A method for determining vacant houses of residents based on daily electricity fluctuation characteristics

Abstract

The present application relates to the technical field of power big data analysis, and discloses a kind of resident empty house determination method based on daily electricity fluctuation characteristics, the method extracts the daily frozen electricity data of resident user recent 14 natural days from electricity information acquisition system, after 0 point, 8 point, 22 point collection integrity check, with 7 days as observation window, the variation coefficient of daily electricity consumption sequence is calculated as core criterion, and daily average electricity consumption, peak-valley difference index, time period distribution entropy and continuous stationary day are introduced as auxiliary index, a five-dimensional weighted scoring system is constructed, risk classification is carried out according to comprehensive score, and the state of continuous operation of equipment is identified when residence is not inhabited.The present application is based on the conventional collection data operation of existing electricity information acquisition system, without adding second-level or 1 minute-level equipment waveform collection hardware, with the advantages of simple calculation, strong interpretability and low deployment cost.

Description

Technical Field

[0001] This invention relates to the field of power data analysis technology, and more specifically, to a method for determining vacant residential properties based on the characteristics of daily electricity consumption fluctuations. Background Technology

[0002] With the acceleration of urbanization in my country, many families are living in a special situation where their homes remain unoccupied for extended periods, yet still have electricity, due to reasons such as migrant work, holiday travel, accompanying elderly relatives in the hospital, or real estate investment. These families typically do not proactively request a power outage before leaving, and appliances such as refrigerators, air purifiers, network routers, aquarium pumps, and security cameras continue to operate, exhibiting a low and stable electricity consumption pattern distinct from that of normally occupied households. In the event of an unplanned power outage, a power outage due to unpaid bills, or a sudden tripping of the circuit breaker during holidays, these cold and heat storage devices will suffer direct economic losses due to food spoilage, medication failure, and even appliance malfunctions caused by the power outage. More seriously, some users are unaware of the power outage due to prolonged absence, leading to escalating losses. Statistics show that complaints about refrigerator power outages account for 8% to 12% of household service complaints received by the power supply service hotline, becoming a significant pain point affecting customer satisfaction.

[0003] Existing power big data analytics primarily focuses on areas such as line loss identification, electricity theft detection, and load forecasting, with relatively little research on proactive identification of vacant residential properties where electrical equipment continues to operate despite being unoccupied. Traditional methods attempt to locate target users by identifying the start-stop characteristics or electrical waveforms of specific electrical equipment, such as refrigerator compressors, through non-intrusive load monitoring. However, these methods typically rely on continuous load curves or electrical waveform data with time resolutions of seconds, minutes, or even higher. Existing electricity consumption information collection systems, i.e., marketing systems, often rely on daily frozen data and 15-minute collected data, which are usually insufficient to directly support device-level waveform analysis. If implemented using existing systems, it requires high standards for collection frequency, data integrity, and data item types, and necessitates complex waveform recognition algorithms, resulting in significant engineering implementation difficulties, high deployment costs, insufficient interpretability, and hindering rapid promotion in grassroots power supply units. Therefore, there is an urgent need to provide a method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics to address these issues. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention provides a method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics. This method uses daily frozen electricity consumption data from existing electricity information collection systems as a foundation, combined with time-segmented electricity consumption data obtained by summarizing customer meter data collected every 15 minutes, as well as status information collected at 0:00, 8:00, and 22:00, to identify whether residential users are unoccupied but their equipment is continuously running. This enables proactive early warning of potential power outage losses and differentiated power supply protection services.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics, comprising:

[0007] Extract daily frozen electricity consumption data for the most recent 14 natural days from all residential users in the target area from the electricity consumption information collection system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption based on the summary of customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00, and 22:00.

[0008] The daily frozen electricity consumption data is preprocessed, including determining whether the user's data was successfully collected at 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points.

[0009] Based on the preprocessed data, the fluctuation characteristics of the user's daily electricity consumption sequence are calculated within the preset observation window as the core criterion, and the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics are calculated as auxiliary indicators.

[0010] The core criteria and the auxiliary indicators are normalized and scored respectively, and then weighted and summed to obtain a comprehensive score.

[0011] Based on the comprehensive score, risk levels are determined, and users are classified as either highly credible vacant, suspected vacant, or normally occupied.

[0012] As a preferred embodiment of the present invention, the time-segmented electricity consumption includes peak electricity consumption and valley electricity consumption for calculating the peak-valley difference index, and three-time-segment electricity consumption for calculating the time-segment distribution entropy; wherein, the peak electricity consumption is the electricity consumption from 08:00 to 22:00, and the valley electricity consumption is the electricity consumption from 22:00 to 08:00 the next day.

[0013] The electricity consumption in the three time periods is calculated using the period from 08:00 on the statistical day to 08:00 on the next day as a period distribution entropy calculation cycle, including the electricity consumption from 08:00 to 16:00 on the statistical day, the electricity consumption from 16:00 to 24:00 on the statistical day, and the electricity consumption from 00:00 to 08:00 on the next day.

[0014] As a preferred embodiment of the present invention, the fluctuation characteristic is the coefficient of variation, which is the ratio of the standard deviation to the mean of the daily electricity consumption sequence of users within the observation window.

[0015] As a preferred embodiment of the present invention, the peak-valley difference index is calculated based on the ratio of the difference between peak electricity consumption and valley electricity consumption to the total electricity consumption of the day;

[0016] The time period distribution entropy is based on the information entropy of the proportion of electricity consumption in each time period calculated from the electricity consumption in each time period.

[0017] The number of consecutive stable days refers to the maximum number of consecutive days during which the fluctuation characteristics meet the preset stability threshold.

[0018] As a preferred embodiment of the present invention, when the time-segmented electricity consumption is unavailable, the peak-valley difference characteristics and the time-segment distribution characteristics are substituted with the default intermediate score.

[0019] As a preferred embodiment of the present invention, the weight allocation of the five-dimensional weighted scoring mechanism is as follows: fluctuation characteristics account for 35%, average daily electricity consumption accounts for 20%, peak-valley difference characteristics account for 25%, time period distribution characteristics account for 10%, and continuous and stable characteristics account for 10%.

[0020] As a preferred embodiment of the present invention, the risk classification sets a first threshold of 75 points and a second threshold of 60 points. When the comprehensive score is not lower than the first threshold of 75 points, the high confidence level is determined to be that the user is in an uninhabited but continuously operating state and an active warning is triggered. When the comprehensive score is between the second threshold of 60 points and the first threshold of 75 points, it is marked as a suspected target pending manual verification. When the comprehensive score is lower than the second threshold of 60 points, it is determined to be normal residential electricity use.

[0021] As a preferred embodiment of the present invention, in the high confidence vacancy state, an extremely high risk level is further distinguished: when the comprehensive score is not lower than 85 points and the number of consecutive stable days is not less than 7 days, it is determined to be an extremely high risk level.

[0022] As a preferred embodiment of the present invention, it further includes: implementing a differentiated service strategy based on the risk classification results, and pushing the list of high-risk users to the electricity business system to achieve power supply early warning and differentiated protection;

[0023] The differentiated service strategy includes:

[0024] For extremely high-risk users, proactive telephone contact confirmation will be implemented based on the user's contact information, advance SMS notification will be sent before overdue payments are made, and they will be included in the differentiated protection list for planned power outages.

[0025] For high-risk users, we will send SMS messages to inform them of their home electricity status, push payment channels, and record the power supply protection list for the area.

[0026] For users at medium risk, they will be included in the suspected list and verified on-site during routine inspections.

[0027] This invention also provides a system for determining vacant residential properties based on daily electricity consumption fluctuation characteristics, comprising:

[0028] The data acquisition module is used to extract the daily frozen electricity consumption data of all residential users in the target area for the most recent 14 natural days from the electricity consumption information acquisition system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption obtained by summarizing customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00 and 22:00.

[0029] The data preprocessing module is used to preprocess the daily frozen electricity consumption data, including determining whether the user's data was successfully collected at the three time points of 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points.

[0030] The feature extraction module is used to calculate the fluctuation characteristics of the user's daily electricity consumption sequence within a preset observation window based on the preprocessed data as the core criterion, and to calculate the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics as auxiliary indicators.

[0031] The comprehensive scoring module is used to normalize and score the core criteria and the auxiliary indicators respectively, and then sum them by weight to obtain a comprehensive score.

[0032] The risk classification module is used to classify users based on the comprehensive score, classifying them as either in a high-confidence vacant state, a suspected vacant state, or a normal resident state.

[0033] The beneficial technical effects of this invention are:

[0034] This invention operates based on daily frozen electricity consumption data obtainable from existing electricity consumption information collection systems. The daily frozen electricity consumption data includes daily electricity consumption, time-segmented electricity consumption data summarized from customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00, and 22:00. This invention does not require the addition of second-level or 1-minute-level device-level waveform acquisition hardware, nor does it require load decomposition of specific equipment such as refrigerator compressors. It can be directly deployed in a lightweight manner based on existing marketing system data.

[0035] This invention uses the coefficient of variation (CV) of daily electricity consumption time series as the core criterion to effectively capture the abnormal stability of daily electricity consumption when electrical equipment is continuously running in unoccupied residential areas. Simultaneously, it introduces daily average electricity consumption range constraints, peak-to-valley difference index, time-period distribution entropy, and the number of consecutive stable days as auxiliary indicators to construct a five-dimensional weighted scoring system. The synergistic effect of these multi-dimensional features significantly reduces the false positive rate caused by single-threshold judgments, improving both accuracy and recall.

[0036] All judgments in this invention are based on clear statistical indicators such as standard deviation, mean, and information entropy, as well as a rule-based scoring system. The judgment results have clear mathematical and physical meanings, which are easy for account managers to understand and explain to customers, and also facilitate auditing and parameter tuning, thus overcoming the obstacles to the promotion of black box models in grassroots applications.

[0037] By identifying potential high-risk users with high confidence, this invention can proactively issue warnings before power outages due to unpaid bills, planned power outages, and power outages due to malfunctions are executed. It also works in conjunction with the power outage management system for unpaid bills, the planned power outage management platform, and the customer service system to implement differentiated service strategies such as extending the payment reminder buffer period, providing differentiated protection, and offering precise reassurance. This effectively reduces complaints such as refrigerator damage and food spoilage caused by power outages that no one is aware of, and significantly improves the customer service experience for residents. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0039] In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0040] Combination Figure 1 The present invention provides the following embodiments:

[0041] Example 1:

[0042] A method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics, comprising:

[0043] Extract daily frozen electricity consumption data for the most recent 14 natural days from all residential users in the target area from the electricity consumption information collection system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption based on the summary of customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00, and 22:00.

[0044] The daily frozen electricity consumption data is preprocessed, including determining whether the user's data was successfully collected at 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points.

[0045] Based on the preprocessed data, the fluctuation characteristics of the user's daily electricity consumption sequence are calculated within the preset observation window as the core criterion, and the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics are calculated as auxiliary indicators.

[0046] The core criteria and the auxiliary indicators are normalized and scored respectively, and then weighted and summed to obtain a comprehensive score.

[0047] Based on the comprehensive score, risk levels are determined, and users are classified as either highly credible vacant, suspected vacant, or normally occupied.

[0048] Furthermore, the time-segmented electricity consumption includes peak-segment electricity consumption and valley-segment electricity consumption for calculating the peak-valley difference index, and three-segment electricity consumption for calculating the time-segment distribution entropy; wherein, the peak-segment electricity consumption is the electricity consumption from 08:00 to 22:00, the valley-segment electricity consumption is the electricity consumption from 22:00 to 08:00 the next day; the three-segment electricity consumption includes the electricity consumption from 08:00 to 16:00, the electricity consumption from 16:00 to 24:00, and the electricity consumption from 24:00 to 08:00 the next day.

[0049] By setting peak and off-peak electricity consumption, the difference between peak and off-peak electricity consumption within a day can be characterized, providing a data basis for calculating the peak-valley difference index. By setting electricity consumption for three time periods—08:00 to 16:00, 16:00 to 24:00, and 00:00 to 08:00 the next day—the degree of dispersion of user electricity consumption across multiple time periods within a day can be characterized, providing a data basis for calculating the time period distribution entropy.

[0050] Furthermore, the fluctuation characteristic is the coefficient of variation, which is the ratio of the standard deviation to the mean of the daily electricity consumption sequence of users within the observation window.

[0051] Using the coefficient of variation as the core criterion, and leveraging its dimensionless characteristic as the ratio of standard deviation to mean, the influence of differences in benchmark electricity consumption levels among different households is eliminated. This allows two-person households and multi-person households to compare electricity consumption stability on the same scale, avoiding misjudgments caused by differences in the absolute value of electricity consumption.

[0052] Furthermore, the peak-valley difference index is calculated based on the ratio of the difference between peak electricity consumption and valley electricity consumption to the total daily electricity consumption.

[0053] The time period distribution entropy is based on the information entropy of the proportion of electricity consumption in each time period calculated from the electricity consumption in each time period.

[0054] The number of consecutive stable days refers to the maximum number of consecutive days during which the fluctuation characteristics satisfy the preset stability threshold.

[0055] Furthermore, when the time-of-use electricity consumption is unavailable, the peak-valley difference characteristic and the time-of-use distribution characteristic are substituted with the default intermediate score.

[0056] The peak-valley difference index measures the difference between peak and valley electricity consumption within a day. When people live in the area, electricity consumption during the day is significantly higher than at night, while when no one lives in the area, electricity consumption is evenly distributed throughout the day. The time period distribution entropy reflects the degree of dispersion of electricity consumption over time. When no one lives in the area, electricity consumption tends to be evenly distributed, and the entropy value tends to be at its maximum. The number of consecutive stable days excludes occasional factors such as short-term business trips, ensuring that the identified state is a long-term vacancy.

[0057] Furthermore, the weight allocation of the five-dimensional weighted scoring mechanism is as follows: fluctuation characteristics account for 35%, average daily electricity consumption accounts for 20%, peak-valley difference characteristics account for 25%, time period distribution characteristics account for 10%, and continuous and stable characteristics account for 10%.

[0058] The coefficient of variation was given the highest weight because it has the highest significance in distinguishing between resident and unresident states; the peak-valley difference index was second because it is sensitive to differences in daily electricity consumption patterns; the average daily electricity consumption constrained the equipment baseline range; and the time period distribution entropy and the number of consecutive stable days served as supplementary verification, together forming a multi-dimensional collaborative scoring system.

[0059] Furthermore, the risk classification sets a first threshold of 75 points and a second threshold of 60 points. When the comprehensive score is not lower than the first threshold of 75 points, the high confidence level is determined to be that the user is in an uninhabited but continuously operating state and an active warning is triggered. When the comprehensive score is between the second threshold of 60 points and the first threshold of 75 points, it is marked as a suspected target pending manual verification. When the comprehensive score is lower than the second threshold of 60 points, it is determined to be normal residential electricity use.

[0060] Setting a score of 75 as a high confidence threshold ensures that the vast majority of users on the warning list are genuine target users, reducing invalid on-site verifications; setting a score of 60 as a suspected threshold creates a buffer zone for manual verification; scores below 60 are considered normal electricity usage, avoiding interference with normal users and achieving precise allocation of business resources.

[0061] Furthermore, in the aforementioned high-confidence vacancy state, an extremely high risk level is further distinguished: when the comprehensive score is not lower than 85 points and the number of consecutive stable days is not less than 7 days, it is determined to be an extremely high risk level.

[0062] The additional condition of having at least 7 consecutive stable days is introduced for high-scoring individuals because users with high overall scores but short periods of stability may only be going out for short periods. A sufficiently long period of consecutive stable days is necessary to confirm that the residence is unoccupied for an extended period, thus avoiding over-warning for occasional situations such as short-term business trips.

[0063] Furthermore, it also includes: implementing differentiated service strategies based on the risk classification results, and pushing the list of high-risk users to the electricity business system to achieve power supply early warning and differentiated protection;

[0064] The differentiated service strategy includes:

[0065] For extremely high-risk users, proactive telephone contact confirmation will be implemented based on the user's contact information, advance SMS notification will be sent before overdue payments are made, and they will be included in the differentiated protection list for planned power outages.

[0066] For high-risk users, we will send SMS messages to inform them of their home electricity status, push payment channels, and record the power supply protection list for the area.

[0067] For users at medium risk, they will be included in the suspected list and verified on-site during routine inspections.

[0068] The reason for adopting differentiated service strategies for different risk levels is that users with extremely high risk are most likely to suffer power outage losses and the most severe losses, requiring rapid response at the director level; users with high risk require routine attention at the grid leader level; and users with medium risk can be included in daily inspections, thereby optimizing the allocation of limited service resources according to risk level.

[0069] Example 2:

[0070] A system for determining vacant residential properties based on daily electricity consumption fluctuation characteristics includes:

[0071] The data acquisition module is used to extract the daily frozen electricity consumption data of all residential users in the target area for the most recent 14 natural days from the electricity consumption information acquisition system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption obtained by summarizing customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00 and 22:00.

[0072] The data preprocessing module is used to preprocess the daily frozen electricity consumption data, including determining whether the user's data was successfully collected at the three time points of 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points.

[0073] The feature extraction module is used to calculate the fluctuation characteristics of the user's daily electricity consumption sequence within a preset observation window based on the preprocessed data as the core criterion, and to calculate the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics as auxiliary indicators.

[0074] The comprehensive scoring module is used to normalize and score the core criteria and the auxiliary indicators respectively, and then sum them by weight to obtain a comprehensive score.

[0075] The risk classification module is used to classify users based on the comprehensive score, classifying them as either in a high-confidence vacant state, a suspected vacant state, or a normal resident state.

[0076] Application example:

[0077] A method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics includes the following steps:

[0078] S1: Data Extraction

[0079] The system extracts daily frozen electricity consumption data for the most recent 14 calendar days from all residential users in the target area from the electricity consumption information collection system, which is also the marketing system. Smart meters automatically freeze and collect electricity consumption data once a day, and the marketing system completes data entry into the database the day after collection. Extracting 14 days of data ensures sufficient data redundancy for the subsequent 7-day observation window, allowing for a maximum of 2 days of missing data within the 14-day period.

[0080] The daily frozen electricity consumption data includes user ID, statistical date, daily electricity consumption, time-period electricity consumption summarized based on customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00, and 22:00.

[0081] Among them, daily electricity consumption is the total electricity consumption of the corresponding calendar day of the statistical date; peak electricity consumption is the electricity consumption from 08:00 to 22:00, and valley electricity consumption is the electricity consumption from 22:00 to 08:00 of the next day; the electricity consumption of the three time periods is calculated with the period from 08:00 of the statistical date to 08:00 of the next day as one time period distribution entropy calculation period, including the electricity consumption from 08:00 to 16:00 of the statistical date, the electricity consumption from 16:00 to 24:00 of the statistical date, and the electricity consumption from 00:00 to 08:00 of the next day.

[0082] S2: Data Preprocessing

[0083] The date of this round of vacancy assessment is used as the assessment date. The integrity of the user's data collection status at 0:00, 8:00, and 22:00 on this assessment date is verified. If there are data collection records at all three time points, the collected values ​​are not empty, and there are no data collection anomaly indicators, the user is considered to have completed data collection on this assessment date, and the user is retained for subsequent feature extraction steps. If there are no data collection records at any of the three time points, the collected values ​​are empty, or the data collection record has a data collection anomaly indicator, the user is considered to have failed data collection on this assessment date, and the user is removed from the assessment sample for this round. The user's daily frozen electricity consumption data for the most recent 14 calendar days is not included in the calculation of the coefficient of variation, average daily electricity consumption, peak-valley difference index, time period distribution entropy, number of consecutive stable days, and comprehensive score for this round. Removal is only used for the model assessment in this round and does not delete the original data from the electricity consumption information collection system.

[0084] S3: Feature Extraction

[0085] Based on the preprocessed data, the following characteristics are calculated for each household with a 7-day observation window. The observation window must contain at least 5 days of valid data. The observation window length can be dynamically adjusted according to the application scenario: 3 days is suitable for rapid early warning during short holidays, 5 days for regular use, 7 days is the default balanced configuration, 14 days is suitable for long-term vacancy identification, and 21 days is suitable for ultra-long-term vacancy identification.

[0086] The core criterion is the coefficient of variation. The coefficient of variation is the ratio of the standard deviation to the mean of the daily electricity consumption series within the observation window. Its calculation formula is: .

[0087] In the formula: For the first Daily electricity consumption, in kilowatt-hours; Within the observation window The arithmetic average of daily electricity consumption; Within the observation window The standard deviation of daily electricity consumption; Let the observation window length be denoted as . .

[0088] The threshold for determining the coefficient of variation is set as follows: when At that time, a high confidence level indicates that the residence is unoccupied but the electrical equipment is continuously operating; when When the condition is deemed a suspected case, further confirmation using auxiliary indicators is required; when At that time, it was determined to be a normal living situation.

[0089] The first auxiliary indicator is the average daily electricity consumption. :

[0090]

[0091] Decision constraints are ,in The kilowatt-hour is the lower limit for equipment baseline estimation and can be adjusted according to region; The kilowatt-hour is the upper limit of the equipment baseline estimate and can be adjusted seasonally. Below that... kilowatt-hours indicates that almost no equipment is operating, which is higher than If the electricity consumption is too high, it is necessary to rule out situations where people are living in the area or high-power equipment is running continuously.

[0092] The second auxiliary indicator is the peak-to-valley difference index. :

[0093]

[0094] In the formula: This represents the electricity consumption during the peak period (8:00 AM to 10:00 PM) of the day. This refers to the electricity consumption during the off-peak period (10 PM to 8 AM the following day). The total electricity consumption for the day, i.e. The threshold for judgment is... Peak and valley data are required for support.

[0095] The third auxiliary indicator is the time-period distribution entropy. The period from 08:00 on the statistical day to 08:00 on the next day is used as a calculation period for the distribution entropy. This period is divided into three time periods: 08:00 to 16:00, 16:00 to 24:00, and 00:00 to 08:00 on the next day. Let the proportion of electricity consumption in the i-th time period to the total electricity consumption in the three time periods be . ( ),but:

[0096]

[0097] In the formula: For the first The proportion of electricity consumption in each time period to the total daily electricity consumption; The number of time periods is used; in this example, it is taken as... The threshold for judgment is... ,Right now This requires time-of-use data support. When people are living in the area, the time-of-use distribution entropy is relatively low, and electricity consumption is highly concentrated in specific time periods; when people are unoccupied, the time-of-use distribution entropy approaches its maximum value, and electricity consumption is evenly distributed.

[0098] The fourth auxiliary indicator is the number of consecutive stable days. :

[0099]

[0100] That is: statistical coefficient of variation Satisfy threshold constraints The maximum number of consecutive days. The threshold for determination is... The number of days can be used as a supplementary criterion for strong confidence, to exclude short-term stability caused by occasional factors such as short-term business trips.

[0101] When time-of-use electricity is unavailable, the peak-valley difference index and time period distribution entropy Substituting the default median score, the model primarily relies on the coefficient of variation. and consecutive stable days Make a judgment.

[0102] S4: Five-dimensional weighted scoring

[0103] After normalizing the features of the above five dimensions, a weighted sum is obtained to obtain the comprehensive score. :

[0104]

[0105] In the formula: Score for coefficient of variation; The score is based on the average daily electricity consumption. The peak-to-valley difference index score; The time period distribution entropy score; The score is based on the number of consecutive stable days.

[0106] The scoring rules for each sub-indicator are as follows:

[0107] Coefficient of variation score: when He scored 20 points at the time; 15 points were scored at the time; 8 points are awarded in the first instance; 0 points are awarded in the other instances.

[0108] Daily average electricity consumption score: When He scored 20 points at the time; or 10 points are awarded for the correct answer; 0 points are awarded for any other answer.

[0109] Peak-to-valley difference index score: when He scored 20 points at the time; He scored 12 points at that time; 5 points are awarded for the correct answer; 0 points are awarded for any other answer.

[0110] Time period distribution entropy score: when He scored 20 points at the time; He scored 12 points at that time; 5 points are awarded for the correct answer; 0 points are awarded for any other answer.

[0111] Score for consecutive stable days: when He scored 20 points at the time; 15 points were scored at the time; 8 points are awarded in the first instance; 0 points are awarded in the other instances.

[0112] S5: Grading Determination and Output

[0113] Based on the overall score Risk grading was performed, with a first threshold of 75 points and a second threshold of 60 points. In time-sharing scenarios, a high-confidence assessment indicates that the user is unoccupied but the equipment is continuously running, triggering an active alert with a high or extremely high risk level. At any given time, it is marked as a suspected target, pending manual verification; the risk level is medium. The electricity usage is classified as normal residential electricity consumption and requires no intervention.

[0114] In a high-confidence vacancy state, further distinguish the extremely high risk level: when Partial and The timing was deemed extremely high-risk.

[0115] when The system outputs a warning list, along with contact information and a power consumption trend chart, triggering a proactive power supply protection warning. In time-sharing mode, a list of suspected cases is output and added to the queue for verification. No output during time-sharing.

[0116] when and The timing is unfavorable, and the risk level is extremely high. Service measures include proactively contacting the customer by phone to confirm payment, sending advance SMS notifications before outstanding payments, inclusion in the planned power outage differentiated protection list, and a response time of within one business day.

[0117] when The risk level is high. Service measures include sending SMS notifications about home electricity usage status, providing payment channels, recording the power supply protection watchlist for the area, and responding within 3 business days.

[0118] when The risk level is medium at this time. The service measure is to include the suspect in the list and conduct an on-site verification during the next routine inspection. The response time is the next inspection.

[0119] The list of high-risk users is pushed to the overdue payment power outage management system. Before the power outage is executed, it is verified whether the user is on the warning list. If so, the payment reminder buffer period is extended and the notification is strengthened.

[0120] The list of high-risk users is pushed to the planned power outage management platform and included in the power outage plan approval reference for dispatchers.

[0121] The list of high-risk users is pushed to the customer service system. When a high-risk user calls the complaint hotline, the customer service can check the warning record immediately through the contact information to achieve precise reassurance and handling.

[0122] Deploy structured query language stored procedures or scheduled tasks directly in the existing marketing system database. The results are automatically calculated and written to the early warning table the day after the data is collected. There is no need to add second-level or minute-level waveform acquisition hardware, nor is there any need to decompose the load of specific equipment, thus achieving low-cost deployment.

[0123] The warning list is confirmed by phone or in person, and the verification results are fed back to the model for subsequent model parameter optimization and threshold adjustment.

[0124] By introducing meteorological factors such as temperature and humidity as covariates, the seasonal electricity consumption benchmark is corrected, and the upper limit threshold of the daily average electricity consumption is dynamically adjusted to reduce the misjudgment rate in summer and winter.

[0125] Accumulate labeled data and gradually introduce a lightweight classifier to replace the manual scoring system to further improve recognition accuracy.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics, characterized in that, include: Extract daily frozen electricity consumption data for the most recent 14 natural days from all residential users in the target area from the electricity consumption information collection system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption based on the summary of customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00, and 22:

00. The daily frozen electricity consumption data is preprocessed, including determining whether the user's data was successfully collected at 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points. Based on the preprocessed data, the fluctuation characteristics of the user's daily electricity consumption sequence are calculated within the preset observation window as the core criterion, and the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics are calculated as auxiliary indicators. The core criteria and the auxiliary indicators are normalized and scored respectively, and then weighted and summed to obtain a comprehensive score. Based on the comprehensive score, risk levels are determined, and users are classified as either highly credible vacant, suspected vacant, or normally occupied.

2. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, The time-segmented electricity consumption includes peak-segment electricity consumption and valley-segment electricity consumption used to calculate the peak-valley difference index, as well as three-segment electricity consumption used to calculate the time-segment distribution entropy; wherein, the peak-segment electricity consumption is the electricity consumption from 08:00 to 22:00, and the valley-segment electricity consumption is the electricity consumption from 22:00 to 08:00 the next day. The electricity consumption in the three time periods is calculated using the period from 08:00 on the statistical day to 08:00 on the next day as a period distribution entropy calculation cycle, including the electricity consumption from 08:00 to 16:00 on the statistical day, the electricity consumption from 16:00 to 24:00 on the statistical day, and the electricity consumption from 00:00 to 08:00 on the next day.

3. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, The fluctuation characteristic is the coefficient of variation, which is the ratio of the standard deviation to the mean of the daily electricity consumption sequence of users within the observation window.

4. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, The peak-valley difference index is calculated based on the ratio of the difference between peak and valley electricity consumption to the total daily electricity consumption. The time period distribution entropy is based on the information entropy of the proportion of electricity consumption in each time period calculated from the electricity consumption in each time period. The number of consecutive stable days refers to the maximum number of consecutive days during which the fluctuation characteristics meet the preset stability threshold.

5. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 4, characterized in that, When the time-of-use electricity consumption is unavailable, the peak-valley difference feature and the time-of-use distribution feature are substituted with the default intermediate point.

6. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, The weighting of the five-dimensional weighted scoring mechanism is as follows: fluctuation characteristics account for 35%, average daily electricity consumption accounts for 20%, peak-valley difference characteristics account for 25%, time period distribution characteristics account for 10%, and continuous stability characteristics account for 10%.

7. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, The risk classification sets a first threshold of 75 points and a second threshold of 60 points. When the comprehensive score is not lower than the first threshold of 75 points, the high confidence level is determined to be that the user is in an uninhabited but the equipment is continuously running and an active warning is triggered. When the comprehensive score is between the second threshold of 60 points and the first threshold of 75 points, it is marked as a suspected target pending manual verification; when the comprehensive score is below the second threshold of 60 points, it is determined to be normal residential electricity use.

8. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 7, characterized in that, In the high-confidence vacancy state, the extremely high risk level is further distinguished: when the comprehensive score is not lower than 85 points and the number of consecutive stable days is not less than 7 days, it is judged as an extremely high risk level.

9. The method for determining vacant residential properties based on daily electricity consumption fluctuation characteristics according to claim 1, characterized in that, Also includes: Based on the risk classification results, a differentiated service strategy is implemented, and a list of high-risk users is pushed to the electricity business system to achieve power supply early warning and differentiated protection. The differentiated service strategy includes: For extremely high-risk users, proactive telephone contact confirmation will be implemented based on the user's contact information, advance SMS notification will be sent before overdue payments are made, and they will be included in the differentiated protection list for planned power outages. For high-risk users, we will send SMS messages to inform them of their home electricity status, push payment channels, and record the power supply protection list for the area. For users at medium risk, they will be included in the suspected list and verified on-site during routine inspections.

10. A system for determining vacant residential properties based on daily electricity consumption fluctuation characteristics, characterized in that, include: The data acquisition module is used to extract the daily frozen electricity consumption data of all residential users in the target area for the most recent 14 natural days from the electricity consumption information acquisition system. The daily frozen electricity consumption data includes user number, statistical date, daily electricity consumption, time-period electricity consumption obtained by summarizing customer meter data collected every 15 minutes, and collection status information at 0:00, 8:00 and 22:

00. The data preprocessing module is used to preprocess the daily frozen electricity consumption data, including determining whether the user's data was successfully collected at the three time points of 0:00, 8:00 and 22:00 based on the collection status information at 0:00, 8:00 and 22:00 on the same day, and removing users whose data collection failed at any of the three time points. The feature extraction module is used to calculate the fluctuation characteristics of the user's daily electricity consumption sequence within a preset observation window based on the preprocessed data as the core criterion, and to calculate the average daily electricity consumption, peak-valley difference characteristics, time period distribution characteristics, and continuous stability characteristics as auxiliary indicators. The comprehensive scoring module is used to normalize and score the core criteria and the auxiliary indicators respectively, and then sum them by weight to obtain a comprehensive score. The risk classification module is used to classify users based on the comprehensive score, classifying them as either in a high-confidence vacant state, a suspected vacant state, or a normal resident state.

Images