Power distribution cabinet line intelligent detection method and system

Abstract

The invention relates to the technical field of power distribution network monitoring, in particular to a power distribution cabinet line intelligent detection method and system, and the method comprises the steps: collecting node voltage, phasor current and active / reactive power of each power distribution cabinet, attaching a sampling period and a quality identifier, and carrying out the arrangement according to a timestamp; carrying out edge calculation on harmonic waves, flicker, unbalance, missing measurement compensation and initial estimation, and generating a straight-through UDP stream and a storage stream; a real-time interface is aligned and fused with a topology version according to a unified time reference, state estimation is carried out, and a node initial value set is output; s4, according to the initial value and topological modeling node digital twinning (index recovery load), obtaining online parameters through four-stage sub-step discrete + UKF joint estimation; accessing a real-time simulation platform through HiL, and recharging in the same period; a variable damping impedance stable interface is adopted, and when consistency and a threshold value are met, a digital snapshot containing an online parameter, a steady-state quantity and a transient abstract is generated. According to the method, low time delay of measurement alignment, credible state estimation, digital twinborn self-adaption, hardware-in-loop stability and verifiable result release can be realized, so that the fault detection, positioning and evaluation capabilities of the power distribution network are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution network monitoring, and particularly relates to a power distribution cabinet line intelligent detection method and system. BACKGROUND

[0002] Current power distribution cabinet line state perception mainly relies on periodic measurements sent by distributed power meters / protective devices / gateways. The collection devices are heterogeneous in origin, the time references are not unified, and the data quality identifiers are missing or inconsistent, which makes it difficult to align the data across devices and feeder lines in terms of event arrival time. Abnormal signals such as voltage sag, short-time flicker, and three-phase imbalance mutation are easily masked or misdetected. The edge side usually lacks the ability to extract power quality indicators in real time and to compensate for missing measurements, resulting in an inefficient process of relying on polling for alarms and manual positioning. The state estimation on the center side is mainly based on static operating conditions and lagging topology information, which makes it difficult to reflect the adjustment of connectivity caused by changes in circuit breaker / isolating switch states in a timely manner, leading to a deviation between estimation and field. On the other hand, traditional offline simulation is difficult to close-loop with field measurements, and when coupled with hardware-in-the-loop, it is prone to interface energy non-conservation and numerical oscillation problems, making it difficult for simulation results to serve as real-time basis for operation and dispatch. SUMMARY

[0003] In view of the above technical problems, the present application provides a power distribution cabinet line intelligent detection method and system to provide a power distribution cabinet line intelligent detection method, realize measurement alignment with low latency, reliable state estimation, adaptive digital twin, stable hardware-in-the-loop, and verifiable result publishing, thereby improving the fault detection, positioning and evaluation capabilities of the power distribution network.

[0004] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the present application.

[0005] According to an aspect of the present application, a power distribution cabinet line intelligent detection method is provided, which comprises:

[0006] Collecting power grid measurement values at each line node of each power distribution cabinet power grid, recording the sampling period and data quality identifier, wherein the power grid measurement values at least include voltage amplitude, phasor current, active power and reactive power, and arranging the power grid measurements into a node measurement sequence according to the time stamp;

[0007] Based on the node measurement sequence, an electric energy quality index including harmonics, flicker and three-phase imbalance is calculated on the edge side of the power distribution cabinet, local missing measurement compensation and primary state estimation are performed, a straight-through bypass channel data stream for low-latency event transmission and a regular warehouse channel data stream for batch processing are generated; the straight-through bypass channel data stream is sent to the real-time interface system in a connectionless message, and the message is packet aggregated and compressed to converge into a fusion candidate set; the regular warehouse channel data stream is written into a central database;

[0008] Based on the real-time interface system, after aligning and checking the fusion candidate set according to a unified time reference and a current topology version number, network-wide state estimation is performed and an initial value set of each node is output, the initial value set at least contains reference voltage, current, power, device type and operating interval, voltage-dependent initial parameter, topology state number and time consistency label, and the power grid simulation model loaded in the real-time simulation platform and the topology version number are synchronously updated according to the initial value set;

[0009] Based on the initial value set and topology information, a node digital twin model is established for each node, the node digital twin model uses an exponential recovery load model to represent the steady-state and transient behavior of the active power and the reactive power relative to voltage changes, the node digital twin model updates active recovery state quantities and reactive recovery state quantities according to a four-level sub-step weighted discretization model every sampling period, and a scaled unscented Kalman filter is used for joint estimation of state and parameters to obtain an online parameter set, so as to correct the steady-state index, transient index and recovery time constant, so that the output of the node digital twin model is consistent with the power grid measurement value, and the unscented Kalman filter takes the initial value set as a priori and parameter constraint boundary;

[0010] The node digital twin model is connected to the real-time simulation platform through a hardware-in-the-loop interface, the real-time simulation platform generates node voltages according to the power grid simulation model and inputs them into the node digital twin model, the node digital twin model outputs equivalent currents and power backfilling to the real-time simulation platform to complete same-cycle solving, and the simulation quantities obtained by solving are returned to the real-time interface system through a monitoring channel and are associated with the online parameter set;

[0011] Based on the variable damping impedance algorithm, the instantaneous equivalent impedance is calculated and damping components are superimposed according to the simulation quantities fed back and associated by the monitoring channel in the previous cycle. When phase oscillation or alternating power direction exceeding the threshold is detected, the damping is increased, and after stabilization, the damping is withdrawn at a set slope to reduce numerical dissipation. The updated instantaneous equivalent impedance is applied to the coupling interface between the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantities. When the simulation quantities and the corresponding power grid measurement values ​​meet the time consistency label and the deviation index is within the preset threshold, the real-time interface system generates a digital snapshot and publishes it to the outside world through the standard protocol. The digital snapshot includes the latest online parameter set, steady-state quantities, and the most recent transient trajectory summary.

[0012] Furthermore, when arranging the node measurement sequence, the local clocks of different distribution cabinets are calibrated and drift corrected according to a unified time reference, the maximum drift and jitter are limited, and the power grid measurement values ​​are rearranged in order of arrival time to obtain the node measurement sequence; the data quality identifier includes at least the types of missing measurement, out-of-bounds, saturation, phase anomaly and mutual inductance coupling suspicion.

[0013] Furthermore, the real-time interface system includes the following functional modules:

[0014] The data acquisition and synchronization module is used to receive and correct voltage, current, active power, and reactive power measurement data uploaded by edge nodes;

[0015] The state estimation and power quality analysis module is used to perform network-wide state estimation based on the synchronized node measurement sequence and extract the power quality index of each node.

[0016] The dynamic model generation module is used to generate the node digital twin model based on the state estimation results;

[0017] The data output module is used to output the parameters of the node digital twin model to the real-time simulation system for simulation.

[0018] Furthermore, the construction of the node digital twin model specifically includes: determining the structure of the exponential recovery load model based on the equipment type and the operating range; and superimposing the power quality index onto the equivalent current in a narrowband injection manner, so that the power grid simulation model replicates the local power quality characteristics in the real-time simulation platform.

[0019] Further, the discretization model of the fourth-order sub-step weighting comprises: sequentially calculating four sub-increments in each sampling period, and summing the four sub-increments with weights of "1, 2, 2, 1" to update the active recovery state quantity and the reactive recovery state quantity, so as to convert continuous dynamics into a discrete process which can be iterated in real time.

[0020] Further, the scaled unscented Kalman filter performs rolling in each sampling period when performing joint estimation of states and parameters, and specifically comprises:

[0021] generating a set of symmetric sampling points according to a current state covariance;

[0022] propagating each symmetric sampling point in the set of symmetric sampling points through the discretization model, and collecting to obtain a prior state mean and a prior state covariance;

[0023] mapping observations of voltage and power to the set of symmetric sampling points to obtain observation mean, observation covariance and state-observation cross-covariance;

[0024] calculating observation residuals and obtaining gains according to the observation residuals, writing back a state vector and its covariance, and outputting the online parameter set of the current period.

[0025] Further, a covariance self-adaptive adjustment strategy is adopted in the joint estimation process:

[0026] statistically calculating variance and related diagnostic quantities of the observation residuals in a sliding window;

[0027] when the observation residuals exceed a threshold, increasing process noise covariance to enhance tracking ability;

[0028] when the observation residuals continuously fall below the threshold in multiple windows, shrinking measurement noise covariance to improve steady-state accuracy;

[0029] assigning a robust weight to a window marked as abnormal to suppress its adverse effect on statistical quantities;

[0030] the adjustment result is used for symmetric sampling point generation and gain calculation in the next sampling period, forming a closed-loop optimization.

[0031] Further, the topology information comprises:

[0032] a set of connectivity data of the power grid at the current time, at least including a connection pair of nodes and branches, open / close states of circuit breakers / switches / sectionalizers and tie devices, phase connection relationships, and feeder affiliation and island division derived therefrom;

[0033] the topology information is used to generate the topology state number and associate the topology version number.

[0034] Further, the steady-state quantities include: steady-state electrical variables aggregated by nodes and branches, including node voltage effective value and phase angle, active power, and reactive power;

[0035] The last transient trajectory summary includes:

[0036] Record the key trajectory from the start of the last disturbance to the recovery of the steady state, including the event time window, extreme value and deviation, rise / fall time, setting time, steady-state error, and parameter change record and quality information identified by the trajectory and written back to the online parameter set.

[0037] According to a second aspect of the present disclosure, a power distribution cabinet line intelligent detection system is provided, the system comprising:

[0038] A data acquisition and timestamp unification module is configured to acquire power grid measurement values at each line node of each power distribution cabinet power grid, and record the sampling period and data quality identifier, wherein the power grid measurement values at least include voltage amplitude, phasor current, active power and reactive power, and the power grid measurement values are arranged into a node measurement sequence according to the time stamp;

[0039] An electric energy analysis and shunt module is configured to calculate electric energy quality indicators based on the node measurement sequence, on the edge side of the power distribution cabinet, wherein the electric energy quality indicators include harmonics, flicker and three-phase imbalance, perform local missing measurement compensation and primary state estimation, generate a straight-through bypass channel data stream for low-latency event transmission and a regular warehouse channel data stream for batch processing; send the straight-through bypass channel data stream to a real-time interface system in the form of a connectionless message, and perform small packet aggregation and compression on the message to converge into a fusion candidate set; write the regular warehouse channel data stream into a central database;

[0040] A communication module is configured to, based on the real-time interface system, after aligning and checking the fusion candidate set according to a unified time reference and a current topology version number, perform network-wide state estimation and output an initial value set of each node, wherein the initial value set at least contains reference voltage, current, power, device type and operating interval, voltage-dependent initial parameters, topology state number and time consistency label, and update the power grid simulation model loaded in the real-time simulation platform and the topology version number according to the initial value set in synchronization;

[0041] A model construction and identification module is configured to establish a node digital twin model for each node based on the initial value set and topology information, wherein the node digital twin model represents the steady-state and transient behavior of the active power, the reactive power and the relative voltage change by using an exponential recovery load model, the node digital twin model updates the active recovery state quantity and the reactive recovery state quantity according to a four-level sub-step weighted discretization model at each sampling period, and a scaled unscented Kalman filter is used to perform joint estimation of the state and the parameter to obtain an online parameter set, so as to correct the steady-state index, the transient index and the recovery time constant, so that the output of the node digital twin model is consistent with the grid measurement value, and the unscented Kalman filter takes the initial value set as the prior and the parameter constraint boundary.

[0042] An interface module is configured to connect the node digital twin model to the real-time simulation platform through a hardware-in-the-loop interface, the real-time simulation platform generates node voltages according to a grid simulation model and inputs the node voltages to the node digital twin model, the node digital twin model outputs equivalent currents and power backflow, and the real-time simulation platform completes the same-period solving, the simulation quantity obtained by solving is returned to the real-time interface system through a monitoring channel and is associated with the online parameter set.

[0043] An interface stabilization module is configured to calculate an instantaneous equivalent impedance by adding a damping component based on a variable-damping impedance algorithm according to the simulation quantity returned by the monitoring channel and associated in the last period, to increase the damping when phase swing or power direction alternation exceeds a threshold, and to reduce the damping according to a set slope after stabilization to reduce numerical dissipation, and to apply the updated instantaneous equivalent impedance to the coupling interface of the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantity.

[0044] A publishing module is configured to generate a digital snapshot and publish it externally through a standard protocol when the simulation quantity and the corresponding grid measurement value meet the time consistency label and the deviation index is within a preset threshold, wherein the digital snapshot includes the latest online parameter set, steady-state quantity and the latest transient trajectory abstract.

[0045] The technical solution of the present application has the following beneficial effects:

[0046] Compared with the prior art, the application significantly reduces the end-to-end delay of event type information through the communication division of the straight-through bypass channel + the conventional warehouse channel, while ensuring the continuous warehousing of batch high-fidelity data; the unified time reference + topology version number / topology state number enables the consistency check of the timing and connectivity of the fusion candidate set before it enters the whole network state estimation, thereby improving the reliability of the initial value set; on this basis, the online identification closed loop composed of the node digital twin model + the exponential recovery load model + the scaled unscented Kalman filter continuously calibrates the steady-state and transient-state coupling relationship of "active / reactive - voltage" in a parameterized manner, generates an online parameter set and writes back constraints, and takes into account the robustness to missing data / noise and the rapid tracking of sudden changes in working conditions; through the hardware-in-the-loop interface + variable damping impedance algorithm, the node digital twin model is stably coupled with the simulation platform, interface oscillation and numerical dissipation are suppressed, and the simulation quantity can be used as a real-time reference for decision detection. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A flowchart of a power distribution cabinet line intelligent detection method in an embodiment of the present specification;

[0048] Figure 2 A structural block diagram of a power distribution cabinet line intelligent detection system in an embodiment of the present specification. DETAILED DESCRIPTION

[0049] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations can be implemented in any number of manners, and are not limited to the examples set forth herein; instead, these implementations are provided so that this disclosure will be more thorough and complete, and will fully convey the concept of example implementations to those skilled in the art. The described features, structures, or characteristics can be combined in one or more implementations in any suitable manner. In the following description, numerous specific details are provided to give a thorough understanding of implementations of the present disclosure. One skilled in the relevant art will recognize, however, that the implementations of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, etc. In other instances, well-known structures have not been described in detail so as not to obscure aspects of the present disclosure.

[0050] Furthermore, the accompanying drawings merely show possible implementations of the present disclosure. The same reference numbers in different drawings identify the same or similar components and therefore repeated description of them will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily have to correspond to physically or logically independent entities. These functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0051] The application provides a product power distribution cabinet line intelligent detection method. Referring to Figure 1 Fig. 1 is a flowchart of a power distribution cabinet line intelligent detection method provided by an embodiment of the application. The method can be applied to electronic devices such as personal computers, servers, controllers, etc. The method can be executed by a device, which can be realized by software and / or hardware. The method can specifically include the following steps S101-S106:

[0052] In step S101, power grid measurement values collected at line nodes of each power distribution cabinet power grid are recorded, and sampling periods and data quality identifiers are recorded. The power grid measurement values at least include voltage amplitude, phasor current, active power and reactive power. The power grid measurement values are arranged into a node measurement sequence according to time stamps.

[0053] In the arrangement of the node measurement sequence, the local clocks of different power distribution cabinets are calibrated and drift-corrected according to a unified time reference, the maximum drift and jitter are limited, the power grid measurement values are reordered according to arrival time sequence, and the node measurement sequence is obtained. The data quality identifier at least includes missing measurement, out-of-bound, saturation, phase anomaly and mutual inductance coupling suspicious types.

[0054] In the arrangement of the node measurement sequence, the local clocks of different power distribution cabinets are calibrated and drift-corrected according to a unified time reference, the maximum drift and jitter are limited, the power grid measurement values are reordered according to arrival time sequence, and the node measurement sequence is obtained. The data quality identifier at least includes missing measurement, out-of-bound, saturation, phase anomaly and mutual inductance coupling suspicious types.

[0055] In the step of arranging the original measurement by timestamp into a node measurement sequence, this step limits the calibration based on a unified time reference and the drift / dither correction of the local clock of different power distribution cabinets, and the measurement frame arriving is reordered by arrival time sequence and then written into the sequence; wherein the data quality identifier is used to mark the missing, out of range, saturation, phase anomaly, mutual inductance coupling suspicious quality events, so as to be excluded or weighted in subsequent estimation and alignment. In order to support this time-ordered and quality-traceable organization, a time-critical straight-through path + conventional warehouse path can be adopted, that is, the straight-through path is carried by UDP, small packet aggregation and compression can be carried out, so as to reduce the end-to-end delay and preferentially transmit time-sensitive event frames; the conventional warehouse path enters the high-speed write / interior read mode of the central database, so as to guarantee the real-time warehousing of field data and the fast read-write of subsequent processing. It should be pointed out that the meaning of unified time reference is to align all node measurements, quality identifiers and subsequent features to the same time axis, so as to facilitate cross-node comparison and synchronous checking of simulation / estimation process; the database level can adopt double mode (two sets of schema), the first set is responsible for high-speed writing of field measurement and fast reading for RTI internal processing, and the second set is responsible for providing processing results to real-time simulation, which guarantees the capture, management and calling of timestamp-ordered data from the system architecture.

[0056] In step S102, based on the node measurement sequence, the power quality indicators including harmonics, flicker and three-phase imbalance are calculated on the edge side of the power distribution cabinet, local missing compensation and preliminary state estimation are performed, and a straight-through bypass channel data stream for low-latency event transmission and a conventional warehouse channel data stream for batch processing are generated; the straight-through bypass channel data stream is sent to the real-time interface system in the form of connectionless message, and small packet aggregation and compression are performed on the message to converge into a fusion candidate set; the conventional warehouse channel data stream is written into the central database.

[0057] Wherein, based on the node measurement sequence obtained in step S101, the power quality phenomena are calculated and condensed on the edge side of the power distribution cabinet, and a node feature set which can be used for subsequent comparison and modeling is formed. The power quality indicators at least cover harmonics and flicker, and can be extended to three-phase imbalance according to engineering needs; wherein, as typical power quality phenomena, harmonics and flicker are extracted as node features, which are used for node-by-node comparison and verification with real-time simulation output, and are used to inject real performance (such as actually measured harmonic components) which has not been included in the dynamic model into subsequent simulation and calibration process, so as to improve the fidelity and traceability of digital twin. In order to ensure the effectiveness of such features in the subsequent process, the edge side organizes and outputs the features with a unified time reference, so as to ensure that the features from different sources have consistent time reference before being transmitted into the real-time simulation platform.

[0058] To improve the continuity and robustness of features, the edge side performs local missing compensation and preliminary state estimation in the same step to make up for the node measurement void caused by missing measurement points or communication frame loss, and to identify and exclude suspicious bad data and model library errors, thereby improving the data reliability entering the subsequent stage. The preliminary state estimation here focuses on the fast completion and health screening of the edge side, producing basic consistency data for subsequent fusion and simulation. After this processing, each node has not only the original voltage / current / power time series, but also the power quality features and the estimated and completed basic electrical parameters aligned with them, providing input for the fusion and verification of the subsequent real-time interface system.

[0059] After the above calculations are completed on the edge side, two data channels with different purposes are generated synchronously: one is a straight-through bypass channel data stream for low-latency event transmission, and the other is a regular storage channel data stream for batch processing. The straight-through bypass channel sends connectionless messages to the real-time interface system, where the connectionless messages correspond to messages carried by the User Datagram Protocol (UDP); this channel is dedicated to time-sensitive information such as voltage fast events and faults to minimize end-to-end delay. On the real-time interface system side, the arriving UDP data are aggregated and compressed on demand to further reduce latency and improve transmission efficiency; at the same time, the regular storage channel writes complete measurements and processed data to the central database. The communication of the real-time interface system is divided into a fast straight-through path and a regular database path, i.e., the former directly processes incoming UDP streams and serves node calculations and grid state monitoring, and the latter relies on the central database to complete data storage and management. To support high-concurrency writing and simulation reading, the database adopts a dual-mode structure: the first set of modes is used to write field data at high speed and read out quickly for the internal processing components of the real-time interface, and the second set of modes is used to write back the processing results and provide them for the real-time simulation platform. The real-time interface system aligns and checks the straight-through bypass data and the database readouts on a unified time reference to form a fusion candidate set required for subsequent processing.

[0060] In step S103, based on the real-time interface system, after aligning and checking the fusion candidate set on a unified time reference and the current topology version number, performing full-network state estimation and outputting the initial value set of each node, the initial value set at least contains the reference voltage, current, power, device type and operating interval, voltage-dependent initial parameters, topology state number and time consistency label, and synchronously updating the grid simulation model loaded in the real-time simulation platform and the topology version number according to the initial value set.

[0061] The real-time interface system comprises the following functional modules: a data acquisition and synchronization module, configured to receive and correct voltage, current, active power, and reactive power measurement data uploaded by an edge node; a state estimation and power quality analysis module, configured to perform network-wide state estimation based on the synchronized node measurement sequence, and extract the power quality indicators of each node; a dynamic model generation module, configured to generate a node digital twin model according to the state estimation result; and a data output module, configured to output the parameters of the node digital twin model to the real-time simulation system for simulation.

[0062] The real-time interface system first receives a fusion candidate set collected by two paths: one is fast pass-through data in a connectionless message bearer, used for node calculation and power grid state monitoring, and the other is a data path relying on central database management. To ensure the effectiveness of subsequent calculations, the real-time interface system aligns and checks the fusion candidate set according to a unified time reference and a current topology version number, that is, different node measurements and known connectivity at the same time are processed consistently. The unified time reference is used to ensure that parameter sets of different sources have the same time reference before being transmitted to the real-time simulation, and the topology version number is used to identify the network structure description currently used, to ensure that subsequent calculations and simulations are based on the same network architecture. The above-mentioned division of labor between the pass-through path and the database path, and the requirement of effectiveness guarantee and alignment based on the unified time reference; the database side adopts a clear division of labor structure to support fast writing and simulation reading, so that the real-time interface system can align the two types of data under the same time reference and enter subsequent processing.

[0063] After completing the consistency processing of time and topology, the real-time interface system performs a network-wide state estimation function, calculates state variables according to measurement quantities and topology information, and thus obtains an initial value set of each node. State estimation is considered a key function in power grid monitoring, which aims to restore the system state using measurement and topology information. In the hierarchical node digital twin system, this function, power quality monitoring, and node modeling based on an observer together constitute a core processing chain, which outputs a parameter-adjustable representation for real-time simulation. The initial value set is oriented to the node level, and at least includes reference voltage, reference current, and reference power, which are used to represent the current steady-state working condition; includes device type and operating range, which are used to limit the physical and operating boundaries of subsequent parameter identification; includes voltage-dependent initial parameters, which are used to give an initial description of active / reactive power variation with voltage in the node digital twin model; and is accompanied by a topology state number and a time consistency label, to explicitly bind the initial value set to a specific connectivity and time window. Through such an organization form, the initial value set can serve as priori for subsequent node modeling, and also as a working condition anchor when loaded into the real-time simulation platform.

[0064] Based on the above initial value set, the real-time interface system synchronously updates the power grid simulation model and the topology version in the real-time simulation platform, so that the grid configuration on the simulation side is consistent with the physical side. When a topology event (such as a switch state change) is detected, the connectivity relationship is continuously updated to ensure that the simulation always reflects the real power grid; at the same time, the central database undertakes the responsibility of collecting and forwarding information to the hardware-in-the-loop interface, so that the node characteristics and state estimation results are stably supplied to the simulation environment. Under this process, the real-time simulation platform is based on the grid consistent with the database, receives the processed node and power grid information for solving, and the subsequent hardware-in-the-loop and filter link further closes the loop for checking.

[0065] Around the functional decomposition of the real-time interface system, the data acquisition and synchronization module is responsible for receiving and correcting the voltage, current, active power and reactive power measurements sent by the edge node, ensuring that the data before entering the fusion candidate set has a unified time reference and necessary quality control; the state estimation and power quality analysis module performs network-wide state estimation under the premise of time and topology consistency, and outputs the node-level power quality characteristics for subsequent comparison and simulation injection; the dynamic model generation module generates the initial parameters and structure description required for the node digital twin model according to the state estimation results; the data output module is responsible for delivering the node digital twin model related parameters and initial value set to the real-time simulation system.

[0066] In step S104, a node digital twin model is established for each node based on the initial value set and topology information, the node digital twin model uses an exponential recovery load model to represent the steady-state and transient behavior of the active power, the reactive power relative to the voltage change, the node digital twin model updates the active recovery state quantity and the reactive recovery state quantity every sampling period according to a four-level sub-step weighted discrete model, and a scaled unscented Kalman filter is used for joint estimation of state and parameters to obtain an online parameter set, so as to correct the steady-state index, the transient index and the recovery time constant, so that the output of the node digital twin model is consistent with the power grid measurement value, and the unscented Kalman filter takes the initial value set as the prior and parameter constraint boundary.

[0067] In addition, the scaled unscented Kalman filter performs rolling every sampling period when performing joint estimation of state and parameters, specifically including: generating a set of symmetric sampling points according to the current state covariance; propagating each symmetric sampling point in the set of symmetric sampling points through the discrete model to obtain the prior state mean and the prior state covariance; mapping the voltage and power to the set of symmetric sampling points as observations to obtain the observation mean, the observation covariance and the state-observation cross-covariance; calculating the observation residual and obtaining the gain accordingly, writing back the state vector and its covariance, and outputting the online parameter set of the current period.

[0068] And the topology information includes: a connected relationship data set of the power grid at the current time, at least including a connection pair of a node and a branch, an opening and closing state of a circuit breaker / disconnector / segment and a tie-in device, a phase connection relationship, and a feeder affiliation and island division derived therefrom; the topology information is used to generate the topology state number and is associated with the topology version number.

[0069] The discretization model of the fourth-order sub-step weighting includes: sequentially calculating four sub-increments in each sampling period, and summing the four sub-increments with weights of “1, 2, 2, 1” to update the active power recovery state quantity and the reactive power recovery state quantity, so as to convert continuous dynamics into a discrete process that can be iterated in real time.

[0070] In the joint estimation process, a covariance self-adaptive adjustment strategy is adopted: the variance of the observation residual and related diagnostic quantities are statistically calculated in a sliding window; when the observation residual exceeds a threshold, the process noise covariance is increased to enhance the tracking ability; when the observation residual is continuously lower than the threshold in multiple windows, the measurement noise covariance is shrunk to improve the steady-state accuracy; a robust weight is given to the window marked as abnormal to suppress its adverse effect on the statistical quantity; the adjustment result is used for generating and gain calculation at the symmetric sampling point in the next sampling period, forming a closed-loop optimization.

[0071] Wherein, based on the initial value set obtained in step S103 and the topology information, a node digital twin model is established for each node. The topology information is used to determine the power source / load attribution of the node under the current connected relationship and the coupling with adjacent branches, so that the model can reflect the dependence of node voltage-power on the state of the grid; the node digital twin model selects an exponential recovery load model (ERL) to represent the steady-state and transient behavior of active power and reactive power with respect to voltage changes, which has a moderate parameter scale, is conducive to online estimation, and meets the calculation constraints of real-time implementation. In continuous time, the active / reactive channels of ERL can be described as follows:

[0072]

[0073] Wherein, U0, P0, Q0 are the voltage and active / reactive power before voltage disturbance, x p , x q are active / reactive recovery state quantities, T p , T q are recovery time constants, and α s , β s correspond to transient exponents. The selection aims to reproduce the dynamic characteristics of the node load online with fewer parameters, so as to more accurately fit the response of the physical power grid in real-time simulation.

[0074] To meet the requirements of real-time discrete solution and rolling updates per sampling period, the node digital twin model adopts a four-level sub-step weighted discretization model to update the recovered state variables, corresponding to weights of "1, 2, 2, 1". Taking the active channel as an example, let the sampling step size be Δt, then:

[0075]

[0076] in:

[0077]

[0078] And provide the corresponding output while maintaining the parameters:

[0079] α s (k+1)=α s (k), α t (k+1)=α t (k), T p (k+1)=T p (k);

[0080] The reactive power channel is discretized using the same approach and updated in conjunction with the active power channel to ensure consistency between the steady-state and transient states of the two channels. The aforementioned "1, 2, 2, 1" weighting transforms continuous dynamics into a discrete process that can be iterated in real time.

[0081] To ensure that the model output and nodal measurements remain consistent under noise and nonlinear conditions, a scaled unscented Kalman filter is used to jointly estimate the state and parameters, generating an online "online parameter set" for the current period. This is represented as a discrete nonlinear state space: x k+1 =f(x) k ,u k )+v k y k =h(x k )+n k ,in,

[0082] Under the assumption of finite voltage disturbance amplitude (e.g., single-node voltage drop not exceeding approximately 10%), measurement noise can be considered additive. The time update and measurement update of the scaled unscented Kalman filter include: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] k-1 Generate symmetrical sampling points Propagation to the priori and The sampling points are obtained through observation mapping. P yy With cross covariance P xy ;calculate And update In implementation, αs ,α t ,β s ,β t ,T p ,T q Unknown parameters such as,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,are incorporated into the extended state vector and covariance, and the filter is modified at each sampling period according to the above discrete equations and measurements, thereby gradually correcting the steady-state index, transient index, and recovery time constant. The initial value set is input as the prior mean and parameter constraint boundary, ensuring that the online identification is physically feasible and limited by prior information such as device type and operating interval.

[0083] In step S105, the node digital twin model is connected to the real-time simulation platform through the hardware-in-the-loop interface. The real-time simulation platform generates node voltages according to the power grid simulation model and inputs them into the node digital twin model. The node digital twin model outputs equivalent currents and power, which are fed back to the real-time simulation platform for synchronous solving. The simulation results are transmitted back to the real-time interface system through the monitoring channel and associated with the online parameter set.

[0084] Wherein, the node digital twin model is connected to the real-time simulation platform through the hardware-in-the-loop interface. At this time, the real-time simulation platform generates voltages for each node according to the power grid simulation model, and inputs these voltage values into the corresponding node digital twin model. The node digital twin model uses the digital twin model constructed in step S104 to dynamically update its output of equivalent current and power at each sampling period according to voltage changes. This process is achieved by simulating the dynamic behavior and electrical characteristics of the node, ensuring that the output of the simulation model can reflect the state of each node in the physical power grid in real time. The output equivalent current and power are fed back to the real-time simulation platform, which further completes the synchronous solving of the power grid according to these data and feeds back the calculation results in real time. This process is based on the coupling between the real-time simulation platform and the node digital twin model, ensuring the synchronization and interaction between the power grid model and the digital twin model, thereby achieving effective simulation of the power grid operating state.

[0085] The solved simulation quantities are transmitted back to the real-time interface system through the monitoring channel and associated with the online parameter set. In this process, the monitoring channel is responsible for transmitting the simulation results to the real-time interface system, ensuring that the simulation quantities can be matched and corrected with the online parameter set generated in the previous steps. This data association link provides key data support for subsequent state estimation and simulation model updating, ensuring that each data point in the simulation process is consistent with the actual power grid measurement, thereby improving the accuracy and reliability of the simulation results. Throughout the process, the stable coupling of the hardware-in-the-loop interface ensures that the closed-loop calculation between the digital twin model and the simulation platform can be executed in real time, greatly improving the response speed and simulation accuracy of dynamic changes in the power grid.

[0086] In step S106, based on the variable damping impedance algorithm, the instantaneous equivalent impedance is calculated according to the simulation quantities returned and associated by the monitoring channel in the last cycle, and a damping component is superimposed; when phase swing or power direction alternation exceeds the threshold, the damping is increased, and after stabilization, the damping is withdrawn according to the set slope to reduce numerical dissipation; the updated instantaneous equivalent impedance acts on the coupling interface of the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantities, when the simulation quantities and the corresponding grid measured values meet the time consistency label, and the deviation index is within the preset threshold, the real-time interface system generates a digital snapshot and publishes it externally through a standard protocol, the digital snapshot includes the latest online parameter set, steady-state quantities, and a recent transient trajectory summary.

[0087] The steady-state quantities include steady-state electrical variables summarized by nodes and branches, including node voltage effective value and phase angle, active power, and reactive power.

[0088] The recent transient trajectory summary includes recording the key trajectory from the start to the recovery of the last disturbance, including event time window, extreme value and deviation, rise / fall time, setting time, steady-state error, and parameter change record and quality information identified by the trajectory and written back to the online parameter set.

[0089] Wherein, when the variable damping impedance algorithm is used to calculate the instantaneous equivalent impedance of the power grid, the simulation quantities returned and associated by the monitoring channel in the last cycle are used to perform related operations. Specifically, based on the simulation quantities, the instantaneous equivalent impedance is calculated, and a damping component is superimposed to the calculation result. This operation mainly adjusts the impedance characteristics of the power grid model dynamically to ensure the stability of the power grid when it is disturbed. In particular, when the phase swing or power direction alternation exceeds the set threshold, the damping is immediately increased to slow down the oscillation and prevent instability. Once stabilized, the damping is gradually withdrawn according to the set slope, the purpose is to reduce numerical dissipation and restore normal operation of the system. The purpose of this is to ensure that the power grid simulation platform can quickly respond and recover to a stable state under different power grid disturbances, while avoiding unnecessary energy loss and calculation time caused by excessive damping.

[0090] The updated instantaneous equivalent impedance acts on the coupling interface of the real-time simulation platform and the node digital twin model in real time. At this time, the real-time simulation platform continues to output simulation quantities according to the updated equivalent impedance, and compares the calculation results with the actual power grid measurement values. When the simulation quantities and the power grid measurement values meet the time consistency label, and the deviation between the simulation quantities and the measurement values is lower than the preset threshold, the real-time interface system generates a digital snapshot and publishes it externally through a standard protocol. The digital snapshot includes the latest online parameter set, the steady-state quantity and the latest transient trajectory summary. The steady-state quantity includes the electrical variables summarized by nodes and branches, such as node voltage effective value and phase angle, as well as active power and reactive power, which are the key electrical characteristics of the power grid under steady-state conditions. The latest transient trajectory summary records the key changes from the disturbance to the recovery of the steady-state of the power grid, including the event time window, the extreme value, the deviation, the rise / fall time, the setting time, the steady-state error, and the parameter change record and quality information written back to the online parameter set through trajectory identification. Through these information, the digital snapshot provides a comprehensive record of the operation state of the power grid, ensuring that the state of the power grid can be traced and analyzed, and ensuring the consistency between the simulation results and the actual measurement values.

[0091] In an embodiment, the node digital twin model comprises, when constructed: determining the structure of the exponential recovery load model based on the device type and the operation interval; and superimposing the power quality indicators on the equivalent current in a narrow-band injection manner, so that the power grid simulation model reproduces the local power quality characteristics in the real-time simulation platform.

[0092] In order to improve the fitting degree of the model to the non-ideal characteristics on site, the power quality indicators extracted on the edge side are superimposed on the equivalent current in a narrow-band manner without changing the structure of the exponential recovery load model, which is used to reproduce the local characteristics such as harmonics and flicker of the node on the real-time simulation platform. Such node characteristics are not only used for comparison and verification of simulation results, but also used to integrate real components (such as actually measured harmonics) that are not explicitly covered by the dynamic model into the simulation environment. The above injection and verification process is coordinated with the node monitoring-computing link: the monitoring side condenses the power quality phenomenon into characteristics, and the computing side completes the system identification of each node and forms an equivalent power flow continuous data stream for simulation according to the characteristics, and finally provides the hardware-in-the-loop coupling simulation side under the unified time reference by the real-time interface.

[0093] Based on the same idea, as shown in Figure 2 Fig. 1 is a structural block diagram of a power distribution cabinet line intelligent detection system provided by an embodiment of the present application. The system comprises:

[0094] The data acquisition and time stamping module 201 is used to collect power grid measurement values ​​at each line node of the power grid in each distribution cabinet, and record the sampling period and data quality identifier. The power grid measurement values ​​include at least voltage amplitude, phasor current, active power and reactive power. The power grid measurements are arranged into a node measurement sequence according to the timestamp.

[0095] The power analysis and distribution module 202 is used to calculate power quality indicators at the edge of the distribution cabinet based on the node measurement sequence. The power quality indicators include harmonics, flicker, and three-phase imbalance. It performs local missing measurement compensation and primary state estimation, and generates a direct bypass channel data stream for low-latency event transmission and a regular inbound channel data stream for batch processing. The direct bypass channel data stream is sent to the real-time interface system with connectionless messages, and the messages are aggregated and compressed to form a fusion candidate set. The regular inbound channel data stream is written to the central database.

[0096] The communication module 203 is used to perform full network state estimation and output the initial value set of each node based on the real-time interface system after aligning and verifying the fusion candidate set with the current topology version number according to a unified time reference. The initial value set includes at least the reference voltage, current, power, equipment type and operating range, voltage-dependent initial parameters, topology state number and time consistency label. The module also updates the power grid simulation model loaded in the real-time simulation platform and the topology version number synchronously according to the initial value set.

[0097] The model construction and identification module 204 is used to build a node digital twin model for each node based on the initial value set and topology information. The node digital twin model uses an exponential recovery load model to represent the steady-state and transient behavior of the active power and reactive power relative to voltage changes. The node digital twin model updates the active power recovery state quantity and reactive power recovery state quantity according to a four-level sub-step weighted discretization model in each sampling period, and uses a scaled unscented Kalman filter to jointly estimate the state and parameters to obtain an online parameter set, so as to correct the steady-state exponent, transient exponent and recovery time constant, so that the output of the node digital twin model is consistent with the grid measurement value. The unscented Kalman filter uses the initial value set as the prior and parameter constraint boundary.

[0098] Interface module 205 is used to connect the node digital twin model to the real-time simulation platform through a hardware-in-the-loop interface. The real-time simulation platform generates the voltage of each node according to the power grid simulation model and inputs it to the node digital twin model. The node digital twin model outputs equivalent current and power and feeds them back to the real-time simulation platform to complete the same-cycle solution. The solved simulation quantities are transmitted back to the real-time interface system through the monitoring channel and associated with the online parameter set.

[0099] The interface stabilization module 206 is used to calculate the instantaneous equivalent impedance and superimpose the damping component based on the simulation quantity returned and associated by the monitoring channel in the previous cycle, according to the variable damping impedance algorithm; when phase swing or alternating power direction over-threshold is detected, the damping is increased; after stabilization, the damping is withdrawn at a set slope to reduce numerical dissipation; the updated instantaneous equivalent impedance is applied to the coupling interface between the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantity;

[0100] The publishing module 207 is used to generate a digital snapshot and publish it to the outside world through a standard protocol when the simulated quantity and the corresponding power grid measurement value meet the time consistency label and the deviation index is within a preset threshold. The digital snapshot includes the latest online parameter set, steady-state quantity, and the most recent transient trajectory summary.

[0101] The specific details of the above system have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0102] Compared with existing technologies, this system significantly reduces end-to-end latency of event-type information by dividing communication tasks through a direct bypass channel and a regular data entry channel, while ensuring continuous data entry of batches of high-fidelity data. A unified time base and topology version / state number enable consistency verification of timing and connectivity relationships in the fusion candidate set before entering the overall network state estimation, improving the reliability of the initial value set. Based on this, an online identification closed loop consisting of a node digital twin model, an exponentially recovering load model, and a scaled unscented Kalman filter continuously calibrates the steady-state and transient coupling relationship of "active / reactive power-voltage" in a parameterized manner, generating an online parameter set and writing back constraints, balancing robustness to missing measurements / noise with rapid tracking of sudden changes in operating conditions. Through a hardware-in-the-loop interface and a variable damping impedance algorithm, the real-time simulation platform and the node digital twin model are stably coupled, suppressing interface oscillations and numerical dissipation, allowing simulated quantities to serve as a real-time reference for decision-making.

[0103] The accompanying drawings are merely illustrative of the processes included in the methods according to exemplary embodiments of this disclosure and are not intended to be limiting. It is readily understood that the processes shown in the drawings do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0104] It should be noted that although several modules or units of the system have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0105] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0106] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for intelligent detection of circuits in a power distribution cabinet, characterized in that, The method includes: The power grid measurements are collected at each line node of the power grid in each distribution cabinet, and the sampling period and data quality identifier are recorded. The power grid measurements include at least voltage amplitude, phasor current, active power and reactive power. The power grid measurements are arranged into a node measurement sequence according to the timestamp. Based on the node measurement sequence, power quality indicators, including harmonics, flicker, and three-phase imbalance, are calculated at the edge of the distribution cabinet. Local measurement loss compensation and primary state estimation are performed, generating a through-bypass channel data stream for low-latency event transmission and a regular inbound channel data stream for batch processing. The through-bypass channel data stream is sent to the real-time interface system with connectionless messages, and the messages are aggregated and compressed into a fusion candidate set. The regular inbound channel data stream is written to the central database. Based on the real-time interface system, after aligning and verifying the fusion candidate set with the current topology version number according to a unified time reference, the system performs full network state estimation and outputs the initial value set for each node. The initial value set includes at least the reference voltage, current, power, equipment type and operating range, voltage-dependent initial parameters, topology state number and time consistency label. The system also updates the power grid simulation model loaded in the real-time simulation platform and the topology version number synchronously according to the initial value set. Based on the initial value set and topology information, a node digital twin model is established for each node. The node digital twin model uses an exponential recovery load model to represent the steady-state and transient behavior of the active power and reactive power relative to voltage changes. The node digital twin model updates the active power recovery state quantity and reactive power recovery state quantity according to a four-level sub-step weighted discretization model in each sampling period. The online parameter set is obtained by joint estimation of state and parameters using a scaled unscented Kalman filter to correct the steady-state exponent, transient exponent and recovery time constant, so that the output of the node digital twin model is consistent with the grid measurement value. The unscented Kalman filter uses the initial value set as the prior and parameter constraint boundary. The node digital twin model is connected to the real-time simulation platform through a hardware-in-the-loop interface. The real-time simulation platform generates the voltage of each node according to the power grid simulation model and inputs it to the node digital twin model. The node digital twin model outputs equivalent current and power and feeds them back to the real-time simulation platform to complete the same-cycle solution. The solved simulation quantities are transmitted back to the real-time interface system through the monitoring channel and associated with the online parameter set. Based on the variable damping impedance algorithm, the instantaneous equivalent impedance is calculated and damping components are superimposed according to the simulation quantities returned and associated by the monitoring channel in the previous cycle. When phase oscillation or alternating power direction exceeding the threshold is detected, the damping is increased, and after stabilization, the damping is withdrawn at a set slope to reduce numerical dissipation. The updated instantaneous equivalent impedance is applied to the coupling interface between the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantities. When the simulation quantities and the corresponding power grid measurement values ​​meet the time consistency label and the deviation index is within the preset threshold, the real-time interface system generates a digital snapshot and publishes it to the outside world through the standard protocol. The digital snapshot includes the latest online parameter set, steady-state quantities, and the most recent transient trajectory summary.

2. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, When arranging the node measurement sequence, the local clocks of different distribution cabinets are calibrated and drift corrected according to a unified time reference, the maximum drift and jitter are limited, and the power grid measurement values ​​are rearranged in order of arrival time to obtain the node measurement sequence; the data quality identifier includes at least the types of missing measurement, out-of-bounds, saturation, phase anomaly and mutual inductance coupling suspicion.

3. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The real-time interface system includes the following functional modules: The data acquisition and synchronization module is used to receive and correct voltage, current, active power, and reactive power measurement data uploaded by edge nodes; The state estimation and power quality analysis module is used to perform network-wide state estimation based on the synchronized node measurement sequence and extract the power quality index of each node. The dynamic model generation module is used to generate the node digital twin model based on the state estimation results; The data output module is used to output the parameters of the node digital twin model to the real-time simulation system for simulation.

4. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The construction of the node digital twin model specifically includes: determining the structure of the exponential recovery load model based on the equipment type and the operating range; and superimposing the power quality index onto the equivalent current in a narrowband injection manner, so that the power grid simulation model replicates the local power quality characteristics in the real-time simulation platform.

5. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The four-level sub-step weighted discretization model includes: calculating four sub-increments sequentially in each sampling period, and summing the four sub-increments with weights of "1, 2, 2, 1" to update the active power recovery state quantity and the reactive power recovery state quantity, thereby transforming continuous dynamics into a discrete process that can be iterated in real time.

6. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The scaled unscented Kalman filter performs joint estimation of state and parameters in a rolling manner every sampling period, specifically including: Generate a symmetric set of sampling points based on the current state covariance; Each symmetric sampling point in the set of symmetric sampling points is propagated through the discretization model, and the prior state mean and prior state covariance are obtained by summing them up. Voltage and power are used as observations and mapped to the symmetrical sampling point set to perform observation mapping, resulting in the observation mean, observation covariance, and state-observation cross-covariance. Calculate the observation residuals and obtain the gain accordingly, write back the state vector and its covariance, and generate the online parameter set for the current period.

7. The intelligent detection method for distribution cabinet lines according to claim 6, characterized in that, An adaptive covariance adjustment strategy is employed in the joint estimation process: The variance of the observed residuals and related diagnostic parameters are statistically analyzed using a sliding window. When the observed residual exceeds the threshold, the process noise covariance is increased to enhance the tracking capability. When the observation residual remains below the threshold for multiple windows, the measurement noise covariance is reduced to improve steady-state accuracy. assign robust weights to windows marked as anomalous to suppress their adverse effects on statistics; The adjustment results are used in the next sampling period for symmetric sampling point generation and gain calculation, forming a closed-loop optimization.

8. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The topology information includes: The current set of power grid connectivity data includes at least the connection pairs between nodes and branches, the opening and closing status of circuit breakers / disconnecting switches / sections and tie devices, phase connection relationships, and the resulting feeder affiliation and island division; The topology information is used to generate the topology status number and associate it with the topology version number.

9. The intelligent detection method for distribution cabinet lines according to claim 1, characterized in that, The steady-state quantities include: steady-state electrical variables summarized by nodes and branches, including the effective value and phase angle of node voltage, active power, and reactive power; The most recent transient trajectory summary includes: Record the key trajectory from the onset of the most recent disturbance to the recovery to steady state, including the event time window, extreme values ​​and deviations, rise / fall time, settling time, steady-state error, as well as parameter change records and quality information identified by the trajectory and written back to the online parameter set.

10. An intelligent detection system for distribution cabinet lines, the system comprising: The data acquisition and time stamping module is used to collect power grid measurement values ​​at each line node of the power grid in each distribution cabinet, and record the sampling period and data quality identifier. The power grid measurement values ​​include at least voltage amplitude, phasor current, active power and reactive power. The power grid measurements are arranged into a node measurement sequence according to the timestamp. The power analysis and distribution module is used to calculate power quality indicators at the edge of the distribution cabinet based on the node measurement sequence. The power quality indicators include harmonics, flicker, and three-phase imbalance. It performs local measurement loss compensation and primary state estimation, and generates a direct bypass channel data stream for low-latency event transmission and a regular inbound channel data stream for batch processing. The direct bypass channel data stream is sent to the real-time interface system with connectionless messages, and the messages are aggregated and compressed to form a fusion candidate set. The regular inbound channel data stream is written to the central database. The communication module is used to perform full network state estimation and output the initial value set of each node based on the real-time interface system after aligning and verifying the fusion candidate set with the current topology version number according to a unified time reference. The initial value set includes at least the reference voltage, current, power, equipment type and operating range, voltage-dependent initial parameters, topology state number and time consistency label. The module also updates the power grid simulation model loaded in the real-time simulation platform and the topology version number synchronously according to the initial value set. The model construction and identification module is used to build a node digital twin model for each node based on the initial value set and topology information. The node digital twin model uses an exponential recovery load model to represent the steady-state and transient behavior of the active power and reactive power relative to voltage changes. The node digital twin model updates the active power recovery state quantity and reactive power recovery state quantity according to a four-level sub-step weighted discretization model in each sampling period, and uses a scaled unscented Kalman filter to jointly estimate the state and parameters to obtain an online parameter set, so as to correct the steady-state exponent, transient exponent and recovery time constant, so that the output of the node digital twin model is consistent with the grid measurement value. The unscented Kalman filter uses the initial value set as the prior and parameter constraint boundary. The interface module is used to connect the node digital twin model to the real-time simulation platform through a hardware-in-the-loop interface. The real-time simulation platform generates the voltage of each node according to the power grid simulation model and inputs it to the node digital twin model. The node digital twin model outputs equivalent current and power and feeds them back to the real-time simulation platform to complete the same-cycle solution. The solved simulation quantities are transmitted back to the real-time interface system through the monitoring channel and associated with the online parameter set. The interface stabilization module is used to calculate the instantaneous equivalent impedance and superimpose the damping component based on the simulation quantity fed back and associated by the monitoring channel in the previous cycle, according to the variable damping impedance algorithm. When phase swing or alternating power direction over-threshold is detected, the damping is increased, and after stabilization, the damping is withdrawn at a set slope to reduce numerical dissipation. The updated instantaneous equivalent impedance is applied to the coupling interface between the real-time simulation platform and the node digital twin model in real time, so that the real-time simulation platform continuously outputs the simulation quantity. The publishing module is used to generate a digital snapshot and publish it to the outside world through a standard protocol when the simulated quantity and the corresponding power grid measurement value meet the time consistency label and the deviation index is within a preset threshold. The digital snapshot includes the latest online parameter set, steady-state quantity, and the most recent transient trajectory summary.

Images