Electric meter misalignment detection method and system based on cumulative indication value of power utilization information collection

Abstract

The present application relates to the technical field of electric power metering and smart grid, and discloses a meter misalignment detection method and system based on power utilization information collection cumulative indication value, a misalignment calculation unit performs data validity check on a cumulative indication value sequence, and invalid samples are eliminated or marked; the cumulative indication value at the same time is taken as a sample, a linear regression model of an indication value domain with an intercept is constructed according to phase grouping; a ridge regression is used to solve the regression model; a normalized equation group coefficient matrix after centering is constructed, decomposition is performed on the coefficient matrix, a diagonal element sequence of the decomposition process is monitored to determine the health degree of the design matrix, and a hierarchical self-healing strategy is triggered according to the health degree until the measurement transfer coefficients of each sub-meter are successfully solved or a pathological alarm is output; the misalignment proportion is calculated according to the measurement transfer coefficients of each sub-meter, the misalignment proportion is compared with a preset hierarchical threshold, a sub-meter misalignment determination result is output and reported; and the present application realizes meter misalignment identification and phase attribution determination.

Description

Technical Field

[0001] This invention relates to the field of power metering and smart grid technology, specifically to a method and system for detecting meter inaccuracies based on the cumulative readings of electricity consumption information. The method involves high-frequency acquisition of meter data via RS-485 bus and the use of a regularized linear regression algorithm to identify meter inaccuracies and determine phase attribution. Background Technology

[0002] In existing electricity metering systems, the detection of meter inaccuracy (measuring deviation exceeding tolerance) mainly relies on the following methods: (1) Regular on-site verification: Staff bring standard devices to the site to compare each meter. It is accurate but costly and inefficient, and cannot cover a large number of meters in operation. (2) Data analysis based on the main station side of the electricity consumption information collection system: Relying on the electricity consumption information collection system for power users, the collection terminal / concentrator remotely reports and aggregates the meter data to the main station. The main station constructs a set of equations based on the energy conservation relationship of the transformer area (total meter power consumption ≈ sum of power consumption of each sub-meter + line loss + fixed loss), and uses least squares or regularized regression to solve for the error coefficients of each meter. This type of scheme often uses daily frozen power consumption or time-period incremental power consumption as the modeling object; (3) Pulse comparison method: Collect the number of energy pulses of the main meter and sub-meters in the meter box, and solve the sub-meter error based on the energy conservation equation; Modeling based on time-period electricity consumption requires first differentiating the cumulative readings of each meter and then aligning them according to time windows. When there are clock discrepancies or inconsistent reading times among the meters, mismatched difference windows directly introduce errors, and the difference operation amplifies reading noise. Errors accumulate and propagate to the regression results window by window. Existing solutions using ridge regression and other regularization methods often employ manually set fixed values ​​for regularization parameters. Excessively large parameters lead to coefficient estimation bias and a systematic underestimation of the inaccuracy ratio; excessively small parameters fail to suppress multicollinearity caused by similar meter consumption behaviors, resulting in unstable coefficient solutions. Furthermore, the optimal parameters vary significantly across different distribution areas and load structures, making manual setting unsuitable. When there are many meters, similar consumption patterns, or some meters have consistently low electricity consumption, the regression... The design matrix is ​​ill-conditioned or even singular. Existing solutions only report an error and interrupt when the solution fails. There is no online detection method for the degree of ill-conditioning, nor is there an automatic recovery strategy, making it unsuitable for unattended edge scenarios. Fixed losses in the transformer area (meter voltage circuit loss, meter box bus loss, etc.) are either substituted as prior constants in existing solutions (requiring pre-experimental measurement and becoming inaccurate with equipment aging) or solved as additional unknowns (increasing the number of unknowns and exacerbating the ill-conditioning of the equation system). Solutions that rely on remote collection and centralized analysis by the electricity information collection system are constrained by the collection cycle and communication bandwidth (usually daily freeze granularity). Uploading all original detailed data is a huge burden and cannot achieve near real-time detection at the meter box. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing a method and system for detecting meter inaccuracies based on the cumulative readings of electricity consumption information.

[0004] The first aspect of this invention provides a method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information, the method comprising the following steps: S1. The measuring switch reads the positive active total cumulative value and status of each sub-meter in the meter box at a high frequency granularity of minute or quarter-hour via RS-485 bus. At the same time, the measuring unit of the measuring switch synchronously accumulates the total power consumption value. The reading record is marked with a timestamp using the real-time clock of the measuring switch as a unified time standard, forming a cumulative value sequence. S2. The inaccuracy calculation unit performs data validity verification on the cumulative indication sequence, and removes or marks invalid samples. S3. Using the cumulative values ​​at the same time as samples, construct a linear regression model of the value range with intercept by grouping by phase: Where y(t) is the total electrical charge reading of the measuring switch, x i (t) represents the cumulative reading of the i-th submeter, β0 is the intercept term, and β i Let ε(t) be the metering transfer coefficient for the i-th sub-meter, and ε(t) be the residual term. S4. Solve the regression model using ridge regression; S5. Construct the coefficient matrix of the centered normal equation system, perform Cholesky decomposition on the coefficient matrix, monitor the diagonal element sequence of the decomposition process to determine the health of the design matrix, and trigger a graded self-healing strategy based on the health until the metering transfer coefficient of each sub-meter is successfully solved or an ill alarm is output. S6. Calculate the inaccuracy ratio based on the metering transfer coefficient of each sub-meter, compare the inaccuracy ratio with the preset classification threshold, output the sub-meter inaccuracy judgment result and report it.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the data validity verification in step S2 includes state quantity verification, continuity verification, indication value rationality verification, and low-activity table pre-screening, wherein, The state quantity verification is used to mark the records of times when the voltage exceeds the normal power supply range, or when there is a phase loss or current loss as invalid. The continuity check is used to discard samples that cross discontinuity intervals when the interval between adjacent valid copying times does not match the preset granularity. The indication rationality check is used to remove records that show a backward cumulative indication, exceed the range, or contradict the magnification. The low-activity meter pre-screening is used to temporarily suspend the inclusion of sub-meters whose cumulative reading changes within the detection cycle are lower than a preset threshold in the current regression, record them separately, and automatically include them after they become active.

[0006] Optionally, in a second implementation of the first aspect of the present invention, a voltage-dependent phase reorientation step is further included between step S2 and step S3: Extract the three-phase voltage sequence U of the measurement switch from the valid dataset. A U B U C The voltage sequence U of the single-phase sub-meter to be identified m ; Calculate U respectively m with U A U B U C The Pearson correlation coefficient; The phase corresponding to the one with the largest correlation coefficient is selected as the assigned phase of the sub-meter, and the sub-meters are grouped according to their assigned phases.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the intercept term β0 in step S3 is used to absorb the sum of the initial reading offsets of each sub-meter and the combined amount of the fixed loss component of the meter box.

[0008] Optionally, in the fourth implementation of the first aspect of the present invention, in step S4, ridge regression is used to solve for the multicollinearity of the design matrix caused by similar electricity consumption behavior of sub-meters: Where α is a regularization parameter, a candidate set with a pre-defined logarithmically uniform distribution is used, K-fold cross-validation is performed based on valid historical samples, and the α value with the smallest validation error is selected as the solution parameter for this round.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, the process of solving the centered normal equation system in step S5 includes: X is obtained by column centering the design matrix X and the observation vector y. c and y c ; Construct a system of normal equations: ; For the coefficient matrix Perform Cholesky decomposition ; After successful decomposition, β is obtained through previous and back substitution, and then... Back substitution to solve for the intercept term; Among them, X c Let y represent the matrix after column-centered X. cLet X represent the vector centered with respect to y, β represent the eigenvalue vector to be determined, α represent the regularization parameter, I represent the identity matrix, and X represent the vector of eigenvalues. c T X c Let A denote the covariance matrix, and L denote the lower triangular matrix obtained by Cholesky decomposition of A. This represents the mean of the original target y. This represents a vector composed of the mean values ​​of each column of the original design matrix X.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the hierarchical self-healing strategy in step S5 includes first-level self-healing, second-level self-healing, third-level self-healing, and a fallback strategy, wherein, The first-level self-healing mechanism is used to rebuild the coefficient matrix and re-solve the problem when the Cholesky decomposition is successful but the decay gradient of the diagonal element sequence exceeds a preset threshold. This is done by increasing the coefficient matrix by a preset boosting factor based on the current regularization parameters, while simultaneously marking the coefficient confidence level as downgraded. The secondary self-healing mechanism is used to automatically extend the sample accumulation window to increase the sample size and then reconstruct the equation for resolution when the Cholesky decomposition fails. The three-level self-healing is used to identify and remove several low-activity sub-meters that are most correlated with other sub-meters when the decomposition still fails after extending the window. After dimensionality reduction, the sub-meters are re-decomposed, and the removed sub-meters are marked as unidentifiable in this round. The fallback strategy is used to abandon the current solution and output an abnormal alarm for the station data if the three-level self-healing fails.

[0011] Optionally, in the seventh implementation of the first aspect of the present invention, the inaccuracy ratio is calculated in step S6 based on the metering transfer coefficient of each sub-meter: Where, δ i ≈0 indicates normal measurement, δ i >0 indicates undercounting, δ i <0 indicates multiple counts; δ i Compared with the grading judgment threshold, the output shows three levels of status: normal / suspected / inaccurate, and suggests a rotation priority order.

[0012] Optionally, in the eighth implementation of the first aspect of the present invention, the reporting information of the sub-meter inaccuracy judgment result in step S6 includes: the inaccuracy ratio of each sub-meter, the three-level status and the suggested rotation priority ranking, wherein the final alarm is upgraded and output only when the inaccuracy direction and amplitude of the same sub-meter are consistent in multiple consecutive detection cycles, and the phase return result, the list of low-activity meters that have been removed and the solvable closed-loop action record are also reported.

[0013] A second aspect of the present invention provides a meter misalignment detection system based on the cumulative reading of electricity consumption information, employing the above-described method, comprising: The measurement switch installed at the inlet of the meter box has a built-in metering unit, a main control unit, an RS-485 communication interface, a high-speed power line carrier remote communication unit, and a real-time clock unit. Several sub-meters are installed at the outlet of the meter box, and the sub-meters are connected to the measuring switch via an RS-485 bus. The main control unit of the measuring switch reads the cumulative readings and status values ​​of each sub-meter via RS-485 bus at high frequency, and sends the total power reading of the local metering unit along with the read data to the inaccuracy calculation unit embedded in the measuring switch. The inaccuracy calculation unit is used to perform data validity verification, adaptive regular regression solution of cumulative indication, and inaccuracy determination, and submits the determination result back to the main control unit, which then reports it to the area concentrator and the main station system via high-speed power line carrier.

[0014] Compared with the prior art, the present invention can achieve at least the following technical effects: 1. Modeling is performed directly in the cumulative indication domain. The intercept absorbs the initial offset and fixed loss of the meter. No differential power or window alignment is performed. The alignment error propagation and differential noise amplification path are cut off at the data organization level. 2. Minute-level high-frequency reading enables thousands of samples per day and night, and the equation system is fully overdetermined; combined with adaptive regularization, it can still stably solve the coefficients of each meter in multicollinearity scenarios where the electricity consumption behavior of each meter is highly similar. 3. Solvability closed loop transforms the ill-conditioning of the matrix into observable and hierarchical events, enabling automatic recovery or accurate alarms even when the edge device is running unattended, thus avoiding "silent failure"; 4. Regularization parameters are automatically determined by cross-validation of distribution area and cycle, eliminating the need for manual parameter adjustment and adapting to distribution areas of different sizes and load structures; 5. Low memory and computational requirements, zero external dependencies; it can run as software / firmware embedded in the measurement switch without requiring independent hardware; it does not upload original details, only the inaccuracy diagnosis results, resulting in low bandwidth pressure and low data security risks. 6. Using the measuring switch as the host, the RS-485 reads and reads existing meter data without modification, suitable for both new and modified meter boxes; inaccurate results are imported into the existing power consumption information acquisition channel (area concentrator → main station) via HPLC, seamlessly connecting with the existing acquisition architecture, flexible deployment, and compatible with the power consumption information acquisition system. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 A flowchart of a meter inaccuracy detection method based on cumulative readings of electricity consumption information provided in an embodiment of the present invention; Figure 2 This is a solvable closed-loop logic diagram provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an adaptive regular regression meter misalignment detection device based on the cumulative reading of electricity consumption information, provided in an embodiment of the present invention. Figure 4 The diagram shows the structure and data flow of the misalignment calculation unit provided in this embodiment of the invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The flowchart of the meter inaccuracy detection method based on cumulative readings from electricity consumption information collection provided in this invention embodiment includes five stages: high-frequency reading acquisition, data validity verification, voltage correlation phase correction (dependent, optional), adaptive regular regression solution of cumulative readings (including CV selection α, Cholesky solvability closed-loop branching and hierarchical self-healing), inaccuracy determination and result reporting, which are executed cyclically. S1, High-frequency indication acquisition (executed by the main control of the measurement switch) A high-frequency electricity consumption information acquisition unit, independent of the main station's remote data collection channel, is built locally in the meter box by the main control unit of the measurement switch. The acquisition objects are the cumulative reading of the electricity meter and the status quantity. The acquisition process is as follows: The main control unit of the measurement switch establishes communication with each sub-meter in the meter box via RS-485 bus to read the meter address and phase type; Using a preset high-frequency granularity period, such as minute-level or quarter-hour-level, the system reads the total cumulative positive active power reading of each sub-meter, as well as status quantities such as voltage, current, and power; the metering unit of the measurement switch synchronously accumulates the total energy reading. Each reading record is in the form of a meter identifier, including the reading time, cumulative reading, and status quantity group. The total cumulative positive active power reading represents the continuously increasing energy reading, not the energy consumption during a specific time period. Using the RTC (Record-Time Controller) of the measurement switch as a unified time standard, timestamps are marked on the reading records. Individual late / missed reading records are merged into the nearest reference time slice according to the reading time. No power differential or window power alignment calculation is performed in this stage, and the cumulative reading value is entered into the subsequent stage as is. The main control unit delivers the aforementioned meter box data to the embedded misalignment calculation unit. The meter box data includes total power consumption, cumulative values ​​of each sub-meter, status parameters, and timestamps.

[0019] S2, Data Validation (executed by the Inaccuracy Calculation Unit) The misalignment calculation unit performs multidimensional verification on the acquired indication sequence, eliminating or marking invalid samples: (1) Status quantity verification: Records of abnormal states such as voltage exceeding the normal power supply range, phase loss, and current loss are marked as invalid; (2) Continuity verification: When the interval between adjacent valid reading times does not match the preset granularity, i.e. when it exceeds the tolerance, the sample that crosses the interval is discarded; (3) Verification of the reasonableness of the indicated value: records with cumulative indicated value regression, over-range jump, or obvious contradiction with the multiplier are removed; (4) Pre-screening of low-activity tables: Tables whose values ​​show almost no change during the detection period have an approximately constant value column, which will be highly collinear with the intercept column, worsening the ill-conditioned nature of the regression. They are temporarily suspended from being included in this round of regression and recorded separately. They will be automatically included after they become active. S3, Voltage-related phase relocation (subordinate / optional) For scenarios where phase information is missing or may be distorted in single-phase meter ledgers, phase reorientation preprocessing is performed: Extract the three-phase voltage sequence U of the measurement switch from the valid dataset. A U B U C The voltage sequence U of the single-phase sub-meter to be identified m ; Calculate U respectively m with U A U B U C Pearson correlation coefficient: Here, r(X,Y) calculates the degree of linear correlation between two variable sequences X and Y, where X and Y represent two sets of voltage data sequences being compared, k represents the sampling point index, and x...k and y k These represent the specific voltage values ​​of sequences X and Y at the k-th time, respectively; The phase with the largest correlation coefficient among the three candidate phases is selected as the assigned phase of the sub-table. The sub-tables are grouped according to the assigned phase, and the subsequent regression is constructed according to the phase group to build their respective indication regression models. The total electricity indication component of the same phase corresponds to the sub-table value of the same phase. The phase identification results are also used to correct the ledger.

[0020] S4. Solving the cumulative indication adaptive regular regression. S4.1, Intercept-based Indication Model For a certain phase group or all sub-meters, let the total electrical quantity reading of the measuring switch be y, and the readings of the P sub-meters in the same group be x1,…,xP. Establish a linear model with intercept at sampling time t: Where: β i (i≥1) is the measurement transfer coefficient of the i-th sub-table, where β is the measurement transfer coefficient when the sub-table is accurate. i ≈1; β i >1 indicates that the table undercounts (runs slowly), β i <1 indicates over-counting (running faster); β0 is the intercept term, which physically represents the sum of the initial readings and meter offsets of each meter, combined with the fixed loss component of the distribution area. Since the initial readings of each meter are constant, their contribution on the time axis can be completely absorbed by the intercept, thus eliminating the need to know the initial readings of each meter, pre-determine fixed losses, or perform any differential calculations; ε(t) is the residual term, containing fluctuation line loss and noise; since no differential calculations are performed, each synchronization moment of high-frequency reading is a set of samples, and N moments constitute an N×P design matrix X. With minute-level granularity, thousands of samples can be obtained in a 24-hour period, naturally forming a sufficiently overdetermined set of equations. Compared with incremental difference modeling, this model eliminates window alignment error and difference noise amplification problems at the data organization level.

[0021] S4.2 Cross-validation adaptively determines the regularization parameter a on the logarithmic candidate set. To address the multicollinearity of the design matrix caused by similar electricity consumption behavior across different meters, ridge regression is used for solution: The regularization parameter α is not a fixed value, but is automatically determined by the miscalculation unit according to the transformer area and the round: on the preset log-uniform candidate set, historical samples are used for cross-validation, and the α with the best validation error is selected as the solution parameter for this round. The range and density of the candidate set are predetermined based on the sample size of the transformer area and the granularity of the copying.

[0022] S4.3, Cholesky decomposition solution + solvability loop closure The solution employs a centered + Cholesky decomposition path: (1) The design matrix X and the observation vector y are column-centered to obtain Xc and yc; (2) Construct a system of normal equations: ; (3) For the coefficient matrix Perform Cholesky decomposition A=LL T In theory, A is always symmetric positive definite. However, when the subdivisions are highly collinear and α is small, the minimum eigenvalue of A approaches zero. The decomposition process may fail due to non-positive diagonal elements, or the diagonal element sequence may be severely decayed, indicating an ill-conditioned state. (4) After successful decomposition, β is obtained through previous and back substitution, and then... Back substitution to solve for the intercept term; Among them, X c Let y represent the matrix after column-centered X. c Let X represent the vector centered with respect to y, β represent the eigenvalue vector to be determined, α represent the regularization parameter, I represent the identity matrix, and X represent the vector of eigenvalues. c T X c Let A denote the covariance matrix, and L denote the lower triangular matrix obtained by Cholesky decomposition of A. This represents the mean of the original target y. This represents a vector composed of the mean values ​​of each column of the original design matrix X.

[0023] Please see Figure 2 The solvability closed-loop logic diagram provided in this embodiment of the invention constructs a coefficient matrix and performs Cholesky decomposition, using the diagonal element sequence as the health measure; if the decomposition is healthy, the solution is solved normally and output; otherwise, a self-healing strategy is triggered in stages, including "increasing α to resolve / extending the sample window to reconstruct the equation / removing highly correlated and low-activity sub-tables to reduce dimensionality and resolve / outputting pathological alarms" and looping back to resolve, with each level of failure being handled in a progressively escalating manner. Solvability loop closure: The diagonal element sequence of Cholesky decomposition is used as an online observation of the health of the design matrix, and a hierarchical self-healing strategy is triggered based on it. Level 1 self-healing: When Cholesky decomposition is successful but the decay gradient of the diagonal element sequence exceeds the preset threshold, the coefficient matrix is ​​reconstructed and re-solved based on the current regularization parameter by increasing the preset boosting coefficient by a value, while the coefficient confidence is downgraded. Secondary self-healing: When Cholesky decomposition fails, the sample accumulation window is automatically extended to increase the sample size before reconstructing the equation and solving it again; Level 3 self-healing: If decomposition still fails after extending the window, identify and remove several low-activity sub-meters that are most correlated with other sub-meters, reduce the dimensionality and re-decompose, and mark the removed sub-meters as unidentifiable in this round. fallback strategy: If the solution fails after three levels of self-healing, abandon the current round of solving and output an abnormal alarm for the station data.

[0024] This closed loop uses the "solver" as a "sensor": the numerical ill-conditioning of the matrix is ​​no longer a fault, but a valid information reflecting the convergence of power consumption patterns in different areas, low meter activity, or data quality degradation. Based on this, the detection process can be automatically restored and alarms can be diverted. The thresholds and boosting coefficients at each level are pre-calibrated according to the scale of the area.

[0025] S5. Accuracy Judgment and Result Reporting S5.1, Judgment of Table Inaccuracy Calculate the inaccuracy ratio based on the measurement transfer coefficient of each sub-meter: Where, δ i ≈0 indicates normal measurement, δ i >0 indicates undercounting, δ i <0 indicates multiple counts; δ i Compared with the tiered judgment threshold, the threshold is pre-calibrated according to the metrological management requirements, with a typical level of percentage. It outputs three levels of status: normal / suspected / inaccurate, and suggests a rotation priority order. Only when the inaccurate mark has the same direction and amplitude for several consecutive cycles will it be upgraded to the final alarm, thus suppressing occasional fluctuations in a single cycle. S5.2 Result Reporting The misalignment calculation unit returns the judgment result to the measurement switch main control unit, stores it locally, and reports it to the station area concentrator via HPLC power line carrier. The concentrator then aggregates and forwards the result to the main station system / maintenance terminal, simultaneously illuminating the grading indicator light. The phase return result, the list of low-activity tables that were removed, and the record of resolvable closed-loop actions are also attached for easy maintenance review.

[0026] S6. Local storage of results + HPLC reporting to the main station of the concentrator in the station area (with phase return results / list of low activity tables / closed-loop action records).

[0027] Please see Figure 3 The present invention provides a schematic diagram of a meter misalignment detection system based on cumulative readings of electricity consumption information. The measuring switch (at the inlet, also serving as a measurement assessment unit, with an embedded misalignment calculation unit) inside the meter box reads the cumulative readings and status values ​​of each sub-meter via a high-frequency RS-485 bus. The measuring switch reports the misalignment calculation results to the distribution area concentrator via HPLC power line carrier, and then forwards them to the main station system / maintenance terminal. This system includes: The measurement switch is a metering device installed at the inlet of the meter box. It has its own metering unit, built-in main control unit, RS-485 communication interface, HPLC remote communication unit and power management. Its host form is not limited to the measurement switch, but can also be the assessment master meter or other meter box level metering devices. The metering unit is used to assess the total power consumption, that is, the source of the total power observation value of the regression model. The misalignment calculation unit, which is embedded in the measurement switch's calculation unit (software / firmware form), receives the meter box data delivered by the main device, independently completes the misalignment detection algorithm calculation, and submits the result back to the main device for reporting. It does not communicate directly with the outside world. The distribution meter is installed at the outlet of the meter box to measure the electricity consumption of each user. There are P units (P≥2) and it provides the total positive active power reading and status quantities such as voltage. The RS-485 bus connects the measurement switch and each sub-meter to form a local high-frequency electricity consumption information acquisition channel at the meter box level. It reads data at the minute or quarter-hour level, which is significantly higher than the conventional centralized reading / daily freezing granularity of the electricity consumption information acquisition system. The measurement switches transmit the misalignment calculation results to the distribution area concentrator via HPLC power line carrier, and the concentrator then aggregates and forwards them to the main station system / maintenance terminal.

[0028] In this embodiment, the metering unit is used to assess the total power consumption of the meter box and provide a total power consumption observation sequence for the regression model; the main control unit is used to read the cumulative readings and status values ​​of each sub-meter at a preset high-frequency granularity via RS-485, stamp them with a unified timestamp, and send the read data along with the local metering data to the misalignment calculation unit; it receives the misalignment calculation results and reports them via HPLC; the RS-485 communication interface is used to communicate with each sub-meter in the meter box, supporting DL / T698.45 or DL / T645 protocols; the HPLC remote communication unit is used to report the misalignment detection results to the area concentrator and forward them to the main station; the power supply and RTC unit is used to draw power from the meter box and stabilize the voltage, providing a unified time reference; please refer to Figure 4 The misalignment calculation unit structure and data flow diagram provided in this embodiment of the invention are as follows: The misalignment calculation unit is embedded in the meter box data (total power reading, cumulative reading of each sub-meter, status quantity) delivered by the main device from the measurement switch input. Internally, it sequentially performs data validity verification, (optional) phase return, adaptive regular regression solution of cumulative reading (CV selects α+Cholesky solvability closed loop), misalignment determination, and outputs a misalignment list to be submitted back to the main device for reporting.

[0029] 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 preferred examples and are not intended to limit 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 the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information, characterized in that, The method includes the following steps: S1. The measuring switch reads the positive active total cumulative value and status of each sub-meter in the meter box at a high frequency granularity of minute or quarter-hour via RS-485 bus. At the same time, the measuring unit of the measuring switch synchronously accumulates the total power consumption value. The reading record is marked with a timestamp using the real-time clock of the measuring switch as a unified time standard, forming a cumulative value sequence. S2. The inaccuracy calculation unit performs data validity verification on the cumulative indication sequence, and removes or marks invalid samples. S3. Using the cumulative values ​​at the same time as samples, construct a linear regression model of the value range with intercept by grouping by phase: Where y(t) is the total electrical charge reading of the measuring switch, x i (t) represents the cumulative reading of the i-th submeter, β0 is the intercept term, and β i Let ε(t) be the metering transfer coefficient for the i-th sub-meter, and ε(t) be the residual term. S4. Solve the regression model using ridge regression; S5. Construct the coefficient matrix of the centered normal equation system, perform Cholesky decomposition on the coefficient matrix, monitor the diagonal element sequence of the decomposition process to determine the health of the design matrix, and trigger a graded self-healing strategy based on the health until the metering transfer coefficient of each sub-meter is successfully solved or an ill alarm is output. S6. Calculate the inaccuracy ratio based on the metering transfer coefficient of each sub-meter, compare the inaccuracy ratio with the preset classification threshold, output the sub-meter inaccuracy judgment result and report it.

2. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, The data validity verification in step S2 includes state quantity verification, continuity verification, value reasonableness verification, and low-activity table pre-screening. The state quantity verification is used to mark the records of times when the voltage exceeds the normal power supply range, or when there is a phase loss or current loss as invalid. The continuity check is used to discard samples that cross discontinuity intervals when the interval between adjacent valid copying times does not match the preset granularity. The indication rationality check is used to remove records that show a backward cumulative indication, exceed the range, or contradict the magnification. The low-activity meter pre-screening is used to temporarily suspend the inclusion of sub-meters whose cumulative reading changes within the detection cycle are lower than a preset threshold in the current regression, record them separately, and automatically include them after they become active.

3. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, Between step S2 and step S3, there is also a voltage-dependent phase reorientation step: Extract the three-phase voltage sequence U of the measurement switch from the valid dataset. A U B U C The voltage sequence U of the single-phase sub-meter to be identified m ; Calculate U respectively m with U A U B U C The Pearson correlation coefficient; The phase corresponding to the one with the largest correlation coefficient is selected as the assigned phase of the sub-meter, and the sub-meters are grouped according to their assigned phases.

4. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, The intercept term β0 mentioned in step S3 is used to absorb the sum of the initial reading offsets of each sub-meter and the combined amount of the fixed loss component of the meter box.

5. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, In step S4, ridge regression is used to solve for the multicollinearity of the design matrix caused by similar electricity consumption behavior in sub-meters: Where α is a regularization parameter, a candidate set with a pre-defined logarithmically uniform distribution is used, K-fold cross-validation is performed based on valid historical samples, and the α value with the smallest validation error is selected as the solution parameter for this round.

6. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, The process of solving the centered normal equation system described in step S5 includes: X is obtained by column centering the design matrix X and the observation vector y. c and y c ; Construct a system of normal equations: ; For the coefficient matrix Perform Cholesky decomposition ; After successful decomposition, β is obtained through previous and back substitution, and then... Back substitution to solve for the intercept term; Among them, X c Let y represent the matrix after column-centered X. c Let represent the vector centered with respect to y, β represent the vector of eigenvalues ​​to be determined, α represent the regularization parameter, and I represent the identity matrix. Let A denote the covariance matrix, and L denote the lower triangular matrix obtained by Cholesky decomposition of A. This represents the mean of the original target y. This represents a vector composed of the mean values ​​of each column of the original design matrix X.

7. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, The graded self-healing strategy described in step S5 includes first-level self-healing, second-level self-healing, third-level self-healing, and a fallback strategy. The first-level self-healing mechanism is used to rebuild the coefficient matrix and re-solve the problem when the Cholesky decomposition is successful but the decay gradient of the diagonal element sequence exceeds a preset threshold. This is done by increasing the coefficient matrix by a preset boosting factor based on the current regularization parameters, while simultaneously marking the coefficient confidence level as downgraded. The secondary self-healing mechanism is used to automatically extend the sample accumulation window to increase the sample size and then reconstruct the equation for resolution when the Cholesky decomposition fails. The three-level self-healing is used to identify and remove several low-activity sub-meters that are most correlated with other sub-meters when the decomposition still fails after extending the window. After dimensionality reduction, the sub-meters are re-decomposed, and the removed sub-meters are marked as unidentifiable in this round. The fallback strategy is used to abandon the current solution and output an abnormal alarm for the station data if the three-level self-healing fails.

8. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 1, characterized in that, In step S6, the misalignment ratio is calculated based on the metering transfer coefficient of each sub-meter: Where, δ i ≈0 indicates normal measurement, δ i >0 indicates undercounting, δ i <0 indicates multiple counts; δ i Compared with the grading judgment threshold, the output shows three levels of status: normal / suspected / inaccurate, and suggests a rotation priority order.

9. The method for detecting meter inaccuracy based on the cumulative reading of electricity consumption information as described in claim 8, characterized in that, The information reported in step S6 regarding the sub-meter inaccuracy determination results includes: the inaccuracy ratio of each sub-meter, the three-level status, and the suggested rotation priority ranking. The final alarm is only upgraded when the inaccuracy direction and amplitude of the same sub-meter are consistent within multiple consecutive detection cycles. At the same time, the phase return result, the list of low-activity meters that have been removed, and the record of solvable closed-loop action are also reported.

10. A meter misalignment detection system based on the cumulative reading of electricity consumption information collection, employing the method described in claim 1, characterized in that, include: The measurement switch installed at the inlet of the meter box has a built-in metering unit, a main control unit, an RS-485 communication interface, a high-speed power line carrier remote communication unit, and a real-time clock unit. Several sub-meters are installed at the outlet of the meter box, and the sub-meters are connected to the measuring switch via an RS-485 bus. The main control unit of the measuring switch reads the cumulative readings and status values ​​of each sub-meter via RS-485 bus at high frequency, and sends the total power reading of the local metering unit along with the read data to the inaccuracy calculation unit embedded in the measuring switch. The inaccuracy calculation unit is used to perform data validity verification, adaptive regular regression solution of cumulative indication, and inaccuracy determination, and submits the determination result back to the main control unit, which then reports it to the area concentrator and the main station system via high-speed power line carrier.

Images