A method for recommending missed measurement points based on the propagation of measurement tree bias

By constructing a hierarchical measurement tree and combining sliding window filtering, adaptive thresholding, and cross-validation, the system automatically locates unmeasured points, solving the problem of identifying unmeasured points in the sub-item measurement system and achieving efficient and accurate recommendation of missed measurement points.

CN123115848APending Publication Date: 2026-11-17NANJING RONGSHU AUTOMATION SYST CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611234660.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

During the construction and improvement phase of the sub-item metering system, how can we automatically and reliably identify unmetered points and recommend replacement locations without relying on full terminal coverage, so as to eliminate systematic deviations between the master meter and the sum of known sub-items?

Method used

A hierarchical metering tree is constructed, and by synchronously collecting electricity values, linear interpolation, sliding window filtering, adaptive thresholding, and hierarchical reverse lookup algorithms, combined with cross-validation, the automatic location and recommendation of unmetered points can be achieved.

Benefits of technology

It effectively suppresses instantaneous load fluctuations and meter freezing noise, reduces false alarms, narrows the scope of investigation, improves efficiency, ensures the interpretability and quantifiability of test results, and facilitates operation and maintenance decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115848A_ABST
    Figure CN123115848A_ABST
Patent Text Reader

Abstract

This invention relates to the fields of energy management, electricity metering, and power distribution monitoring. It discloses a method for recommending missed metering points based on metering tree deviation propagation. The method includes constructing a hierarchical metering tree, synchronously collecting and calculating electricity consumption, calculating node deviation rates, skipping cycles under low load, sliding window t-test, dynamic adaptive thresholding, recursive unexplained deviation, and cross-validation of the recommended list. This method treats metering coverage integrity diagnosis as an independent problem, systematically identifying missed metering rather than electricity theft. It suppresses spurious deviations caused by instantaneous fluctuations and clock asynchrony through sliding window statistical filtering and adaptive thresholding, reducing invalid investigations. It recursively locks unmetered levels through a hierarchical reverse lookup algorithm, narrowing the scope of verification. It eliminates the risk of division by zero and noise amplification through threshold upper and lower limits and a low-load suspension mechanism, ensuring numerical stability and reducing false alarms. Cross-validation and confidence scoring make the results interpretable and quantifiable, facilitating prioritization and re-installation decisions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of energy management, electricity sub-metering and power distribution monitoring, specifically a method for recommending missed metering points based on metering tree deviation propagation. Background Technology

[0002] Sub-item metering is the foundation of refined energy management. It breaks down electricity consumption within the same power distribution circuit into multiple sub-items such as lighting, air conditioning, and power, based on function, area, or equipment type, and measures them separately to support energy efficiency analysis, quota management, and carbon emission accounting. Sub-item metering presupposes that the metering tree (from the main meter to each individual collecting meter) fully covers the actual electricity load. However, in engineering practice, sub-item metering coverage is usually implemented gradually: initially, meters are only installed for major loads such as air conditioning, lighting, and power, while a large number of scattered loads (such as sockets, small equipment, temporary power, network cabinets, etc.) are not included in the sub-item metering scope for a long time, resulting in a systematic deviation between the main meter reading and the sum of the known sub-items.

[0003] This type of deviation differs from deviations caused by line loss or electricity theft in both its cause and handling. It is not due to losses on the transmission line, nor is it illegally stolen; rather, it represents normal electricity consumption that has been consumed but not individually metered by any meter. It needs to be eliminated by installing additional metering points. Existing metering diagnostic technologies primarily focus on line loss location and electricity theft detection, generally relying on a step-by-step comparison method between the master meter and user meters, or waveform feature analysis, to locate anomalies. These methods assume full coverage of end-user meters, which is fundamentally different from the aforementioned project scenario of gradual coverage of individual meters. Therefore, how to automatically and reliably determine where there is unmetered electricity, provide confidence levels, and recommend replacement locations based on the deviation between the master meter and the sum of known individual meters, without relying on full end-user coverage, is a practical technical problem currently faced by individual metering systems during their construction and improvement phases. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for recommending missed metering points based on metering tree deviation propagation. It primarily solves the problem of how to automatically and reliably locate unmetered items and their corresponding metering tree levels from the deviation between the total meter reading and the sum of known sub-items when metering coverage is incomplete during the construction and improvement phases of a sub-metering system. Furthermore, it distinguishes between genuine unmetered items and pseudo-deviations caused by instantaneous load fluctuations or metering asynchrony, and provides recommended solutions for supplementing metering points, achieving complete automatic diagnosis from deviation detection to point recommendation.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for recommending missed measurement points based on the propagation of measurement tree bias includes the following steps: S1: Construct a hierarchical measurement tree; the hierarchical measurement tree takes the master table as the root node and is divided into first-level items and second-level items according to function to form a tree structure; S2: Synchronously collect the electricity freezing value of each node, and use linear interpolation to normalize the electricity value of each node to the same reference time; S3: For each metered parent node, calculate its node deviation and deviation rate; the node deviation is the difference between the parent node's electricity value and the sum of the electricity values ​​of all its child nodes, and the deviation rate is the ratio of the node deviation to the parent node's electricity value. S4: When the parent node's cycle power is lower than the preset lower limit, skip the deviation calculation for that cycle, and that cycle will not be included in the valid samples for subsequent statistics. S5: Calculate the window mean of the deviation rate within the sliding window, and perform a one-sample t-test on the sequence within the window. Only when the window mean is greater than the adaptive threshold and the t-test is significant, it is determined to be a persistent significant deviation. S6: Dynamically calculate the adaptive threshold based on the basic threshold, baseline load, current power, lower limit of the denominator, and time period coefficient; S7: When the root node is determined to have a persistent significant deviation, the unexplained deviation of each node is recursively calculated. The unexplained deviation is the difference between the node deviation of the node and the sum of the node deviations of all its measured child nodes. If the unexplained deviation is greater than the preset tolerance, it is determined that there are unmeasured sub-items in the level where the node is located. S8: Cross-validate the judgment results with equipment ledger changes, recent work orders and time period consistency information, and output a list of unmetered items with confidence scores and recommended metering points based on the cross-validation results.

[0006] Preferably, in step S2, the synchronous acquisition is based on NTP time synchronization results, and the electricity values ​​of each node are uniformly calculated to the same reference time through linear interpolation.

[0007] Preferably, in step S5, the width of the sliding window is W, and the window mean of the deviation rate is calculated within the sliding window of width W. Perform a one-sample t-test on the sequences within the window to test whether the window mean is significantly greater than the adaptive threshold. Only when the window mean Greater than the adaptive threshold If the t-test is significant, it is determined to be a persistent significant deviation; the width W of the sliding window is a configurable parameter.

[0008] Preferably, in step S6, the adaptive threshold The calculation formula is: in, Based on the threshold, As the baseline load, For the current power, The lower limit of the denominator For time period coefficients, This represents the upper limit of the threshold.

[0009] Preferably, in step S4, the preset lower limit is an absolute lower limit of battery power. When the parent node's periodic power... Less than When this period occurs, the deviation calculation for that period is skipped, and that period is not included in the effective sample count of the sliding window.

[0010] Preferably, in step S7, the unexplained bias The calculation formula is: in, parent node The node deviation, parent node All metered child nodes The sum of node deviations; like Greater than the preset tolerance If so, it is determined that there are unmeasured sub-items in the level where the node is located.

[0011] Preferably, in step S8, the cross-validation includes: ledger comparison, querying equipment ledger change records; work order comparison, querying recent temporary power consumption or construction work orders; and time period consistency check, determining whether the deviation only occurs in a fixed time period.

[0012] Preferably, in step S8, the confidence score is calculated based on the normalized bias, cross-validation score, and volatility penalty term.

[0013] Compared with existing technologies, this invention provides a method for recommending missed metering locations based on metering tree deviation propagation, which has the following advantages: This method separates metering coverage integrity diagnosis from traditional line loss analysis and electricity theft detection, and proposes it as an independent diagnostic problem, providing a systematic method for discovering missed metering rather than electricity theft; it effectively suppresses spurious deviations caused by instantaneous load fluctuations and asynchronous meter freezing times through sliding window statistical filtering and adaptive thresholds, reducing ineffective investigations by maintenance personnel; it recursively locks the level of unmetered locations from coarse to fine through a hierarchical reverse lookup positioning algorithm, narrowing the scope requiring manual verification and improving investigation efficiency; it fundamentally eliminates the risk of division by zero and noise amplification under extremely low or zero loads through threshold upper and lower limits and a low-load suspension mechanism, ensuring the numerical stability of the judgment and further reducing false alarms; cross-validation and confidence scoring make the detection results interpretable and quantifiable, facilitating priority ranking and reinstallation decisions by maintenance personnel. Attached Figure Description

[0014] Figure 1This is a flowchart illustrating the overall method of the method for recommending missed measurement points based on the deviation propagation of the measurement tree in this invention. Detailed Implementation

[0015] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0016] The following is in conjunction with the accompanying drawings. Figure 1 Taking a sub-metering system in a power distribution station as an example, the steps of the present invention will be described in detail; the embodiments do not constitute a limitation on the present invention.

[0017] Example 1: Construction of a hierarchical metering tree The electrical equipment within the station is grouped according to physical wiring and function, and a hierarchical metering tree is constructed. The root node is the main incoming meter of the power distribution room (denoted as...). M0 The tree structure is divided into several levels, with the next level being functional sub-items (e.g., lighting M1, air conditioning M2, power M3). Each sub-item can be further subdivided into secondary sub-items (e.g., air conditioning main unit M2a, air conditioning terminal M2b), and so on until the leaf node, which represents a single device or circuit with an installed meter. Nodes with installed meters are marked as metered, while logical group nodes without installed meters are marked as pending overwriting. The tree structure is stored in a relational database's sub-item table (including the parent node's code field).

[0018] Example 2: Synchronous data acquisition and node deviation calculation The scheduled task reads the total electricity consumption and frozen values ​​of each metered node from the time-series database every cycle (default 15 minutes, aligned with the meter freeze cycle). To eliminate comparison errors caused by clock drift between different meters, linear interpolation is used to uniformly calculate the electricity consumption of each node to the same reference time based on NTP time synchronization results. .

[0019] For each metered parent node in the metering tree Calculate its nodal deviation Deviation rate : in, This represents the electricity consumption reading of the parent node during period t. This represents the electricity reading for the child node. The value range is [0,1], with values ​​close to 0 indicating complete coverage; a positive deviation indicates that the sum of the child nodes is less than that of the parent node, indicating that there are unmetered loads; a negative deviation indicates that the sum of the child nodes is greater than that of the parent node (which may indicate metering errors, abnormal CT wiring, or data asynchrony). This invention mainly handles the positive deviation case.

[0020] Example 3: Sliding Window Timing Statistical Filtering Original deviation rate The current data contains a significant amount of noise caused by instantaneous load fluctuations and minute differences in meter freeze time, which can lead to numerous false alarms when directly compared to a fixed threshold. This invention employs a sliding window statistical filter to suppress instantaneous fluctuations: within a sliding window of width W hours (typically W=6), the window mean of the deviation rate is calculated. A one-sample t-test was performed on the sequences within the window to test whether the window mean was significantly greater than the adaptive threshold. Only if: > And the t-test was significant (α=0.05, one-tailed). Only when the deviation of a node is considered a sustained significant deviation is the sliding window considered to be present. Brief jumps are smoothed out by the sliding window, while systematic deviations lasting for several hours are preserved, effectively reducing the false alarm rate. The sliding window width W is configurable: W = 4–8 hours for locations with large fluctuations, and W = 2–4 hours for locations with stable conditions, balancing response speed and stability.

[0021] Example 4: Adaptive Threshold Fixed thresholds are insufficient to accommodate differences in load levels and time periods: At night, total load is low, and even small absolute deviations can result in high percentage deviation rates; during peak hours, total load is high, and even a 5% deviation represents a considerable value. This invention uses an adaptive threshold... Designed as follows: in, The basic threshold is 3% to 5%. This is the baseline load (average monthly total power). For the current power, These are time-based coefficients (1.0 for working hours, 1.3–1.5 for light nighttime loads, and 0.8–0.9 for peak hours). To avoid… When the denominator approaches zero, the ratio diverges. Set a lower limit :when When the value is less than 1% of full scale, it is calculated as 1% of full scale. =0.01× Equivalently, this can also be achieved by adding minimal positive constants in the denominator. In addition, for Set upper limit (Typical 50%), ensuring that the threshold always falls within a physically reasonable range.

[0022] Example 4-B: Low-load suspension rule When a site or node is under minimal or zero load, the deviation rate When dividing a small quantity by another small quantity, the result is almost entirely dominated by metering noise, making any judgment based on the ratio unreliable. Therefore, this invention introduces a low-load suspension mechanism: setting an absolute lower limit for the electrical charge. (For example, 1% of full scale, or several times the minimum resolution of the meter). In each judgment cycle, first check the electricity consumption of the parent node for that cycle. ,like < If the node is marked as having insufficient data / not yet determined in this period, the deviation calculation for this period is skipped, and this period is not included in the effective sample count for the continuous N periods or the sliding window W; wait until the load recovers to After the above, the system automatically resumes normal judgment. This mechanism bypasses the noise amplification caused by 0 / 0 and small-to-small-quantity division from the data source, ensuring that the statistical results are dominated by normal load periods, thus significantly reducing the false alarm rate.

[0023] Example 5: Hierarchical Reverse Lookup Location Algorithm Once the root node M0 is determined to have a persistent and significant deviation, the location should be determined layer by layer downwards by comparing the relationship between the deviation of each node and the sum of the deviations of its child nodes. Let the node... The deviation is If the set of measured child nodes is C, then calculate the unexplained bias of that node: like ≤Tolerance ,illustrate The deviation can be fully explained by the deviation of the child nodes; the unmeasured nodes are located at lower levels. Significantly greater than ,illustrate Each layer contains independent unmeasured loads. The algorithm recursively descends the tree until it finds all nodes with significant unexplained biases, i.e., void layers.

[0024] Example 6: Cross-validation and confidence scoring To avoid misclassifying legitimate temporary loads or metering asynchrony as unmetered, cross-validation is performed before output: (a) Record comparison: Check changes in equipment records. If new equipment has recently been put into production but has not been included in the measurement, the confidence level is increased; (b) Work order comparison: Check if there are any temporary power supply or construction work orders for this area recently. If so, mark them as explainable deviations and reduce the confidence level. (c) Time period consistency check: If the deviation only occurs in a fixed time period, it may be a periodic external load and is marked as a time period related deviation.

[0025] Confidence score The final output is a list of unmetered items, each containing a metering tree path, estimated unmetered electricity consumption, confidence score, time period characteristics, and recommended metering points.

[0026] Example 7: Numerical Case Study For example: The total monthly consumption M0 is 1000 kWh, and we know that lighting M1 = 300 kWh, air conditioning M2 = 350 kWh, and power M3 = 220 kWh. Root node =1000-(300+350+220)=130kWh, deviation rate 13%, exceeding... =5%. After confirming a continuous deviation of W=6h through sliding window filtering, a layered reverse lookup was initiated: Looking at layer M2, its child nodes M2a=320, M2b=30, Δ M2 =0; =130−(0+0+0)=130kWh> Conclusion: Node M0 exhibits an independent unexplained bias; the unmetered load is located at layer M0 and does not belong to any metered sub-nodes. Cross-validation revealed a recent IT cabinet power supply work order (confidence +10), with the bias evenly distributed throughout the day. Recommendation: Add two metering points at layer M0: one for IT equipment and one for socket circuits. This is estimated to recover approximately 130 kWh of unmetered electricity consumption per month.

[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for recommending missed measurement points based on the propagation of measurement tree bias, characterized in that, Includes the following steps: S1: Construct a hierarchical measurement tree; the hierarchical measurement tree takes the master table as the root node and is divided into first-level items and second-level items according to function to form a tree structure; S2: Synchronously collect the electricity freezing value of each node, and use linear interpolation to normalize the electricity value of each node to the same reference time; S3: For each metered parent node, calculate its node deviation and deviation rate; the node deviation is the difference between the parent node's electricity value and the sum of the electricity values ​​of all its child nodes, and the deviation rate is the ratio of the node deviation to the parent node's electricity value. S4: When the parent node's cycle power is lower than the preset lower limit, skip the deviation calculation for that cycle, and that cycle will not be included in the valid samples for subsequent statistics. S5: Calculate the window mean of the deviation rate within the sliding window, and perform a one-sample t-test on the sequence within the window. Only when the window mean is greater than the adaptive threshold and the t-test is significant, it is determined to be a persistent significant deviation. S6: Dynamically calculate the adaptive threshold based on the basic threshold, baseline load, current power, lower limit of the denominator, and time period coefficient; S7: When the root node is determined to have a persistent significant bias, the unexplained bias of each node is recursively calculated. The unexplained bias is the difference between the node bias of the node and the sum of the node biases of all its measured child nodes. If the unexplained deviation is greater than the preset tolerance, it is determined that there are unmeasured sub-items in the level where the node is located; S8: Cross-validate the judgment results with equipment ledger changes, recent work orders and time period consistency information, and output a list of unmetered items with confidence scores and recommended metering points based on the cross-validation results.

2. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S2, the synchronous acquisition is based on the NTP time synchronization results, and the electricity values ​​of each node are uniformly calculated to the same reference time through linear interpolation.

3. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S5, the width of the sliding window is W, and the window mean of the deviation rate is calculated within the sliding window of width W. Perform a one-sample t-test on the sequences within the window to test whether the window mean is significantly greater than the adaptive threshold. Only when the window mean Greater than the adaptive threshold If the t-test is significant, it is determined to be a persistent significant deviation; the width W of the sliding window is a configurable parameter.

4. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S6, the adaptive threshold The calculation formula is: in, Based on the threshold, As the baseline load, For the current power, The lower limit of the denominator For time period coefficients, This represents the upper limit of the threshold.

5. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S4, the preset lower limit is the absolute lower limit of battery power. When the parent node's periodic power... Less than When this period occurs, the deviation calculation for that period is skipped, and that period is not included in the effective sample count of the sliding window.

6. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S7, the unexplained bias The calculation formula is: in, parent node The node deviation, parent node All metered child nodes The sum of node deviations; like Greater than the preset tolerance If so, it is determined that there are unmeasured sub-items in the level where the node is located.

7. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S8, the cross-validation includes: ledger comparison, querying equipment ledger change records; work order comparison, querying recent temporary power consumption or construction work orders; and time period consistency check, determining whether the deviation only occurs in a fixed time period.

8. The method for recommending missed measurement points based on the propagation of measurement tree deviation according to claim 1, characterized in that: In step S8, the confidence score is calculated based on the normalized bias, cross-validation score, and volatility penalty term.