sudden changes in power are monitored to determine the start and stop time of the device; load data is analyzed using existing non-invasive load monitoring (NILM) methods (such as FHMM, Seq2Seq, etc.) to obtain device-level power consumption information; after identification is completed, data including device start and stop times and operating duration are extracted as the basis for analyzing user power consumption behavior.

[0051] The data acquisition module cleans and preprocesses the data, storing the cleaned data in a database and providing it to subsequent modules via an API or data interface. Because smart meter data may contain missing values, outliers, or noise, missing data points are filled using missing value processing methods such as interpolation, mean filling, or adjacent time point completion, and threshold detection methods are used to identify outliers to ensure data accuracy and stability.

[0052] The goal of the user load clustering analysis module is to classify the overall electricity consumption patterns of users in order to explore the typical electricity consumption behavior characteristics of different user groups. This module clusters user load curves through machine learning algorithms, combines them with time series feature analysis, identifies the daily electricity consumption patterns of different types of users, and provides a basis for subsequent user profile construction.

[0053] The load curve data of the user load cluster analysis module is stored in the form of time series, with 96 sampling points per day (15-minute sampling interval); in order to adapt to cluster analysis, the module converts the load data into a standard format:

[0054] For multiple days of load data of a single user i

[0055]

[0056] Where: P i,t (d) represents the power value of user i at time point t on day d, where D is the number of observation days for the user (e.g., 30 days).

[0057] In order to improve the effect of the clustering algorithm, the feature extraction of the user load clustering analysis module requires the extraction of key user load features, including load statistical features and time series features, in addition to the original load curve data.

[0058] The load statistical characteristics are shown in the following table:

[0059]

[0060] The timing characteristics are shown in the table below:

[0061]

[0062] The clustering effect under different K values ​​is evaluated by methods such as the elbow rule or silhouette coefficient, and the optimal K value is selected; the silhouette coefficient measures the compactness within the same cluster and the separation between different clusters.

[0063] Load curve clustering, based on the time series characteristics of the load curve, DTW is used to measure the curve similarity, combined with K-means for clustering, and multiple evaluation indicators are used to evaluate the clustering quality to determine the optimal cluster number k.

[0064] Calinski-Harabasz index: assesses inter-cluster separation and intra-cluster compactness.

[0065] Davies-Bouldin index: measures the similarity between clusters. The smaller the value, the better the clustering effect.

[0066] Finally, several typical daily load curves are obtained.

[0067] The device-level behavior analysis module analyzes the energy consumption contribution of each device type based on the non-intrusive load monitoring (NILM) method. This classifies residential electrical devices by scenario and analyzes the contribution of each type of device to the user's overall electricity consumption. By analyzing the device operating time and power consumption in different scenarios and combining it with the user's overall load data, the energy consumption contribution rate of each scenario is calculated, providing fine-grained support for profiling user electricity behavior.

[0068] (1) Equipment scenario classification

[0069] In order to analyze users' electricity usage behavior more specifically, household appliances are divided into the following five categories according to usage scenarios:

[0070] Infrastructure: Equipment that maintains basic home operations, such as lighting, routers, air conditioners (constant temperature), refrigerators, etc.

[0071] Entertainment and leisure category: equipment mainly used for leisure and entertainment, such as televisions, game consoles, stereos, projectors, etc.

[0072] Kitchen cooking: equipment used for food processing and cooking, such as rice cookers, microwave ovens, induction cookers, ovens, etc.

[0073] Personal care: equipment related to personal health and care, such as hair dryers, electric shavers, massagers, etc.

[0074] Cleaning and care products: equipment related to housework, such as washing machines, dishwashers, vacuum cleaners, etc.

[0075] (2) Calculation of equipment energy consumption

[0076] ①Calculate the energy consumption of each device type

[0077] Based on the device power data, calculate the total energy consumption of each type of equipment at different time periods:

[0078]

[0079] in:

[0080] E type is the total daily energy consumption (kWh) of a certain equipment category;

[0081] P type (t) is the power of the equipment of this category at time t (W);

[0082] Δt is the sampling time interval (hours). If the data is at 15-minute intervals, then Δt = 0.25 hours.

[0083] ②Calculate the energy consumption ratio of equipment types

[0084]

[0085] in:

[0086] E total is the total electricity consumption for the day;

[0087] R type Reflects the energy consumption ratio of this type of equipment;

[0088] ③ Statistics on equipment energy consumption in different time periods

[0089] To analyze the energy consumption distribution of devices at different times, we divide a day into multiple periods and calculate the energy consumption share of different device categories in each period:

[0090]

[0091] Time Time Range early morning 00:00-06:00 morning 06:00-12:00 afternoon 12:00-18:00 night 18:00-24:00

[0092] The device-level behavior analysis module's single-device usage period preference mining aims to analyze patterns in user usage of specific devices over different time periods, identifying regular usage periods, peak usage periods, and abnormal usage behaviors. By constructing a time series probability model, it captures the cyclical characteristics and random fluctuations of device usage, providing data support for intelligent load management, energy efficiency optimization, and user behavior modeling.

[0093] (1) Constructing device usage probability distribution

[0094] Count the usage probability of each device at 96 time points in a day:

[0095]

[0096] P(t) is the probability that the device is in use at time t;

[0097] S(d,t) indicates whether the device is turned on at time t on day d (1 for on, 0 for off);

[0098] N is the number of days included in the data;

[0099] (2) Identify high-frequency periods of device use

[0100] By analyzing the device usage probability distribution, we can determine the high-frequency usage periods (main active times):

[0101] Definition of peak usage period: When the device usage probability is higher than P_mean + 1.5*std (mean + 1.5 times the standard deviation), this time period is considered a high-frequency usage period.

[0102] The device usage probability is lower than P_mean-1.5*std, which can be regarded as a low activity period.

[0103] (3) Classification of device usage patterns

[0104] Based on the device usage time distribution curve, the device usage pattern can be classified

[0105]

[0106] The device-level behavior analysis module's multi-device collaborative pattern mining aims to analyze the collaborative usage of household electrical devices and identify usage associations and temporal combination patterns between different devices. Based on an improved Apriori algorithm, this module introduces time series features to traditional association rule mining. It mines high-frequency collaborative usage patterns between devices and further analyzes their temporal dependencies to reveal users' true electricity usage habits.

[0107] (1) Equipment usage data modeling

[0108] To mine device collaboration patterns, we need to construct appropriate transaction data and time series data. We convert the user's daily load data into a set of transactions. Each transaction represents device usage within a specific time window. Each row represents a time window, and the device's on / off status within that window is represented by 1 (on) and 0 (off).

[0109] Time window air conditioner television Audio washing machine dryer … 00:00-00:15 1 0 0 0 0 … 00:15-00:30 1 1 1 0 0 … 00:30-00:45 1 1 1 1 0 … … … … … … … …

[0110] To analyze the timing dependencies of devices, it is necessary to record the device startup time. The device startup time can be used to analyze the start and stop order of different devices, and the device duration can be used to determine the overlapped operation time between devices.

[0111] equipment Startup time Duration (minutes) air conditioner 00:30 120 television 00:45 90 Audio 00:45 60 washing machine 01:15 45 dryer 01:15 30

[0112] (2) Improved Apriori association rule algorithm

[0113] The Apriori algorithm uses frequent item set mining to identify high-frequency co-occurrence patterns between devices:

[0114] Build a transaction database, where each transaction represents a set of enabled devices within a certain time window;

[0115] Calculate the support of device combinations:

[0116]

[0117] Calculate the confidence (the probability that device B appears when device A is turned on):

[0118]

[0119] Set the minimum support and minimum confidence to filter out highly relevant device combinations;

[0120] Generate device usage association rules, for example: {air conditioner} → {television} (support 0.6, confidence 0.85);

[0121] {Washing machine} → {Dryer} (support 0.4, confidence 0.95)

[0122] In order to consider the time dependency of device usage, the Apriori algorithm is improved to mine the timing correlation pattern of devices. The time interval between the startup of devices A and B is calculated as follows:

[0123] Time_Gap(A,B)=T B -T A

[0124] Set a maximum time window (e.g., 60 minutes) to filter out strongly associated device pairs with a short time interval. While standard support only considers whether devices appear together, sequential support considers the start and stop order of devices: for example: washing machine → 30 minutes → dryer (sequential support 0.85); TV → 15 minutes → stereo (sequential support 0.75).

[0125]

[0126] Use sliding windows to count the start and stop sequences of devices and extract typical device collaboration patterns. The mined device collaboration patterns can be applied to:

[0127] Smart home energy-saving optimization uses time-series association rules to adjust device start-stop logic, such as automatically delaying the start of a dryer to avoid peak grid load. Personalized device control, such as automatically activating the speaker when the TV is turned on, improves the user experience.

[0128] The user profile construction module aims to build a multi-dimensional, fine-grained user electricity consumption behavior portrait by integrating the user load curve clustering results with device-level behavior characteristics. This portrait not only reflects the user's overall electricity consumption pattern (such as peak hours, load shape, etc.), but also reveals the user's preferences and coordination rules in the use of specific devices. The ultimate goal of this module is to provide power companies or energy management systems with accurate user portraits to support personalized energy efficiency optimization, demand response strategy formulation, and smart home control. The generated user portraits include:

[0129] Overall load characteristics: user power consumption, load curve shape, peak and off-peak periods;

[0130] Equipment energy consumption structure: the contribution ratio of each equipment type to the total energy consumption;

[0131] Device usage preferences: the device's common startup time and usage probability during the day;

[0132] Device collaboration mode: Common multi-device linkage modes and time sequence dependencies;

[0133] Using radar charts, pie charts, heat maps and other charts, it intuitively displays the user's performance in dimensions such as overall load, equipment energy consumption, time preference and collaboration mode.

[0134] The present invention also includes a system application module, which is a terminal application layer and converts the multi-dimensional information generated by the data acquisition module, the user load clustering analysis module, the device-level behavior analysis module and the user portrait construction module into an operational decision-making basis; through a comprehensive analysis of the user's overall power consumption pattern, device-level behavior characteristics and collaborative pattern, the module supports power companies in formulating personalized energy efficiency optimization plans, and significantly improves the demand-side response capability, thereby realizing intelligent regulation and precise energy-saving management.

[0135] (1) Development of personalized energy efficiency optimization plan

[0136] Refined user grouping: Based on load curve clustering and device behavior characteristics, users are divided into different types, such as peak type, balanced type, and low energy consumption type. Exclusive energy-saving plans are developed for different groups;

[0137] Energy consumption composition optimization: Combined with energy consumption analysis by device type, this system provides users with energy efficiency optimization suggestions based on the device's perspective. For example, it provides optimization strategies or replacement suggestions for high-energy-consuming devices in the home (such as air conditioners and water heaters) to reduce energy consumption at the source.

[0138] (2) Demand-side response

[0139] Real-time load monitoring and early warning: Through real-time data collection and analysis, the system monitors peak power consumption across the entire network and in local areas. The system can automatically identify abnormal loads and sudden increases, and issue timely early warnings, helping power companies take regulatory measures in advance.

[0140] Incentive mechanism design: Based on the device usage preferences and collaboration patterns reflected in user profiles, we accurately identify user groups suitable for demand response. Through customized incentive programs (such as electricity rate discounts and point rewards), we encourage users to adjust their electricity usage during off-peak hours, thereby balancing supply and demand and alleviating pressure on the power grid.

[0141] (3) Decision support and information visualization

[0142] Data dashboard and reporting system: The system provides an intuitive visual interface that displays key information such as user grouping, power consumption curves, device collaboration modes, and time series change trends. Managers can use the interactive dashboard to monitor power consumption in each area in real time and make quick decisions based on charts and reports.

[0143] Scenario simulation and prediction: Using historical data and machine learning models, we simulate electricity consumption under different regulation strategies, predict the effect of demand-side response, and provide a scientific basis for power companies to formulate dynamic regulation strategies.

[0144] (4) Intelligent control and feedback mechanism

[0145] Smart home interconnection: Connect user profiles with the smart home control platform to achieve automatic device adjustment. For example, during peak demand response periods, the system can control the operating status of air conditioners, water heaters, and other devices to reduce overall power load.

[0146] Feedback and continuous optimization: The system records user responses and energy-saving effects after energy efficiency optimization in real time, and continuously updates user profiles and control strategies to form a closed-loop feedback loop to ensure continuous improvement and optimization.

Claims

1. A user electricity usage behavior profiling system based on non-intrusive load monitoring, characterized by: It includes data acquisition module, user load clustering analysis module, device-level behavior analysis module and user portrait construction module. The data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to analyze device-level power consumption information. The user load clustering analysis module uses a clustering algorithm to classify user load curves and determine the clustering results of user load curves. The device-level behavior analysis module determines the device-level behavior characteristics based on the device-level power consumption information. The user portrait construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user power consumption behavior portrait.

2. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 1 is characterized by: The data acquisition module is responsible for collecting users' smart meter data and analyzing device-level electricity usage information through the non-intrusive load monitoring (NILM) method. The module's tasks include: collecting raw electricity usage data, including total load data; analyzing the operating information of various electrical appliances, including start and stop times, operating duration, and power characteristics; and performing data cleaning and preprocessing.

3. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 2 is characterized by: The data acquisition module collects raw electricity consumption data and obtains the total load data of the user's home through a smart meter or distributed energy consumption monitoring terminal; the smart meter records the user's electricity power, voltage and current at a high frequency and stores them in a local or cloud database; continuously records data at a set sampling interval; when a sudden change in electricity consumption is detected, high-frequency acquisition is triggered to capture the start and stop events of electrical appliances; the data acquisition module analyzes the operating information of various electrical appliances, decomposes the total load data through the non-intrusive load monitoring (NILM) method, and identifies the electricity consumption patterns of major electrical appliances in the home; uses the power change characteristics of electrical appliances to identify different devices; monitors power mutations to determine the start and stop time of the equipment; uses the non-intrusive load monitoring (NILM) method to analyze load data and obtain device-level electricity consumption information; After identification is completed, data including equipment start and stop time and operating time are extracted as the basis for user electricity consumption behavior analysis.

4. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 1 is characterized by: The user load clustering analysis module classifies users' overall electricity consumption patterns to explore the typical electricity consumption behavior characteristics of different user groups and identify the overall electricity consumption patterns of different types of users for user grouping and labeling, energy-saving potential assessment, and demand response design. This module clusters user load curves through machine learning algorithms and combines them with time series feature analysis to identify the daily electricity consumption patterns of different types of users.

5. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 4 is characterized by: The load curve data of the user load cluster analysis module is stored in the form of time series, with 96 sampling points per day and a sampling interval of 15 minutes. In order to adapt to cluster analysis, the module converts the load data into a standard format: For a single user Multi-day load data ； Where: P i,t (d) represents the power value of user i at time point t on day d, and D is the number of observation days for the user.

6. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 1 is characterized by: The device-level behavior analysis module analyzes the energy consumption proportion of device types based on the non-intrusive load monitoring (NILM) method, classifies residential electrical equipment by scenario, and analyzes the proportion of each type of equipment in the user's overall electricity consumption; by counting the equipment operating time and power consumption in different scenarios, and combining it with the user's overall load data, the energy consumption contribution rate of each scenario is calculated.

7. The user electricity behavior profiling system based on non-intrusive load monitoring according to claim 1 is characterized by: The device-level behavior analysis module mines preferences for single-device usage periods by analyzing patterns in how users use specific devices in different time periods to identify regular device usage periods, peak usage periods, and abnormal usage behaviors; and by constructing a time series probability model, it captures the periodic characteristics and random fluctuations of device usage.

8. The user electricity usage behavior profiling system based on non-intrusive load monitoring according to claim 1 is characterized by: The user profile building module builds a multi-dimensional, fine-grained user electricity consumption behavior profile by integrating the user load curve clustering results with device-level behavior characteristics; The generated user portraits include: Overall load characteristics: user power consumption, load curve shape, peak and off-peak periods; Equipment energy consumption structure: the contribution ratio of each equipment type to the total energy consumption; Device usage preferences: the device's common startup time and usage probability during the day; Device collaboration mode: Common multi-device linkage modes and time sequence dependencies; Radar charts, pie charts and heat maps are used to intuitively display users' overall load, equipment energy consumption, time preferences and collaboration patterns.

9. A user electricity usage behavior profiling method based on non-intrusive load monitoring, using the user electricity usage behavior profiling system based on non-intrusive load monitoring according to any one of claims 1 to 8, characterized in that The system includes the following steps: the data acquisition module collects user smart meter data and uses the non-intrusive load monitoring (NILM) method to parse the device-level electricity consumption information; the user load clustering analysis module uses a clustering algorithm to classify user load curves and determine the user load curve clustering results; the device-level behavior analysis module determines the device-level behavior characteristics based on the device-level electricity consumption information; the user portrait construction module integrates the user load curve clustering results with the device-level behavior characteristics to construct a multi-dimensional user electricity consumption behavior portrait.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method for profiling user electricity consumption behavior based on non-intrusive load monitoring as described in claim 9.

Images