Intelligent monitoring system and method for intelligent insulation power distribution cabinet

Abstract

The invention relates to the technical field of intelligent power grids, in particular to an intelligent monitoring system and method for an intelligent insulation power distribution cabinet, and the system comprises a local signal obtaining module which is used for obtaining the multi-dimensional state data of a local circuit; the associated signal receiving module is used for synthesizing network side operation reference data representing the macroscopic working condition of the power grid; the remote indication signal generation module is used for performing feedforward decoupling operation on local data so as to peel off self disturbance of a load, performing differential comparison on the local data and a network side operation reference, and generating remote regulation and control indication data capable of representing a real interaction state between a power grid and a local node; and the remote control module is used for generating a remote control instruction according to the remote regulation and control indication data. According to the invention, through carrying out differential analysis on the physical state of the local load and the network operation reference, predictive monitoring and high-reliability active regulation and control on the potential risk of the power distribution cabinet are realized.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, specifically to an intelligent monitoring system and method for an intelligent insulated distribution cabinet. Background Technology

[0002] In power distribution networks, especially in scenarios like data centers where downtime is unacceptable, the operational reliability of distribution cabinets is crucial. Existing circuitry for remote indication and control of power distribution networks allows maintenance personnel to check the on / off status of switches and perform remote control, but its functionality has significant limitations.

[0003] The remote indication function of these devices is limited to reporting an electrical final state that has already occurred. For example, they can only indicate that a switching device has been opened, but cannot indicate any physical processes that may lead to a change in state, and therefore lack predictability. Specifically, the indicating device cannot remotely indicate gradual physical structural changes at busbar connection points due to continuous thermal effects, which are critical precursors to network stability.

[0004] Similarly, this indicator cannot remotely indicate the physical state of the internal mechanical components of a switch that has been closed for a long time. In a data center, the internal mechanism of a switch may change its operating performance due to prolonged inactivity. When a remote control signal needs to be responded to, the deterministic nature of its execution cannot be guaranteed, and traditional remote indicator circuits cannot detect this.

[0005] Therefore, existing remote indicating circuit devices are limited to indicating basic electrical on / off results and lack the ability to remotely indicate the physical status of related components in the power distribution network. Consequently, the associated remote control circuit devices, unable to obtain indications of the equipment's physical status, lack the necessary basis for their control decisions, making it difficult to meet the data center's need for predictive management of network operation status.

[0006] Therefore, this invention proposes an intelligent monitoring system and method for intelligent insulated distribution cabinets. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent monitoring system and method for intelligent insulated distribution cabinets. By performing differential analysis on the physical state of the local load and the network operating benchmark, it can achieve predictive monitoring and highly reliable proactive control of potential risks in the distribution cabinet.

[0008] To achieve the above objectives, the present invention provides the following technical solution: An intelligent monitoring system for an intelligent insulated distribution cabinet includes: The local signal acquisition module obtains reference voltage data from the input terminal of the data center's insulated distribution cabinet, and simultaneously acquires three-phase load current data, load harmonic content data, and ambient temperature data of the local circuit. The associated signal receiving module receives candidate network status data containing timestamp information from the network, and dynamically filters and synthesizes the candidate network status data according to the preset association logic to generate network-side operation benchmark data. The remote indication signal generation module uses reference voltage data as a benchmark to perform synchronous rotating coordinate transformation on the three-phase load current data to obtain the original load coordinate data. Based on the original load coordinate data, load harmonic content data, and ambient temperature data, it performs feedforward decoupling calculation on the original load coordinate data to generate decoupled local load status data. It calculates the state deviation between the local load status data and the network-side operating reference data to generate differential component data. The magnitude and phase angle of the differential component data together constitute the remote control indication data. The remote control module generates remote control commands based on remote control instruction data.

[0009] Preferably, the local signal acquisition module includes: a synchronization reference generation unit, which extracts a synchronization trigger pulse from the reference voltage data; a multi-channel data conversion unit, which uses the synchronization trigger pulse as a conversion start reference and synchronously acquires three-phase load current parameters characterizing the state of the power grid connection point; a spectrum analysis unit, which performs a fast Fourier transform operation on the conversion result of the three-phase load current to generate load harmonic content parameters reflecting the local load characteristics; and an environmental data acquisition unit, which acquires the ambient temperature data and adds a timestamp associated with the synchronization trigger pulse to the ambient temperature data.

[0010] Preferably, the associated signal receiving module includes: a protocol decoding unit, which decodes the candidate network status data and associated timestamps from data packets that follow a precise time protocol format; a data filtering unit, which removes data that does not meet preset timeliness and integrity requirements by judging the time deviation of the timestamps and verifying the integrity of the data packets; and a data synthesis unit, which applies preset weighting coefficients to all the candidate network status data that have passed the filtering and performs a weighted average calculation to generate smoothed network-side operating reference data that characterizes the average operating conditions of the power grid.

[0011] Preferably, the process of generating the decoupled local load state data includes: performing synchronous reference generation processing on the reference voltage data to generate a rotating coordinate system reference with the same frequency and phase as the reference voltage data; using the rotating coordinate system reference, performing synchronous coordinate transformation on the three-phase load current data to obtain original load coordinate data composed of d-axis and q-axis components; using the original load coordinate data, load harmonic content data, and ambient temperature data as multi-dimensional address inputs, generating a disturbance compensation vector through multilinear interpolation in a preset dynamic feedforward decoupling model; and generating the decoupled local load state data by performing vector subtraction operations on the d-axis and q-axis components of the original load coordinate data with the corresponding components of the disturbance compensation vector.

[0012] Preferably, the dynamic feedforward decoupling model is further dynamically corrected through an online adaptive correction process. This correction process includes: defining the decoupled local load state data as the internal state to be estimated; generating expected original load coordinate data based on the currently estimated internal state, and using the dynamic feedforward decoupling model and the measured original load coordinate data, load harmonic content data, and ambient temperature data; comparing the original load coordinate data actually obtained by synchronous coordinate transformation at the same time with the expected original load coordinate data to generate a prediction residual; and recursively correcting the parameters in the dynamic feedforward decoupling model using the prediction residual and according to a preset recursive estimation algorithm, while simultaneously updating the internal state to be estimated and minimizing the prediction residual.

[0013] Preferably, the local load status data and the network-side operating reference data are used to calculate the state deviation and generate differential component data. The magnitude and phase angle of the differential component data together constitute the remote control indication data, including: defining the decoupled local load status data as a local coordinate dataset, which includes a d-axis local component and a q-axis local component; defining the network-side operating reference data as a network coordinate dataset, which includes a d-axis network component and a q-axis network component; generating differential component data composed of d-axis difference components and q-axis difference components by performing vector difference operations on the local coordinate dataset and the network coordinate dataset; obtaining the original magnitude and the original phase angle by performing a Cartesian to polar coordinate transformation on the differential component data; and performing phase compensation operations on the original phase angle according to preset power grid topology transmission delay parameters to obtain the adjusted phase angle. The adjusted phase angle and the original magnitude together constitute the remote control indication data.

[0014] Preferably, the process of the adjusted phase angle and the original modulus value together constituting the remote control indication data further includes: using a preset control strategy partition plane divided by multiple operating area boundaries as a grading criterion, wherein the radial coordinate of the control strategy partition plane corresponds to the original modulus value and the polar coordinate corresponds to the adjusted phase angle; the operating area boundaries divide the control strategy partition plane into multiple control level areas; and by determining the region to which the real-time coordinate point formed by the original modulus value and the adjusted phase angle belongs within the control strategy partition plane, an additional control level identifier corresponding to the region is added to the remote control indication data.

[0015] Preferably, the remote control module includes: a strategy matching unit that matches the control level identifier in the remote control instruction data with a preset dispatch response strategy library; an instruction generation unit that, based on the matching result, indexes the instruction corresponding to the control level identifier from the dispatch response strategy library and outputs it; the dispatch response strategy library pre-stores instruction sets corresponding to different control level identifiers, the instruction sets including protective trip instructions with target switch equipment addresses and disconnection operation codes, topology reconfiguration instructions defining the coordinated operation sequence and target state of multiple switch equipment, and load transfer instructions setting the target tap position of on-load tap-changing transformers; and a communication encapsulation unit that encapsulates the generated specific instructions and target equipment addresses into data packets conforming to the power automation communication protocol.

[0016] A method for intelligent monitoring of an intelligent insulated distribution cabinet, comprising: Reference voltage data is obtained from the input terminal of the data center's insulated distribution cabinet, and three-phase load current data, load harmonic content data, and ambient temperature data of the local circuit are obtained simultaneously. The system receives candidate network state data containing timestamp information from the network and dynamically filters and synthesizes the candidate network state data according to a preset correlation logic to generate network-side operation baseline data. Using reference voltage data as a benchmark, the three-phase load current data is synchronously rotated to obtain the original load coordinate data. Based on the original load coordinate data, load harmonic content data, and ambient temperature data, a feedforward decoupling operation is performed on the original load coordinate data to generate decoupled local load status data. The state deviation between the local load status data and the network-side operating reference data is calculated to generate differential component data. The magnitude and phase angle of the differential component data together constitute the remote control indication data. Based on the remote control instruction data, generate remote control commands.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention synchronously acquires local multidimensional data (three-phase current, harmonics, temperature) and network-side operating benchmarks, and uses a dynamic feedforward decoupling model to extract local load state data reflecting the internal physical condition of the equipment. This data is then differentially compared with the network-side operating benchmark data. This design quantifies gradual physical changes caused by accumulated thermal effects at busbar connection points or performance degradation of the switching mechanism into remotely visible differential component signals. This fundamentally changes the existing technology's passive reporting of the final state of a switch being disconnected, enabling accurate prediction of fault precursors and providing a basis for predictive maintenance decisions.

[0018] 2. The differential component data (magnitude and phase angle) generated by the aforementioned predictive monitoring not only indicates the degree of deviation but also reveals its nature. By mapping this data to the control strategy partition plane, the system can automatically assess the risk level and match corresponding scheduling response strategies (such as protective tripping, topology reconfiguration, etc.). Compared to traditional remote control, which lacks a basis for control due to the inability to know the physical state of the equipment, the control commands of this invention are based on a deep understanding of the equipment's health status. Its decision-making is well-founded, and the determinism and security of its execution are significantly improved in high-reliability scenarios such as data centers.

[0019] 3. The dynamic feedforward decoupling model designed in this invention possesses online adaptive correction capabilities. It continuously compares the model's predicted output with the actual measured values, using the resulting prediction residuals to recursively correct the model parameters. This core correlation design enables the monitoring system to autonomously learn and adapt to slow changes in load characteristics or the operating environment, continuously ensuring the accuracy of the physical state decoupled from complex electrical signals, and guaranteeing the long-term stable and accurate operation of the entire monitoring and control system. Attached Figure Description

[0020] Figure 1 This is a block diagram of an intelligent monitoring system for an intelligent insulated distribution cabinet according to the present invention; Figure 2 This is a flowchart illustrating the control level division based on the control strategy partition plane according to an embodiment of the present invention. Figure 3 This is a flowchart of an intelligent monitoring method for an intelligent insulated distribution cabinet according to the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Other embodiments obtained by those skilled in the art based on the ideas in this specification without creative effort all fall within the protection scope of this invention.

[0022] Reference Figures 1 to 3 The present invention provides an intelligent monitoring system and method for intelligent insulated distribution cabinets, the specific technical solution of which is as follows.

[0023] An intelligent monitoring system for an intelligent insulated distribution cabinet, referring to Figure 1 ,include: The local signal acquisition module obtains reference voltage data from the input terminal of the data center's insulated distribution cabinet, and simultaneously acquires three-phase load current data, load harmonic content data, and ambient temperature data of the local circuit. The associated signal receiving module receives candidate network status data containing timestamp information from the network, and dynamically filters and synthesizes the candidate network status data according to the preset association logic to generate network-side operation benchmark data. The remote indication signal generation module uses reference voltage data as a benchmark to perform synchronous rotating coordinate transformation on the three-phase load current data to obtain the original load coordinate data. Based on the original load coordinate data, load harmonic content data, and ambient temperature data, it performs feedforward decoupling calculation on the original load coordinate data to generate decoupled local load status data. It calculates the state deviation between the local load status data and the network-side operating reference data to generate differential component data. The magnitude and phase angle of the differential component data together constitute the remote control indication data. The remote control module generates remote control commands based on remote control instruction data.

[0024] Example 1: This embodiment provides an intelligent monitoring system for intelligent insulated distribution cabinets applied to intelligent power distribution networks within a Tier 4 data center (IDC). In this data center scenario, the requirements for power supply continuity are extremely stringent; any unforeseen power outage could lead to incalculable losses. The intelligent monitoring system proposed in this invention is deployed in intelligent insulated distribution cabinets at critical power nodes within the data center to achieve in-depth monitoring and predictive management of the cabinet's operational status.

[0025] Furthermore, the local signal acquisition module includes: a synchronization reference generation unit, which extracts a synchronization trigger pulse from the reference voltage data; a multiplex data conversion unit, which uses the synchronization trigger pulse as a conversion start reference and synchronously acquires three-phase load current parameters characterizing the state of the power grid connection point; a spectrum analysis unit, which performs a fast Fourier transform operation on the conversion result of the three-phase load current to generate load harmonic content parameters reflecting the local load characteristics; and an environmental data acquisition unit, which acquires the ambient temperature data and adds a timestamp associated with the synchronization trigger pulse to the ambient temperature data.

[0026] Specifically, the local signal acquisition module is deployed within the intelligent insulated distribution cabinet of the data center, and its workflow is as follows: The synchronization reference generation unit continuously monitors the A-phase reference voltage waveform acquired from the incoming side of the distribution cabinet through a high-impedance input terminal. This unit integrates a hardware zero-crossing detection circuit based on a high-speed comparator. When this circuit detects the precise moment when the instantaneous value of the A-phase reference voltage crosses zero potential from negative to positive, it outputs a synchronization trigger pulse with a pulse width of 100 nanoseconds.

[0027] The multiplexer unit receives this synchronous trigger pulse and uses it as the reference for parallel conversion initiation. This unit contains three independent 24-bit high-precision analog-to-digital converters (ADCs), connected to the secondary sides of the current transformers in phases A, B, and C, respectively. The synchronous trigger pulse is distributed to the conversion initiation pin of each ADC, ensuring that the sampling and quantization of the three-phase load currents occur physically simultaneously. This design eliminates measurement errors introduced by timing differences between channels.

[0028] The spectrum analysis unit incorporates a digital signal processor (DSP) to process data from the multiplexing unit. Within one power frequency cycle (20 milliseconds for a 50Hz grid), the DSP acquires 256 synchronous sampling points of the three-phase load current. After one cycle, the DSP applies a Hanning window function to the data sequence containing the 256 sampling points to suppress spectral leakage, and then performs a Fast Fourier Transform (FFT) operation. The result of this operation is the amplitude and phase of the DC component and each harmonic component. The unit extracts the amplitude data of the 3rd, 5th, 7th, 9th, 11th, and 13th harmonics and packages them into a load harmonic content parameter.

[0029] The environmental data acquisition unit obtains ambient temperature data through a PT100 platinum resistance temperature sensor. This sensor is fixed to the insulation surface of the copper busbar of phase A in the distribution cabinet, 10 cm away from the circuit breaker terminal, to monitor the temperature at a critical location reflecting conductor resistivity and equipment performance. The system associates a high-precision system clock count value as a timestamp with each generated synchronization trigger pulse. This timestamp is linked to the ambient temperature data acquired within the same power frequency cycle, forming ambient temperature data with a precise time reference.

[0030] By strictly locking all data acquisition events to a unified reference voltage zero-crossing pulse, this module obtains a highly synchronized, multi-dimensional state snapshot containing both electrical and non-electrical quantities. This snapshot provides a high-quality, time-delay-free data foundation for subsequent accurate coordinate transformations and state analysis.

[0031] In the aforementioned correlation signal receiving module, its core function is to dynamically filter and synthesize candidate network state data based on preset correlation logic. Here, correlation logic refers to the rules used to determine how to filter data and how to allocate data synthesis weights (i.e., weighting coefficients). This logic can have different technical foundations: for example, it can be static logic based on physical topology, such as allocating weights based on the electrical distance between nodes in the power grid topology; or it can be dynamic logic based on upper-level operational objectives, such as dynamically selecting different synthesis strategies based on the current operating mode of the data center (e.g., "energy-saving mode" or "high-reliability mode").

[0032] Furthermore, the associated signal receiving module includes: a protocol decoding unit, which decodes the candidate network status data and associated timestamps from data packets that follow a precise time protocol format; a data filtering unit, which removes data that does not meet preset timeliness and integrity requirements by judging the time deviation of the timestamps and verifying the integrity of the data packets; and a data synthesis unit, which applies preset weighting coefficients to all the candidate network status data that have passed the filtering and performs a weighted average calculation to generate smoothed network-side operating reference data that characterizes the average operating conditions of the power grid.

[0033] Specifically, the associated signal receiving module is used to collect and process operational data from other nodes in the data center power distribution network. The protocol decoding unit receives data packets via an Ethernet interface. These data packets are encapsulated according to the IEEE 1588 (Precise Time Protocol) standard, ensuring that the time synchronization accuracy between data sources reaches the sub-microsecond level. The protocol decoding unit parses a data structure from the payload of each data packet, which is the candidate network state data, containing the grid frequency, effective voltage value, and voltage phase angle information of the source node; at the same time, the unit extracts a high-precision associated timestamp from the header of the data packet.

[0034] The data filtering unit verifies the validity of the decoded data. First, the unit compares the associated timestamp in the data packet with the local system's synchronization clock to calculate the time deviation. A preset timeliness requirement is set at 500 microseconds. If the time deviation exceeds this threshold, it indicates that the data has lost its real-time nature due to network congestion or other reasons, and the data filtering unit discards it. Subsequently, the unit performs integrity verification on the data packets that pass the timeliness check. By calculating the cyclic redundancy check (CRC) code of the data packet content and comparing it with the check field at the end of the packet, any mismatched data packet is considered corrupted and discarded.

[0035] The data synthesis unit processes all filtered candidate network state data. Based on a pre-defined configuration table, this unit assigns different weighting coefficients to candidate network state data from different source nodes. These pre-defined weighting coefficients are configured according to the electrical distance of each candidate network state data source within the data center's power grid topology. For example, the weighting coefficient is set to 0.4 for monitoring point data deployed within the same ring network unit; 0.3 for data from the upstream substation; and 0.15 for data from two different adjacent ring networks. The data synthesis unit multiplies each component of each candidate network state data (frequency, voltage, phase angle) by its corresponding weighting coefficient, then sums all the weighted components to generate a single, smoothed network-side operational baseline data. To achieve the above configuration based on electrical distance, the configuration process is as follows: First, the system establishes the node admittance matrix of the data center power grid offline; second, it calculates the potential distribution coefficient from other candidate data source nodes in the network to this distribution cabinet node; finally, the preset weighting coefficient is configured to be proportional to the magnitude of the potential distribution coefficient and is normalized to ensure that nodes with closer electrical connections receive higher weights.

[0036] In a preferred embodiment, the correlation logic is dynamic, and the correlation signal receiving module is further configured to: receive a mode identifier representing the current operating mode of the data center from an external management system; and, based on the mode identifier, select the correlation logic corresponding to the current operating mode and the target value of the network-side operating benchmark data from a preset benchmark strategy library to generate network-side operating benchmark data adapted to the current operating mode.

[0037] Specifically, the associated signal receiving module also subscribes to operating mode broadcast messages from the data center infrastructure management system. The baseline strategy library pre-stores different strategies for two operating modes: "High Reliability Mode" and "Energy Saving Mode." In "High Reliability Mode," the correlation logic defined in the baseline strategy library is as follows: candidate network status data from the power grid inlet and UPS are assigned a weighting coefficient of 0.7, and the target voltage value for the network-side operating baseline data is set to the rated voltage of 220V. In "Energy Saving Mode," the correlation logic defined in the baseline strategy library becomes: candidate network status data from distribution nodes associated with large chiller units are assigned a weighting coefficient of 0.8, and the target voltage value for the network-side operating baseline data is set to 225V. When the associated signal receiving module receives a mode identifier from the DCIM indicating a switch from "High Reliability" to "Energy Saving," it automatically loads the latter strategy from the baseline strategy library and uses it to synthesize the network-side operating baseline data.

[0038] This solution deeply couples the underlying electrical monitoring with the upper-level IT operation strategy, making the network-side benchmark no longer a fixed electrical target, but a dynamic and intelligent reference system that can adapt to the overall operational intent of the data center.

[0039] This module transforms discrete network information from multiple sources, which contains time delays and noise, into a single, stable, and accurate reference signal that reflects the macroscopic operating conditions of the entire data center power grid, providing a reliable reference system for subsequent high-sensitivity state deviation calculations.

[0040] Further, the process of generating the decoupled local load state data includes: performing synchronous reference generation processing on the reference voltage data to generate a rotating coordinate system reference with the same frequency and phase as the reference voltage data; using the rotating coordinate system reference, performing synchronous coordinate transformation on the three-phase load current data to obtain original load coordinate data composed of d-axis and q-axis components; using the original load coordinate data, load harmonic content data, and ambient temperature data as multi-dimensional address inputs, generating a disturbance compensation vector through multilinear interpolation in a preset dynamic feedforward decoupling model; and generating the decoupled local load state data by performing vector subtraction operations between the d-axis and q-axis components of the original load coordinate data and the corresponding components of the disturbance compensation vector.

[0041] Specifically, this process is executed within the remote indication signal generation module. First, the module performs synchronous reference generation processing on the A-phase reference voltage data provided by the local signal acquisition module. This processing is implemented by a software phase-locked loop (SPLL), whose output consists of two orthogonal sine and cosine digital signals. The frequencies and phases of these two signals are precisely locked to the fundamental component of the A-phase reference voltage. These two orthogonal signals together constitute a coordinate system reference that rotates synchronously with the grid voltage (a rotation matrix from the α-β coordinate system to the dq coordinate system).

[0042] Subsequently, using the generated rotating coordinate system reference, the module performs a synchronous coordinate transformation, or Park transformation, on the three-phase load current data (ia, ib, ic) synchronously acquired by the local signal acquisition module. This transformation projects the instantaneous current values ​​in the three-phase stationary coordinate system onto the synchronously rotating dq coordinate system. The calculation process is as follows: through a series of algebraic operations, the three-phase current data is converted into original load coordinate data composed of d-axis and q-axis components. The d-axis component represents the projection of the current in the flux linkage direction, and the q-axis component represents the projection of the current in the torque direction. Ideally, for a purely resistive load, the q-axis component should be zero.

[0043] Next, the module invokes a preset dynamic feedforward decoupling model, which is physically a high-dimensional lookup table stored in non-volatile memory. The multiple input address dimensions of this lookup table correspond to the key state parameters of the local load: the first set of dimensions is the original load coordinate data obtained from the synchronous coordinate transformation, consisting of d-axis and q-axis components; the second set of dimensions is the load harmonic content data, specifically a multi-dimensional vector composed of the amplitudes of the 3rd, 5th, 7th, and 11th harmonics generated by the spectrum analysis unit; and the third dimension is the ambient temperature data.

[0044] At each grid point of this high-dimensional lookup table, a pre-calculated disturbance compensation vector is stored. This vector contains a d-axis compensation component (d_comp) and a q-axis compensation component (q_comp). These compensation components were obtained through extensive offline electromagnetic transient simulations of typical server racks, uninterruptible power supplies (UPS), and precision air conditioners in data centers. The simulations covered all possible combinations of the aforementioned current components, multidimensional harmonic vectors, and temperature, recording the non-ideal disturbances in the dq-axis current caused by the nonlinear characteristics of the load and temperature effects under various operating conditions, and storing these disturbances as compensation values ​​in the lookup table.

[0045] To construct the lookup table, the offline electromagnetic transient simulation was specifically performed in the PSCAD / EMTDC environment. The typical nonlinear load model parameters used in the simulation were modeled for the UPS load according to the IEC 62040-3 standard. The server rack was equivalent to a parallel model with 3rd, 5th, and 7th harmonic current sources, whose harmonic injection amplitudes showed a nonlinear proportional relationship with the fundamental current. The simulation operating range was set as follows: the fundamental current was 0.1 to 1.5 times the rated value; the amplitudes of the 3rd, 5th, 7th, and 11th harmonics in the load harmonic content data were set to vary independently within the ranges of 0% to 15%, 5% to 40%, 0% to 10%, and 0% to 8% of the fundamental amplitude, respectively; and the ambient temperature data was 25°C to 55°C. The input address dimensions of this high-dimensional lookup table are divided as follows: the fundamental d-axis and q-axis current dimensions are divided in 5% steps of the rated current; the amplitudes of the 3rd, 5th, 7th, and 11th harmonics are divided in 2% steps of the fundamental amplitude; and the ambient temperature is divided in 5°C steps. The disturbance compensation vector output by the lookup table is normalized. During real-time system operation, the module inputs the currently obtained raw load coordinate data, load harmonic content data, and ambient temperature data as high-dimensional addresses into the lookup table. Since real-time measurements typically do not fall exactly on the grid points of the lookup table, the module performs multilinear interpolation. Specifically, the system locates the smallest hypercube containing the current input address, reads the disturbance compensation vectors stored at all vertices of the hypercube, and then performs linear interpolation calculations sequentially in multiple dimensions based on the relative position of the current address point within the hypercube, ultimately accurately calculating the disturbance compensation vector that perfectly corresponds to the current operating condition.

[0046] Finally, the module performs vector subtraction. The d-axis component of the disturbance compensation vector generated by the interpolation is subtracted from the d-axis component of the original load coordinate data; simultaneously, the q-axis component of the disturbance compensation vector is subtracted from the q-axis component of the original load coordinate data. The result of this operation is the decoupled local load status data.

[0047] This process uses a model based on offline simulation and online interpolation to accurately remove disturbance components caused by the load's own characteristics (such as nonlinearity and temperature drift), thereby extracting a decoupled internal state that only reflects the essence of the interaction between the load and the external power grid, which greatly improves the sensitivity and accuracy of subsequent state deviation calculations.

[0048] Furthermore, the dynamic feedforward decoupling model is dynamically corrected through an online adaptive correction process. This correction process includes: defining the decoupled local load state data as the internal state to be estimated; generating expected original load coordinate data based on the currently estimated internal state, and using the dynamic feedforward decoupling model and the measured original load coordinate data, load harmonic content data, and ambient temperature data; comparing the original load coordinate data actually obtained by synchronous coordinate transformation at the same time with the expected original load coordinate data to generate a prediction residual; and recursively correcting the parameters in the dynamic feedforward decoupling model using the prediction residual and according to a preset recursive estimation algorithm, while simultaneously updating the internal state to be estimated and minimizing the prediction residual.

[0049] Specifically, to ensure that the dynamic feedforward decoupling model can adapt to the slow changes in data center load characteristics (such as server upgrades or equipment aging) over the long term, the system periodically performs an online adaptive correction process, for example, once every 100 power frequency cycles. This process uses the decoupled local load state data calculated in the previous moment as the best estimate of the current internal state to be estimated.

[0050] The first step in the correction process is state prediction. The system uses the estimated value of the current internal state, combined with the currently measured raw load coordinate data, load harmonic content data, and ambient temperature data, to apply a dynamic feedforward decoupling model in reverse. Specifically, the estimated internal state (d-axis and q-axis components), the measured harmonic multidimensional vector, and the temperature are used as address inputs. The current disturbance compensation vector is obtained from the model (i.e., a high-dimensional lookup table) through multilinear interpolation. Subsequently, this disturbance compensation vector is added to the internal state to be estimated using vector addition. The result is defined as the expected raw load coordinate data. This data represents the coupling state that the model predicts should exist in the dq coordinate system based on the current state.

[0051] The second step is to generate the prediction residual. The system performs a vector subtraction operation between the actual raw load coordinate data, which is directly calculated from the synchronous coordinate transformation at the same time, and the expected raw load coordinate data generated above. The differences between the components of the two vectors on the d-axis and q-axis together constitute a two-dimensional prediction residual vector. The magnitude and direction of this residual vector quantify the deviation between the current model prediction and the actual situation.

[0052] The third step is recursive correction. The system employs a pre-defined recursive estimation algorithm, specifically an unscented Kalman filter (UKF), to process the prediction residual, achieving joint estimation of the state and parameters. To achieve this, the augmented state vector of the UKF is defined to simultaneously include: the internal state to be estimated (i.e., the decoupled d-axis and q-axis components) and the model parameters to be corrected (i.e., the perturbation compensation vector values ​​at the lookup table grid points participating in the current multilinear interpolation operation). The state transition equation assumes that the augmented state vector remains constant within a correction period, with only process noise superimposed. The process noise covariance matrix is ​​set as a diagonal matrix, where the diagonal element values ​​corresponding to the internal state (e.g., set to the order of one ten-thousandth) are greater than the diagonal element values ​​corresponding to the model parameters (e.g., set to the order of one ten-millionth), reflecting the prior knowledge that the internal state changes faster than the model parameter aging. The observation vector is the actual measured raw load coordinate data (d-axis and q-axis components). The observation equation is a nonlinear function that represents the observation vector as the vector sum of the internal state components in the augmented state vector and a perturbation compensation vector, superimposed with observation noise. The perturbation compensation vector is obtained by using the internal state components in the augmented state vector, measured load harmonic content data, and ambient temperature data as inputs, combined with the model parameter components in the augmented state vector, and the perturbation compensation vector obtained by lookup table interpolation from the dynamic feedforward decoupling model, superimposed with observation noise. The observation noise covariance matrix is ​​set to a constant diagonal matrix (e.g., diagonal elements are set to the order of 1%) based on the accuracy of the analog-to-digital converter (ADC). Because the observation equation (which relies on multilinear interpolation operations using a high-dimensional lookup table) is highly nonlinear with respect to the augmented state vector and is not differentiable at the grid point boundaries of the lookup table, this embodiment employs an unscented Kalman filter. The filter propagates the mean and covariance of the state through a set of deterministically sampled sigma points. This process is achieved through an unscented transform, eliminating the need to calculate the partial derivative matrix (Jacobi matrix), and then performs the standard update step of Kalman filtering. The Kalman gain calculated in this update step determines the extent to which the predicted residuals can be used to correct the system state and model parameters.

[0053] The fourth step is model parameter updating. Here, the model parameters are the perturbation compensation vector values ​​stored at each grid point in the dynamic feedforward decoupling model (high-dimensional lookup table). The UKF output is a new, more accurate estimate of the augmented state vector after correction. This new estimate serves as the initial value for the next correction cycle, where the updated internal state components are used for subsequent state bias calculations, and the updated model parameter components are written back to the corresponding grid points in the high-dimensional lookup table. The weighting of the correction amount is inversely proportional to the weight of each vertex in the interpolation operation. To ensure convergence stability, a preset learning rate parameter (e.g., 0.05) is used to scale the correction amount, avoiding excessive adjustments to the model from a single bias. Through continuous recursive correction, the system continuously micro-investigates the data in the lookup table, ensuring that the model's predicted output continuously approximates the actual measurement results, thereby minimizing the prediction residual.

[0054] This adaptive correction process enables the decoupled model to have self-learning capabilities, allowing it to autonomously track and compensate for modeling errors introduced by changes in load characteristics or environmental drift, thus ensuring high accuracy and high reliability of the entire monitoring system during long-term operation.

[0055] Further, a state deviation calculation is performed on the local load status data and the network-side operating reference data to generate differential component data. The magnitude and phase angle of the differential component data together constitute the remote control indication data, including: defining the decoupled local load status data as a local coordinate dataset, which includes a d-axis local component and a q-axis local component; defining the network-side operating reference data as a network coordinate dataset, which includes a d-axis network component and a q-axis network component; generating differential component data composed of d-axis difference components and q-axis difference components by performing vector difference operations on the local coordinate dataset and the network coordinate dataset; obtaining the original magnitude and original phase angle by performing a Cartesian to polar coordinate transformation on the differential component data; and performing phase compensation operations on the original phase angle according to preset power grid topology transmission delay parameters to obtain the adjusted phase angle. The adjusted phase angle and the original magnitude together constitute the remote control indication data.

[0056] Specifically, this calculation process is executed in the remote indication signal generation module. First, the module directly defines the decoupled local load status data generated above as a local coordinate dataset, which consists of the d-axis local component and the q-axis local component.

[0057] Simultaneously, the module converts the network-side operating reference data (which includes grid frequency, effective voltage value, and voltage phase angle information) generated by the associated signal receiving module, representing the average operating conditions of the power grid, into a network coordinate dataset. Specifically, this conversion process involves using a rotating coordinate system reference synchronized with the local A-phase reference voltage, generated by the synchronous reference generation process (i.e., software phase-locked loop), to perform a synchronous rotating coordinate transformation (Park transformation) on the effective voltage value and phase angle information in the network-side operating reference data. This projects the data onto the same dq coordinate system as the local coordinate dataset, generating a network coordinate dataset containing d-axis network components and q-axis network components (representing the dynamic operating conditions of the power grid).

[0058] Subsequently, the module performs vector difference operations. The d-axis local component of the local coordinate dataset is subtracted from the d-axis network component of the network coordinate dataset to generate the d-axis difference component; the q-axis local component of the local coordinate dataset is subtracted from the q-axis network component of the network coordinate dataset to generate the q-axis difference component. These two difference components together constitute the differential component data, which accurately represents the degree and direction of dynamic deviation of the actual operating point of the local load from the macroscopic operating reference of the power grid in the dq coordinate system.

[0059] Next, the module performs a transformation from rectangular coordinates to polar coordinates on the differential component data. The d-axis difference component and the q-axis difference component are used as the x and y coordinates of a point on a two-dimensional plane. The distance from the point to the origin is calculated, and this distance is the original magnitude. At the same time, the angle between the line connecting the point and the origin and the positive direction of the d-axis is calculated, and this angle is the original phase angle.

[0060] After obtaining the original phase angle, the module performs a phase compensation calculation. This calculation aims to eliminate the impact of the inherent transmission delay of the power line from the intelligent insulated distribution cabinet to the main incoming transformer of the data center on control commands. A preset power grid topology transmission delay parameter, valued at 150 microseconds, is stored in the system. This parameter is calculated offline based on the actual length of the line segment, cable type, and impedance parameters. The phase compensation calculation converts this delay parameter into a phase lead compensation amount, calculated by multiplying the delay time (150 microseconds) by the power grid angular frequency (2π × 50 radians / second for a 50Hz power grid) to obtain a positive phase compensation angle. This compensation angle is then added to the original phase angle to obtain the adjusted phase angle.

[0061] Ultimately, the phase angle adjusted by phase compensation and the original unadjusted magnitude are combined to form a data structure, which is the remote control indication data. This data is transmitted to the remote control module for subsequent decision-making and control.

[0062] This process explicitly transforms high-dimensional, implicit state deviations into a two-dimensional polar coordinate command with clear physical meaning. The magnitude of the command intuitively reflects the severity of the deviation, while the phase angle, after time-delay compensation, precisely indicates the nature of the deviation (such as active / reactive disturbances), providing a high-quality decision-making basis for achieving precise and lag-free power grid control.

[0063] In a preferred embodiment, the remote indication signal generation module is further configured to: construct a time series from the differential component data generated at multiple consecutive times; and extrapolate the time series using a preset time series prediction model to generate a predictive deviation vector characterizing the evolution trend of the differential component within a preset time window; the remote control indication data also includes the predictive deviation vector.

[0064] Specifically, the remote indication signal generation module internally caches differential component data calculated over the past 200 power frequency cycles, forming a two-dimensional time series. A preset time series prediction model, an autoregressive integral moving average model (ARIMA(p, d, q)), is used to process this time series. The parameters (p, d, q) of this model are obtained through offline training on historical operating data from the data center. During real-time operation, the module inputs the latest time series into the ARIMA model, performs extrapolation calculations, and predicts the differential component data for the 5th power frequency cycle (100 milliseconds later). The predicted difference between the d-axis and q-axis components together constitute a predictive deviation vector. This vector is appended to the current remote control indication data and transmitted to the remote control module.

[0065] This approach enables the system to not only respond to current deviations but also anticipate their short-term development trends, achieving a shift from reactive control to predictive control and providing better control margins for handling slow disturbances with a development process.

[0066] Furthermore, the process of the adjusted phase angle and the original modulus value together constituting the remote control indication data also includes: using a preset control strategy partition plane divided by multiple operating area boundaries as a grading criterion, wherein the radial coordinate of the control strategy partition plane corresponds to the original modulus value and the polar coordinate corresponds to the adjusted phase angle; the operating area boundaries divide the control strategy partition plane into multiple control level areas; and by determining the region to which the real-time coordinate point formed by the original modulus value and the adjusted phase angle belongs within the control strategy partition plane, an additional control level identifier corresponding to the region is added to the remote control indication data.

[0067] Please see Figure 2 , Figure 2This is a schematic diagram of the control level division process based on the control strategy partition plane in an embodiment of the present invention. After obtaining the remote control indication data consisting of the original modulus and the adjusted phase angle, it also includes a process of adding a control level identifier.

[0068] Specifically, this process uses a preset control strategy partitioning plane to classify the risk level of remote control indication data. This partitioning plane is a polar coordinate system with the positive d-axis as the 0-degree reference. The radial coordinates of the plane are mapped to the original magnitude, representing the magnitude of the state deviation; the polar coordinates of the plane are mapped to the adjusted phase angle, representing the nature of the state deviation. This plane is divided by multiple preset operating area boundaries.

[0069] In this embodiment, the boundary of the operating area consists of two concentric circular arcs and four radial straight lines. The radial coordinates (original modulus values) of the two concentric circular arcs are R1=0.08 and R2=0.25, respectively. The polar coordinates (adjusted phase angles) of the four radial straight lines are θ1=45 degrees, θ2=135 degrees, θ3=225 degrees, and θ4=315 degrees, respectively. These boundaries divide the entire plane into five key control level regions, each corresponding to a unique control level identifier: Safety Zone: The central circular area with a radial coordinate less than R1 (0.08). Areas belonging to this zone are identified as "LEVEL_0_NORMAL (Normal State)".

[0070] Reactive power alarm zone: Two fan-ring regions with radial coordinates between R1 and R2 and polar coordinates between θ1 and θ2 or between θ3 and θ4. The identifier for this region is "LEVEL_1_REACTIVE_ADJUST (Reactive Power Alarm)".

[0071] Active alarm zone: The other two fan-ring regions whose radial coordinates are between R1 and R2 and whose polar coordinates are between θ2 and θ3 or between θ4 and θ1 are identified as “LEVEL_2_ACTIVE_ADJUST (Active Alarm)”.

[0072] Emergency Control Zone: All areas outside the ring with radial coordinates greater than R2 (0.25). The identifier for this zone is "LEVEL_3_URGENT_CONTROL (Emergency Control)".

[0073] Invalid data region: When radial or polar coordinates show non-numerical or other abnormal conditions, the region is identified as "LEVEL_E_INVALID (invalid data)".

[0074] During system operation, the system constructs a real-time coordinate point by combining the calculated radial coordinates and polar coordinates. Then, the system executes a region determination logic: first, it checks if the radial coordinate is less than R1; if so, it adds a "LEVEL_0_NORMAL (Normal State)" flag; if not, it checks if the radial coordinate is less than R2; if so, it adds either a "LEVEL_1_REACTIVE_ADJUST (Reactive Power Alarm)" or "LEVEL_2_ACTIVE_ADJUST (Active Power Alarm)" flag based on the phase angle interval; if the radial coordinate is greater than R2, it directly adds a "LEVEL_3_URGENT_CONTROL (Emergency Control)" flag. Through this process, a clear, machine-readable control level flag is added to the remote control indication data.

[0075] This classification process transforms continuously changing deviation values ​​into discrete classification identifiers with clear scheduling implications. It provides a clear decision-making entry point for remote control modules, enabling them to quickly match and execute differentiated control strategies based on different risk levels, thereby achieving classified and precise responses to power grid disturbances.

[0076] Furthermore, the remote control module includes: a strategy matching unit, which matches the control level identifier in the remote control instruction data with a preset dispatch response strategy library; an instruction generation unit, which, based on the matching result, indexes the instruction corresponding to the control level identifier from the dispatch response strategy library and outputs it; the dispatch response strategy library pre-stores instruction sets corresponding to different control level identifiers, the instruction sets including protective trip instructions with target switch equipment addresses and disconnection operation codes, topology reconfiguration instructions defining the coordinated operation sequence and target state of multiple switch equipment, and load transfer instructions setting the target tap position of on-load tap-changing transformers; and a communication encapsulation unit, which encapsulates the generated specific instructions and target equipment addresses into data packets conforming to the power automation communication protocol.

[0077] Specifically, the remote control module is the final execution end of the entire monitoring and control system. The strategy matching unit receives remote control instruction data with an attached control level identifier from the remote indication signal generation module. This unit internally maintains a pointer to the scheduling response strategy library. The scheduling response strategy library is a data structure preset in non-volatile memory, which uses the control level identifier as the primary key and the corresponding instruction set as the value.

[0078] The strategy matching unit extracts the control level identifier from the remote control instruction data and uses it as an index to search in the strategy library. For example, when the received identifier is "LEVEL_1_REACTIVE_ADJUST", the strategy matching unit locates the instruction set entry associated with that identifier in the strategy library.

[0079] The instruction generation unit reads the matched instruction set and generates the specific instruction. An example of an instruction set pre-stored in the scheduling response policy library is shown below: The instruction corresponding to "LEVEL_1_REACTIVE_ADJUST" is a load transfer instruction. Its specific content is: {Instruction type: load transfer, target device address: TR01_OLTC (Data Center No. 1 on-load tap changer), target status: tap position +1}.

[0080] The corresponding instruction for "LEVEL_2_ACTIVE_ADJUST" is a topology reconfiguration instruction. Specifically: {Instruction type: Topology reconfiguration, cooperating devices: [CB_A07, CB_B15], operation timing: [t0: disconnect CB_A07, t0+50ms: close CB_B15], target state: the load on server rack A is carried by bus B}.

[0081] The instruction corresponding to "LEVEL_3_URGENT_CONTROL" is a protective trip instruction. The specific content is: {Instruction type: protective trip, target device address: CB_F04 (outgoing circuit breaker No. 4 of this intelligent insulated distribution cabinet), operation code: TRIP}.

[0082] For “LEVEL_0_NORMAL” and “LEVEL_E_INVALID”: No instructions are generated, and the current state is maintained.

[0083] Based on the matching results, the instruction generation unit outputs the structured instruction content to the communication encapsulation unit.

[0084] The communication encapsulation unit receives the instruction content and converts it into a data packet conforming to the communication specifications of data center automation systems. In this embodiment, the unit uses the IEC 61850 standard for encapsulation. It maps the target device address in the instruction to a logical node address conforming to the standard, converts the opcode or target state into corresponding control block parameters, and adds necessary authentication and verification information to the entire data packet, ultimately generating a standard MMS protocol data packet. This data packet is then sent via industrial Ethernet to the corresponding switching equipment or control terminal in the data center to perform the final physical operation.

[0085] In a preferred embodiment, the remote control module further includes: an instruction arbitration unit, which, after the remote control module generates internal control instructions based on the remote control instruction data, receives external policy instructions from the external management system, and performs arbitration logic on the internal control instructions and the external policy instructions based on a preset instruction priority and security policy library, and outputs a conflict-free final remote control instruction.

[0086] Specifically, the instruction arbitration unit acts as the final security checkpoint before instruction output. It connects to the data center's infrastructure management system (DCIM) via the OPC UA standard interface to receive external policy instructions from the DCIM. These instructions include the target device address and an operation status code representing a 'manual lockout' state. Instruction priorities are predefined within the security policy library for different sources and types of instructions. The library defines the following priorities: P1 (highest): "protective trip instruction" generated by this system; P2: "manual lockout" instruction from the DCIM system; P3: "topology reconfiguration instruction" generated by this system; P4 (lowest): "economic mode operation" instruction from the DCIM system. The arbitration logic executes as follows: When the instruction generation unit outputs a "topology reconfiguration instruction" (P3 level), the instruction arbitration unit first queries the security policy library to check if a higher-priority "manual lockout" instruction (P2 level) exists for the target switching device. If it does, the arbitration logic outputs that the topology reconfiguration instruction should not be executed, and an instruction conflict event is reported to the system log. If there are no conflicting instructions with higher or equal priority, the "topology reconfiguration instruction" is recognized as the final remote control instruction and passed to the communication encapsulation unit.

[0087] By introducing an instruction arbitration unit, conflicts between the system's automated decision-making and higher-level management strategies and manual intervention in the data center are avoided, greatly improving the system's operational security and collaboration in complex management environments.

[0088] This module transforms abstract control levels into specific, executable equipment operation commands, which are then distributed via standardized communication protocols, forming a complete link from state perception to closed-loop control, ensuring a rapid, accurate, and automated response to power grid disturbances.

[0089] The intelligent monitoring system provided by this invention achieves accurate differentiation of disturbance sources by establishing a differential comparison framework between the locally decoupled load state and the network-side operating benchmark. Its feedforward decoupling operation removes dynamic components generated by internal factors such as load nonlinearity and temperature changes, resulting in differential component data that more accurately reflects the true interaction between the local node and the external power grid. Compared to traditional direct measurement or simple threshold judgment methods, this system has higher detection sensitivity for weak or early fault symptoms from the power grid side, while also exhibiting stronger anti-interference capabilities against normal large fluctuations in local load, reducing the probability of false alarms.

[0090] Example 2: This embodiment provides an intelligent monitoring method for intelligent insulated distribution cabinets applied to intelligent power distribution networks within T4-level data centers. In this scenario, the method is applied to an intelligent insulated distribution cabinet that supplies power to a specific server rack row in a certain area of ​​the data center, enabling refined monitoring and control of the power supply status of this critical load.

[0091] A method for intelligent monitoring of intelligent insulated distribution cabinets, referring to Figure 3 Specifically, it includes the following steps: First, local signal synchronization is performed. At the incoming line of the distribution cabinet, the A-phase reference voltage waveform is continuously acquired at a sampling rate of 25.6kHz, and its zero-crossing point from negative to positive is precisely captured as the synchronization trigger reference. Under the trigger of this synchronization reference, the instantaneous values ​​of the three-phase load current are synchronously acquired; for example, at a certain moment, the acquired values ​​are ia=180A, ib=-75A, and ic=-105A. Simultaneously, a Fast Fourier Transform is performed on the current data for one complete power frequency cycle (20 milliseconds) to calculate the load harmonic content data; for example, the current total harmonic current is 42.5A. In addition, the temperature sensor readings installed near the main busbar of the distribution cabinet are acquired, and the obtained ambient temperature data is 45°C. A timestamp associated with the synchronization reference is added to this temperature data.

[0092] Secondly, the generation of network-side operational baseline data is performed. Data packets conforming to the IEEE 1588 precision time protocol format are received from the industrial Ethernet within the data center. For example, candidate network status data from an upstream UPS output monitoring point with a specific number and a neighboring distribution cabinet with a specific number are received simultaneously. By comparing the timestamps of the data packets with the local clock, a data packet with a time deviation greater than 500 microseconds is eliminated. The data packets that pass the timeliness and CRC integrity checks are then weighted and averaged. Based on a preset correlation logic, a weighting coefficient of 0.6 is assigned to the data from the upstream UPS, and a weighting coefficient of 0.4 is assigned to the data from the neighboring distribution cabinet. Finally, a smoothed network-side operational baseline data is generated, representing the average operating conditions of the power grid as follows: frequency 50.02Hz, effective voltage 220.8V.

[0093] Next, the remote control instruction data is generated. This process first uses the acquired A-phase reference voltage data to generate a rotating coordinate system reference with the same frequency and phase via a software phase-locked loop. Then, using this reference, a Park transformation is performed on the synchronously acquired three-phase load current data to obtain the original load coordinate data composed of d-axis and q-axis components. For example, the calculated d-axis component is 210A and the q-axis component is -35A. Based on this, the obtained original load coordinate data (e.g., d-axis component 210A, q-axis component -35A), load harmonic content data (e.g., a multidimensional vector containing [3rd harmonic amplitude: 5.2A, 5th harmonic amplitude: 25.0A, 7th harmonic amplitude: 10.1A]), and ambient temperature data (45°C) are input as high-dimensional addresses into a pre-defined high-dimensional lookup table model established through offline simulation. By performing multilinear interpolation in the model, a disturbance compensation vector is generated to compensate for the load's own characteristics. For example, its d-axis compensation component is 3A and its q-axis compensation component is -40A. Subsequently, this disturbance compensation vector is subtracted from the original load coordinate data through vector subtraction to generate decoupled local load state data (d-axis: 210-3=207A; q-axis: -35-(-40)=5A). Then, this decoupled local load state data (d-axis: 207A; q-axis: 5A) is defined as a local coordinate dataset. At the same time, the network-side operating reference data (frequency 50.02Hz, effective voltage 220.8V) generated in the "second" step, which characterizes the average operating condition of the power grid, is synchronously transformed using the local rotating coordinate system reference to generate a network coordinate dataset. For example, its d-axis network component is calculated to be 206.95A and its q-axis network component is 4.9A. Subsequently, vector difference operations were performed on the local coordinate dataset and the network coordinate dataset to generate differential component data (d_diff=207A-206.95A=0.05A; q_diff=5A-4.9A=0.1A). Finally, a Cartesian to polar coordinate transformation was performed on the differential component data to obtain an original magnitude of 0.112 and an original phase angle of 63.4 degrees. Based on a preset transmission delay parameter of 150 microseconds, a 2.7-degree phase lead compensation was applied to the original phase angle to obtain an adjusted phase angle of 66.1 degrees. The final original magnitude and the adjusted phase angle together constitute the remote control indication data.

[0094] Finally, the remote control command is generated. The remote control instruction data (modulus 0.112, phase angle 66.1 degrees) generated in the previous step is mapped to the preset control strategy partition plane for judgment. Since the modulus 0.112 is greater than R1 (0.08) and less than R2 (0.25), and the phase angle 66.1 degrees is between θ1 (45°) and θ2 (135°), a control level identifier is added based on the region where this coordinate point is located, such as "LEVEL_1_REACTIVE_ADJUST". Based on this identifier, a match is performed in the preset scheduling response strategy library, and the corresponding command is indexed as "Set the target tap position of the on-load tap changer to +1". This command and the target device address are encapsulated into an MMS protocol data packet conforming to the IEC 61850 standard and sent over the network.

[0095] The intelligent monitoring method provided by this invention achieves accurate differentiation of disturbance sources by implementing a differential comparison process between the locally decoupled load state and the network-side operating benchmark. The feedforward decoupling operation step included in this method effectively removes dynamic components generated by internal factors such as load nonlinearity and temperature changes. This allows the final differential component data to more purely reflect the true interaction state between the local node and the external power grid. Compared to other methods based solely on direct measurement or simple threshold judgment, this method has higher detection sensitivity for weak or early fault symptoms from the power grid side, while also exhibiting stronger anti-interference capability against normal large fluctuations in local load, reducing the probability of false alarms.

[0096] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the scope of protection defined in the claims.

Claims

1. A smart monitoring system for a smart insulated power distribution cabinet, characterized in that, The application relates to a remote control system for power distribution cabinets, which comprises the following modules: a local signal acquisition module for acquiring reference voltage data from the input end of an insulated power distribution cabinet in a data center and synchronously acquiring three-phase load current data, load harmonic content data and environment temperature data of a local circuit; an associated signal receiving module for receiving candidate network state data containing time stamp information from a network and dynamically screening and synthesizing the candidate network state data according to a preset association logic to generate network side operation reference data; a remote indication signal generating module for taking the reference voltage data as a reference to synchronously rotate coordinate transform the three-phase load current data to obtain original load coordinate data, performing feedforward decoupling operation on the original load coordinate data based on the original load coordinate data, the load harmonic content data and the environment temperature data to generate decoupled local load state data, and performing state deviation calculation on the local load state data and the network side operation reference data to generate differential component data; the modulus and phase angle of the differential component data jointly constitute remote control indication data; a remote control module for generating remote control instructions according to the remote control indication data.

2. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 1, characterized in that, The local signal acquisition module comprises a synchronous reference generating unit for extracting a synchronous trigger pulse from the reference voltage data, a multi-channel data conversion unit for taking the synchronous trigger pulse as a conversion starting reference to synchronously collect three-phase load current parameters representing the state of a power grid connection point, a spectrum analysis unit for performing fast Fourier transform operation on the conversion results of the three-phase load current to generate load harmonic content parameters reflecting local load characteristics, and an environment data collection unit for collecting the environment temperature data and adding a time stamp associated with the synchronous trigger pulse to the environment temperature data.

3. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 1, characterized in that, The associated signal receiving module comprises a protocol decoding unit for decoding the candidate network state data and associated time stamps from data packets in a format complying with a precise time protocol, a data screening unit for eliminating data not meeting preset time effectiveness and integrity requirements by judging the time deviation of the time stamps and checking the integrity of the data packets, and a data synthesis unit for applying preset weighting coefficients to all screened candidate network state data and performing weighted average operation to generate the network side operation reference data representing the average operation condition of a power grid after smoothing.

4. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 1, characterized in that, The process of generating the decoupled local load state data comprises the following steps: performing synchronous reference generation processing on the reference voltage data to generate a rotating coordinate system reference in the same frequency and phase as the reference voltage data; using the rotating coordinate system reference to synchronously coordinate transform the three-phase load current data to obtain original load coordinate data composed of d-axis components and q-axis components; taking the original load coordinate data, the load harmonic content data and the environment temperature data as multi-dimensional address inputs, generating a disturbance compensation vector in a preset dynamic feedforward decoupling model through multilinear interpolation operation; and performing vector subtraction operation on the d-axis components and q-axis components of the original load coordinate data and the corresponding components of the disturbance compensation vector to generate the decoupled local load state data.

5. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 4, characterized in that, The dynamic feedforward decoupling model is further dynamically corrected through an online adaptive correction process, the correction process comprising: defining the decoupled local load state data as an internal state to be estimated; generating expected original load coordinate data based on the current estimated internal state, and using the dynamic feedforward decoupling model and the measured original load coordinate data, load harmonic content data and ambient temperature data; comparing the original load coordinate data actually obtained by synchronous coordinate transformation at the same time with the expected original load coordinate data to generate a prediction residual; and recursively correcting parameters in the dynamic feedforward decoupling model according to a preset recursive estimation algorithm through the prediction residual, and updating the internal state to be estimated to minimize the prediction residual.

6. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 1, characterized in that, State deviation calculation is performed on the local load state data and the network-side operation reference data to generate differential component data; The modulus and phase angle of the differential component data jointly constitute remote control instruction data, comprising: defining the decoupled local load state data as a local coordinate data set, the local coordinate data set including a d-axis local component and a q-axis local component; defining the network-side operation reference data as a network coordinate data set, the network coordinate data set including a d-axis network component and a q-axis network component; generating differential component data composed of a d-axis difference component and a q-axis difference component through vector difference operation on the local coordinate data set and the network coordinate data set; obtaining an original modulus and an original phase angle through polar coordinate transformation on the differential component data; performing phase compensation operation on the original phase angle according to a preset power grid topology transmission delay parameter to obtain an adjusted phase angle; and the adjusted phase angle and the original modulus jointly constitute the remote control instruction data.

7. The intelligent monitoring system of the intelligent insulating power distribution cabinet according to claim 6, characterized in that, The process of jointly constituting the remote control instruction data by the adjusted phase angle and the original modulus further comprises: taking a preset control strategy partition plane divided by a plurality of operation region boundaries as a hierarchical criterion, the radial coordinate of the control strategy partition plane corresponding to the original modulus and the polar angle coordinate corresponding to the adjusted phase angle; the operation region boundaries divide the control strategy partition plane into a plurality of control level regions; and by judging the belonging region of a real-time coordinate point composed of the original modulus and the adjusted phase angle in the control strategy partition plane, the remote control instruction data is additionally attached with a control level identifier corresponding to the belonging region. 8.The intelligent monitoring system of the intelligent insulation power distribution cabinet according to claim 1, characterized in that, The remote control module comprises: a policy matching unit that matches a regulation level identifier in remote regulation and control instruction data with a preset dispatch response policy library; an instruction generation unit that indexes out an instruction corresponding to the regulation level identifier from the dispatch response policy library according to a matching result and outputs the instruction; the dispatch response policy library pre-stores instruction sets corresponding to different regulation level identifiers, the instruction sets comprising protective tripping instructions of target switch device addresses and disconnection operation codes, topology reconstruction instructions defining a plurality of switch device cooperative operation time sequences and target states, and load transfer instructions setting target tap positions of on-load voltage regulating transformers; and a communication packaging unit that packages the generated specific instruction and the target device address into a data packet conforming to a power automation communication protocol.

9. A method for intelligently monitoring an intelligent insulated power distribution cabinet, characterized in that, Comprise: acquire reference voltage data from an input end of an insulated power distribution cabinet of a data center, and synchronously acquire three-phase load current data, load harmonic content data and environment temperature data of a local circuit; receive candidate network state data containing timestamp information from a network, and generate network side operation reference data by dynamically screening and synthesizing the candidate network state data according to a preset relevance logic; take the reference voltage data as a reference to perform synchronous rotating coordinate transformation on the three-phase load current data, and obtain original load coordinate data; perform feedforward decoupling operation on the original load coordinate data based on the original load coordinate data, the load harmonic content data and the environment temperature data, to generate decoupled local load state data; perform state deviation calculation on the local load state data and the network side operation reference data, to generate differential component data; and the modulus and phase angle of the differential component data jointly constitute remote regulation and control instruction data; generate remote control instructions according to the remote regulation and control instruction data.

Images