Microgrid coordination control system based on decentralized control

Abstract

This invention relates to the field of distributed control of microgrids and coordinated control of power systems, specifically a coordinated control system for microgrids based on distributed control. The system includes: collecting distributed node voltage amplitude, frequency, active power, reactive power, node connection relationships, node load data, and timestamped consistent status messages; performing timing alignment; constructing ideal state variables and ideal power reference values ​​based on the current topology and initial load; injecting delay, drift, and bias factors to generate simulated state data; forming actual residual vectors and theoretical residual vectors; calculating matching values ​​based on consistent status messages and outputting state evaluation results; generating control commands and issuing them to the corresponding nodes. This invention can distinguish between normal source-load disturbances and coordinated control anomalies, reducing the risk of false alarms and missed alarms.

Description

Technical Field

[0001] This invention relates to the field of distributed control of microgrids and coordinated control of power systems, specifically a coordinated control system for microgrids based on distributed control. Background Technology

[0002] In the existing microgrid collaborative control system, the microgrid includes photovoltaic nodes, energy storage nodes, diesel generator nodes and load nodes. Each distributed node exchanges status messages through communication links and combines operating parameters such as voltage, frequency, active power and reactive power to complete power distribution and voltage and frequency regulation under decentralized control, so that the microgrid can maintain stable operation in islanded or weak grid scenarios.

[0003] Existing systems typically monitor and judge the operating status of nodes based on the power measurement results or local control information uploaded by each node. Some solutions also combine node topology relationships, control command information or preset control models to identify abnormal fluctuations, which are used to determine whether frequency offset, voltage fluctuation or power sharing imbalance in the microgrid belongs to abnormal operating conditions.

[0004] However, in distributed control scenarios, factors such as load fluctuations, communication delays, sensor drift, message misordering, and control parameter mismatch are often coupled with each other. Judging solely based on real-time measurements or simple threshold rules can easily lead to misjudging normal load changes and output fluctuations as control layer coordination failures, or mistaking real coordination loss of synchronization for environmental disturbances. This results in insufficient accuracy of microgrid state assessment, causing problems such as untimely adjustment of control strategies or excessive intervention. Summary of the Invention

[0005] The purpose of this invention is to provide a microgrid collaborative control system based on distributed control, and to solve the following technical problems:

[0006] It avoids misjudging normal microgrid source-load changes as control layer faults and makes it easier to accurately distinguish between environmental fluctuations and collaborative control anomalies in high-reliability power supply scenarios, thereby effectively reducing the risk of false alarms and false alarms in system status assessment and enabling targeted adaptive collaborative handling.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A distributed control-based microgrid collaborative control system includes:

[0009] The data acquisition module is used to collect the voltage amplitude, frequency, active power, reactive power, node connection relationship information, node load data at a preset start time, and node consistency status messages with timestamps from distributed nodes.

[0010] The timing alignment module is used to perform time alignment of the voltage amplitude, frequency, active power, and reactive power based on the timestamp;

[0011] The ideal benchmark construction module is used to determine the current topology based on the node connection relationship information, determine the initial load based on the node load data, and solve the ideal state variables and ideal power reference values ​​according to the preset control equations.

[0012] The parameter injection module is used to introduce delay factors, drift factors, and bias factors that characterize the heterogeneous communication hysteresis and physical measurement errors of the microgrid into the ideal state variables and the ideal power reference values, and generate simulation state data through time-domain iterative calculation.

[0013] The dual-track residual generation module is used to generate real residual vectors and theoretical residual vectors based on the time-aligned voltage amplitude, frequency, active power, reactive power, ideal state variables, ideal power reference values, and simulation state data, according to the correspondence of state parameters with the same physical meaning and dimensions.

[0014] The coupling decision module is used to calculate the matching value between the actual residual vector and the theoretical residual vector by combining the consistency status messages between the nodes and output the status evaluation result;

[0015] The adaptive coordination module is used to generate control commands based on the state assessment results and send them to the corresponding distributed node underlying controllers via a peer-to-peer communication network link.

[0016] Furthermore, the inter-node consistency status message also includes a virtual voltage regulation command sequence and a virtual frequency regulation command sequence;

[0017] The inter-node consistency status message is used to characterize the peer-to-peer collaborative control process between adjacent distributed nodes, and is used by the coupled decision module to verify the collaborative consistency between distributed nodes.

[0018] Furthermore, the preset control equations include droop control equations and / or virtual synchronous generator equations;

[0019] The droop control equation is used for droop control nodes, and the virtual synchronous generator equation is used for virtual synchronous generator nodes.

[0020] The ideal state variables include one or more of the following: node voltage magnitude, node frequency, and node phase angle.

[0021] Furthermore, the ideal reference construction module is used to reconstruct the ideal state variables and the ideal power reference values ​​under the conditions of zero communication delay, zero sensor error, and zero line parameter mutation, based on preset Kirchhoff constraints.

[0022] Furthermore, the parameterized injection module is used to map the delay factor into a time-shift operator acting on the node state time series.

[0023] The drift factor is mapped to a colored noise matrix with the same dimension as the node state variables.

[0024] The bias factor is mapped to a step bias function that acts on the corresponding node state variable.

[0025] The discrete-time state update calculation is then performed based on the mapping result to generate the simulation state data.

[0026] Furthermore, the dual-track residual generation module is specifically used for:

[0027] The time-aligned voltage amplitude and frequency are subtracted from the corresponding ideal state variables, and the time-aligned active power and reactive power are subtracted from the corresponding ideal power reference values. These are then combined to generate the actual residual vector.

[0028] The theoretical residual vector is generated by subtracting the corresponding state variables in the simulation state data from the ideal state variables and the ideal power reference values ​​using the same physical state parameters.

[0029] Furthermore, the coupling decision module is used to employ a multidimensional dynamic time warping algorithm to calculate the matching value between the actual residual vector and the theoretical residual vector in the time dimension and the node association structure dimension, and to generate the state evaluation result based on the matching value.

[0030] Furthermore, the coupling decision module is preset with a first matching threshold and a second matching threshold for the associated spatiotemporal matching dimension, and the first matching threshold is greater than the second matching threshold;

[0031] The first matching threshold is determined by the lower boundary of the matching value calculated from the historical known communication layer fault sample set injected into the system in the offline state; the second matching threshold is determined by the upper boundary of the matching value calculated by the system under the normal source-load disturbance sample set.

[0032] When the matching value is greater than the first matching threshold, the control layer collaborative fault result is output.

[0033] When the matching value is less than the second matching threshold, the environmental fluctuation result is output.

[0034] When the matching value is greater than or equal to the second threshold and less than or equal to the first threshold, the result to be reviewed is output.

[0035] Furthermore, the adaptive coordination module is used to perform communication link reconstruction or control parameter resetting when the state evaluation result is the control layer coordination failure result;

[0036] When the state assessment result is the environmental fluctuation result, the current collaborative strategy shall be maintained;

[0037] When the state evaluation result is the result to be reviewed, the data acquisition module is triggered to increase sampling density, the time alignment module is triggered to update the time alignment result, and the coupling decision module is triggered to recalculate the matching value.

[0038] The control commands are used for node power allocation correction, frequency regulation correction, or voltage regulation correction.

[0039] The beneficial effects of this invention are:

[0040] 1. This invention constructs an ideal operating benchmark and actively injects disturbances such as delay and drift to generate simulation state data. By comparing the dual-track residuals of actual measurements and simulation states deviating from the ideal benchmark, the system can accurately distinguish between normal source-load physical fluctuations and control layer coordination anomalies. This effectively avoids misjudging load changes as control faults or mistaking coordination loss for environmental disturbances, and significantly improves the accuracy of microgrid state assessment and the timeliness of control strategy adjustment.

[0041] 2. This invention introduces a virtual voltage and frequency adjustment command sequence into the consistency status message, which explicitly preserves the dynamic process characteristics of inter-node collaborative negotiation. This mechanism not only verifies the final steady-state recovery result, but also traces the initiation order and adjustment direction of node response actions, thereby accurately distinguishing whether system fluctuations originate from normal physical follow-up responses or control imbalances caused by communication delays and misordering, thus enhancing the ability to trace anomalies.

[0042] 3. This invention establishes an ideal model by matching heterogeneous preset control equations to the dynamic characteristics of different distributed power devices; it uses droop control to simulate the rapid response of energy storage nodes and uses virtual synchronous machine equations to deduce the damping and inertial transition process of generator nodes; this mechanism enables the ideal benchmark to accurately adapt to the heterogeneous physical mechanism of the underlying multi-energy equipment, avoiding the misjudgment of the inherent mechanical inertial delay of the equipment as a control layer fault due to the fixed and single model.

[0043] 4. Under the assumptions of zero communication delay and zero sensor error, this invention reconstructs an ideal benchmark synchronously with the current network structure based on Kirchhoff constraints. This overcomes the data lag problem caused by traditional predictions based on historical averages, and makes the reconstructed ideal state strictly conform to the power conservation and distribution law of the real physical power grid. This mechanism establishes a reference surface that is not affected by historical states, providing a highly reliable comparison benchmark for measuring actual deviations.

[0044] 5. This invention transforms the control layer problem into a reusable disturbance template, mapping delay, drift, and bias factors to time-shift operators, colored noise, and step functions, respectively, thereby generating various engineering simulation trajectories. This mechanism transforms abstract control anomalies into intuitive and comparable dynamic evolution characteristics, enabling the system to specifically characterize and identify the causes of specific faults, rather than merely remaining at the shallow stage of detecting fluctuations.

[0045] 6. This invention adopts a dual-track residual generation mechanism to project the actual operation and the simulation operation with abnormal characteristics into a coordinate system that deviates from the ideal reference. By combining and comparing multi-dimensional data such as voltage, frequency and power, it not only prevents high-amplitude physical quantities from masking key small deviation changes, but also realizes the characteristic homology verification across dimensions, nodes and anomalies.

[0046] 7. This invention employs a multidimensional dynamic time warping algorithm to calculate the spatiotemporal coupling matching degree of the dual-track residuals. This algorithm allows for finite discrete misalignment of the residuals in the time dimension, while rigorously verifying the correlation direction and temporal order of the spread of abnormal deviations in the network topology. This accurately captures the unique local misalignment and spatial homogeneity characteristics of decentralized control anomalies, significantly improving the tolerance to small random delays and the robustness of discrimination. Attached Figure Description

[0047] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 A schematic diagram of a distributed control-based microgrid collaborative control system provided in an embodiment of this application. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Please see Figure 1 The microgrid collaborative control system based on distributed control includes: a data acquisition module, used to acquire the voltage amplitude, frequency, active power, reactive power, node connection relationship information, node load data at a preset start time, and node consistency status messages with timestamps of distributed nodes;

[0051] The timing alignment module is used to align voltage amplitude, frequency, active power, and reactive power based on timestamps.

[0052] The ideal baseline construction module is used to determine the current topology based on node connection relationship information, determine the initial load based on node load data, and solve the ideal state variables and ideal power reference values ​​according to the preset control equations.

[0053] The parameter injection module is used to introduce delay factors, drift factors, and bias factors that characterize the heterogeneous communication hysteresis and physical measurement errors of the microgrid into the ideal state variables and ideal power reference values, and generate simulation state data through time-domain iterative calculation.

[0054] The dual-track residual generation module is used to generate real residual vectors and theoretical residual vectors based on the time-aligned voltage amplitude, frequency, active power, reactive power, ideal state variables, ideal power reference values, and simulation state data, according to the correspondence of state parameters with the same physical meaning and dimensions.

[0055] The coupling decision module is used to calculate the matching value between the actual residual vector and the theoretical residual vector by combining the consistency state messages between nodes and output the state evaluation result; the adaptive coordination module is used to generate control commands based on the state evaluation result and send them to the corresponding distributed node underlying controller through the peer-to-peer communication network link.

[0056] This embodiment provides a microgrid collaborative control mechanism based on distributed control; specifically, the system is deployed in an island general hospital in an isolated power supply scenario. The hospital has photovoltaic nodes, energy storage nodes, diesel generator nodes, and several key load nodes, including the operating room, intensive care unit, medical imaging center, and central oxygen production station.

[0057] The distributed nodes do not rely on a single master station to make real-time power allocation, but maintain voltage and frequency stability through peer-to-peer communication between adjacent nodes;

[0058] Because such scenarios involve both fluctuations in photovoltaic output caused by rapid shading from island clouds and control layer anomalies such as communication link congestion, sensor aging, and control message offset, the system incorporates physical layer power and information layer collaborative signaling into the analysis link to avoid misjudging normal source load changes as control faults.

[0059] In the specific control logic processing steps, the data acquisition module acquires the voltage amplitude, frequency, active power and reactive power of each node in each sampling period, and at the same time reads the node connection relationship information, initial load data and consistency status messages with timestamps.

[0060] The node connections here not only reflect who is connected to whom, but also reflect the hospital's current operating mode, such as the actual network structure after switching from daytime photovoltaic integration to nighttime energy storage as the main mode and diesel engine hot standby.

[0061] The timing alignment module performs unified timing adjustment on data streams from different sources. For example, if node N1 uploads frequency data at 10:00:00.100, node N2 uploads active data at 10:00:00.120, and node N3 uploads consistency messages at 10:00:00.118, the system aligns them to the same analysis window to avoid mistaking the difference in sampling order as control oscillation.

[0062] The ideal baseline construction module, given the current topology and initial load, forms the voltage, frequency, and power distribution state that each node should exhibit at this moment if there is no communication delay, no measurement deviation, and no sudden changes in line parameters.

[0063] The parameter injection module does not directly judge the real data, but actively adds control layer perturbations such as delay, drift and bias to the above ideal benchmark to obtain a set of simulation state data, which is used to express how the system will usually evolve if there is a control layer anomaly.

[0064] The dual-track residual generation module forms two parallel analysis paths: one consists of the deviation of real-time measurements from the ideal benchmark, reflecting the comprehensive fluctuations in reality;

[0065] Another path consists of the deviation of the simulation state from the ideal baseline, reflecting the typical multidimensional characteristic distribution of control layer anomalies; the coupled decision module combines the consistency state message to compare the similarity of the two residual trajectories in the time evolution and node propagation direction, and outputs the state evaluation result.

[0066] The adaptive coordination module issues control commands based on the evaluation results, which can perform power redistribution, frequency correction or voltage correction for the corresponding node;

[0067] In the system's fault tolerance mechanism, if a node loses contact for a short period of time or there is a missing measurement in a certain sampling period, the system does not directly give a conclusion of control layer failure. Instead, it marks the missing data as a low-confidence sample and maintains a temporary benchmark with the most recent effective window.

[0068] If the missing node persists for more than the preset time, it will be downgraded and removed from the current coupled calculation. Only the remaining connected subnets will be analyzed to avoid individual node anomalies from causing bias in the overall network evaluation.

[0069] If the topology information changes within the analysis window, for example, when the diesel engine switches from hot standby to load, the system first freezes the judgment results of the previous window, and then reconstructs the ideal baseline according to the new topology to avoid using the old baseline to interpret the new operating conditions during the topology transition.

[0070] If both the actual residual and the theoretical residual are weak, and the consistency message does not exhibit abnormal propagation characteristics, then the existing coordination strategy will be maintained, and unnecessary control reconfiguration will not be triggered.

[0071] On the evening before the typhoon made landfall, the island's general hospital entered disaster prevention power supply mode; the rooftop photovoltaic system was blocked by thick clouds, and the load on the radiology CT unit and negative pressure ward increased simultaneously, with energy storage nodes and diesel generator nodes sharing the frequency support.

[0072] At this time, the data acquisition module found that the frequency of multiple nodes dropped slightly, but the transmission order and update time of consistency messages between nodes remained normal.

[0073] The ideal benchmark construction module forms an ideal operating reference based on the current topology and initial load of photovoltaic power output reduction, energy storage compensation, and diesel engine gradual increase; the parameterized injection module constructs several possible control layer abnormal trajectories, one of which is that the frequency correction action of adjacent nodes after the message is delayed spreads in a relay manner.

[0074] After system comparison, it was found that although the actual residual fluctuates, its propagation direction mainly follows the location of the load surge and does not show the topological propagation characteristics of control commands being delayed layer by layer. Therefore, for output environmental fluctuation-related results, the adaptive coordination module only maintains the current distributed control strategy and does not directly cut off nodes or reconstruct communication.

[0075] The purpose of this step is to break down complex fluctuations in reality into interpretable physical deviations and verifiable control anomaly data features. By combining ideal benchmarks, parametric simulations, and dual-track residuals, the system can distinguish between normal source-load disturbances and coordinated control anomalies in high-reliability power supply scenarios such as hospitals, thereby reducing the risk of false alarms and missed alarms.

[0076] In a preferred embodiment of the present invention, the inter-node consistency status message further includes a virtual voltage adjustment command sequence and a virtual frequency adjustment command sequence; the inter-node consistency status message is used to characterize the peer-to-peer collaborative control process between adjacent distributed nodes, and is used to couple the decision module to verify the collaborative consistency between distributed nodes.

[0077] This embodiment provides a message enhancement mechanism for inter-node collaborative consistency; specifically, in the aforementioned island general hospital scenario, simply collecting results such as voltage, frequency, and power may still encounter identification bottlenecks where the performance characteristics are similar but the driving mechanisms are different.

[0078] Because both normal load fluctuations and control out-of-synchronization can cause frequency shifts, but the mutual coordination and adjustment processes between nodes are different; therefore, this embodiment further retains virtual voltage adjustment command sequences and virtual frequency adjustment command sequences in the consistency status message to record the cooperative correction intentions sent by each node to its neighboring nodes.

[0079] In the specific control logic processing steps, the virtual voltage regulation command sequence can be understood as the reactive power support suggestion trajectory issued by a node to adjacent nodes after detecting a local bus voltage deviation; the virtual frequency regulation command sequence is used to record the collaborative intent related to active power support and frequency recovery.

[0080] Taking a local subnet consisting of three nodes N1, N2, and N3 as an example, if N1 detects that the operating room load has increased and the frequency has dropped, it sends a frequency adjustment command to N2. N2 then decides whether to continue to propagate the support request to N3 based on its local energy storage capacity.

[0081] The system not only verifies the final steady-state recovery result of the network frequency, but also verifies the consistency of the order of node adjustment actions, the order of response actions, and the direction of adjustment. This kind of process information can help the coupled decision module distinguish whether a frequency fluctuation is a normal coordinated following response caused by physical load, or an imbalance in coordinated control caused by message communication delay, reordering, or instruction tampering.

[0082] In the system's fault tolerance mechanism, if the message only contains the running result status but lacks the adjustment instruction sequence, the system can still make a preliminary judgment based on the basic state residual, but will mark this time window as medium confidence level to avoid making excessive intervention fault conclusions when there is insufficient observation information.

[0083] If the virtual voltage regulation command sequence has breaks or missing segments, but the virtual frequency regulation command sequence is complete, then the frequency coordination command chain will be checked for consistency first, and the voltage coordination information will only be used as an auxiliary feature reference.

[0084] If the neighboring node's message record has executed the cooperative action, but the actual physical power response output locally has not appeared for a long time, the system will mark the situation where the instruction exists but the physical response is missing as a highly abnormal cooperative mechanism mismatch, and hand it over to the subsequent coupled calculation module for key review;

[0085] During the typhoon, the output power of the hospital's rooftop photovoltaic system dropped sharply, and the energy storage nodes continuously sent frequency support coordination requests to the diesel generator nodes.

[0086] If only the frequency state curve is observed, the mechanical ramp-up hysteresis of the diesel engine itself and the communication link delay may both cause the system frequency recovery to lag.

[0087] After introducing virtual frequency regulation command sequence observation, the system found that the coordination requests of energy storage nodes had large-scale timing misalignment in the consistency messages of adjacent nodes, and some nodes repeatedly executed the instructions of the previous cycle. This indicates that the core reason for the abnormal fluctuations was not simply caused by the drop in photovoltaic output, but rather a serious failure in the synchronization of messages in the coordination chain.

[0088] Conversely, if the interactive command chain is complete and the response order of each node is completely consistent with the physical distance of the communication topology, it indicates that the system is more likely to be in a steady-state coordinated adjustment process under normal source-load disturbance.

[0089] The purpose of this mechanism is to explicitly retain and record the characteristics of the collaborative negotiation process between control nodes in a distributed control system, thereby enabling traceable verification of the evolution of peer-to-peer collaborative relationships. This allows the system to trace the basis of the interactive instructions for the current adjustment actions output by each node and accurately analyze the driving root of the dynamic response process.

[0090] In a preferred embodiment of the present invention, the preset control equations include droop control equations and / or virtual synchronous generator equations; droop control equations are used for droop control nodes, and virtual synchronous generator equations are used for virtual synchronous generator nodes; ideal state variables include one or more of node voltage magnitude, node frequency, and node phase angle.

[0091] This embodiment provides an ideal control characterization mechanism for heterogeneous distributed power sources; specifically, in the microgrid of a general hospital on an island, the dynamic characteristics of different power supply equipment are not the same.

[0092] Energy storage converters typically have rapid droop regulation capabilities. In order to enhance the stability of grid voltage and frequency support under islanded grid operation, diesel generators are often configured with control strategies featuring virtual rotational inertia characteristics.

[0093] If a single, identical ideal control model is used to represent all nodes, the constructed ideal reference benchmark will produce a serious distortion deviation from the actual physical response, greatly interfering with the subsequent residual generation and accurate identification of control layer anomalies.

[0094] In the specific control logic processing steps, droop control is more suitable for characterizing fast power sharing nodes. Its physical meaning is that when the system frequency is low, the node immediately increases active power support according to the pre-set droop slope; when the voltage is low, the node increases reactive power support in response.

[0095] Virtual synchronous generator control is more suitable for characterizing nodes with a synchronous machine-like operating mechanism and inertial response characteristics. It not only considers the steady-state power sharing calculation among multiple sources, but also introduces virtual rotor motion equations to simulate the process characteristics of speed inertia, system damping coefficient and power angle evolution.

[0096] Therefore, before constructing an ideal baseline, the system needs to first identify the type of power equipment configured for distributed nodes, then apply droop control logic to simulate energy storage nodes, and match virtual synchronous generator logic to perform deduction calculations for diesel engine or synchronous motor nodes.

[0097] Accordingly, the ideal state variables of the calculated output are not limited to node voltage amplitude and node frequency parameters, but can also include node output phase angle data, which can be used to dynamically reflect the direction of active power exchange between nodes and the dynamic control response rhythm at different times.

[0098] To illustrate with a simplified scenario example, if N1 is an energy storage node with a fast response speed and N2 is a diesel generator node with rotating parts, then the control characteristics of N1 focus on the rapid sharing mechanism of transient power when disturbance occurs, while the control characteristics of N2 focus on the smooth transition and overshoot suppression mechanism when disturbance occurs. The ideal evolution trajectory of the two when the system is disturbed should be constructed strictly according to their physical mechanism differences.

[0099] In the system's fault tolerance mechanism, if the control type of a certain distributed node is missing in the system's equipment file, the system can first perform temporary type clustering classification based on its historical response characteristics. For example, it can prioritize observing the speed of its response time constant to step frequency deviation input and whether there is an obvious damped inertial transition process in the output waveform, and select a control equation that better fits its dynamic characteristics.

[0100] If the operating mode of a certain energy storage node undergoes active reconfiguration switching within the current analysis time window, such as seamlessly switching from a constant power grid-connected mode following the grid to a grid-type control mode with the dominant frequency, the system will strictly separate the time window data before and after the switching breakpoint for segmented modeling and solution, avoiding mixing the steady-state results of two completely different underlying control mechanisms into the ideal benchmark construction system with the same time domain period.

[0101] If the power angle or phase angle data of some nodes cannot be directly measured due to the lack of measurement conditions of the equipment, the system can temporarily form a basic set of ideal state variables based only on the node voltage and node frequency parameters. When the high-precision synchronous phasor measurement device deployed later is online and available, the phase angle dimension data can be expanded and fused to further converge the ideal reference estimation error and improve the identification accuracy and robustness of the corresponding control abnormal events.

[0102] During the emergency power supply phase of the hospital's isolated grid at night, the energy storage nodes undertake the tasks of rapid active power compensation and high-frequency frequency stabilization, while the large-capacity diesel generators undertake the tasks of basic power support and continuous safe power supply. When the large heavy-duty purification air conditioning equipment in the hospital's operating room is started, the global frequency of the system will be impacted and experience short-term fluctuations and drops.

[0103] If the diesel generator node adopts the same droop model and ideal benchmark construction method as the energy storage node, the system will incorrectly presuppose that it can instantly complete the support output action with the same slope as the fast response energy storage without delay. This will lead to the generator's normal mechanical inertia delay transition process being incorrectly judged as communication delay or control instability or other abnormal behavior of the control layer for this node in the subsequent dual-track residual comparison and judgment process.

[0104] After adopting the heterogeneous control precise characterization mechanism in this embodiment, the ideal state of the energy storage node calculation output is mainly presented as an instantaneous step injection fast response, while the ideal state of the diesel generator node calculation output is presented as a gradually damped dynamic convergence trajectory of phase angle and frequency parameters over time. Both types of power nodes with different physical characteristics participate in the bidirectional residual comparison judgment according to the real physical operation constraint parameters of their respective underlying devices.

[0105] The fundamental purpose of this mechanism is to achieve high-precision adaptation between the ideal benchmark model and the heterogeneous control mechanism at the bottom layer of real multi-energy power equipment, thereby realizing accurate mathematical modeling and characterization of the dynamic physical behavior of heterogeneous power nodes in microgrids, and avoiding the risk of system tolerance error judgment caused by the inherent physical properties of equipment due to the overly singular and rigid assumptions of the underlying ideal model.

[0106] In a preferred embodiment of the present invention, the ideal reference construction module is used to reconstruct ideal state variables and ideal power reference values ​​under the conditions of zero communication delay, zero sensor error and zero line parameter mutation, based on preset Kirchhoff constraints.

[0107] This embodiment provides a constraint establishment mechanism for ideal benchmark reconstruction. Specifically, in the aforementioned scenario, if the ideal benchmark is obtained only from historical averages or short-term moving averages, then when the hospital's operation mode undergoes topological reconstruction, the old data will cause the new reasonable state estimate to have historical dependence, making the benchmark itself lagging. To overcome this problem, this embodiment uses Kirchhoff constraints and ideal control boundaries to construct a reference field that changes synchronously with the current network structure.

[0108] In the specific control logic processing steps, the so-called zero communication delay, zero sensor error and zero line parameter mutation does not mean that the actual system can achieve this state. Rather, it is to artificially set a reference environment without additional interference to determine the following theoretical benchmark: under the current node connection relationship, the current starting load and the given control mode, how the system should distribute power, maintain voltage and stabilize frequency.

[0109] Kirchhoff constraints ensure that the current, voltage, and power exchange of each branch satisfy the basic conservation relationship of the power grid, so that the reconstructed ideal state is not based on empirical prediction, but conforms to the laws of the physical network itself. For hospital microgrids, this benchmark can truly reflect the reasonable results of which node should bear more support according to its topological position and which node should follow with a smaller amplitude when the load of the imaging center increases.

[0110] Specifically, the pre-set Kirchhoff constraints are manifested in the construction of the node admittance matrix of the power grid, which equates each distributed node to an injection source, and the load node to a constant power or constant impedance model.

[0111] The system solves the preset control equations and the node voltage equations based on the admittance matrix simultaneously through a numerical iterative algorithm. Under the assumption that all communication and sensing are ideal, the theoretical voltage, frequency and power flow of each node are calculated iteratively, which are used as ideal state variables and ideal power reference values.

[0112] A simplified scenario can be used to illustrate this: If the local network only contains photovoltaic nodes, energy storage nodes, and the operating room load bus, and the energy storage is closer to the load bus, then the ideal power reference value will usually prioritize the pattern of energy storage supporting first and photovoltaic supplementing according to its output capacity, rather than simply distributing it equally.

[0113] In the system's fault tolerance mechanism, if the line parameters cannot be updated in real time, the system can call the most recently verified parameter file to establish an ideal benchmark, and mark the reliability level of the benchmark.

[0114] If subsequent line maintenance, switch switching, or cable temperature rise causes parameter changes to exceed expectations, a baseline reconstruction will be triggered immediately. If a subnet is temporarily isolated and forms an independent operating unit, the system will apply Kirchhoff constraints to that subnet separately to avoid solving the problem as if it were still connected to the whole network.

[0115] If there are obvious abnormal jumps in the load start data, such as a false alarm code from the acquisition equipment causing the oxygen production station load to double instantaneously, the system will first cross-verify with adjacent metering points before deciding whether to adopt the data into the ideal baseline.

[0116] In the first hour after the hospital entered the disaster-time isolated grid operation, the load on the disinfection supply center and the intensive care unit gradually increased; if the average value of past sunny daytime is used as a reference directly, it will be wrong to assume that photovoltaic is still the main support source;

[0117] After adopting this embodiment, under the current conditions of thick cloud cover, online energy storage, and diesel engine with base load, the system reconstructs the ideal state based on node connections and load locations, and obtains a power reference value that is closer to the physical laws at this moment.

[0118] If a deviation from reality occurs subsequently, the system can determine whether the deviation originates from external load changes or from an anomaly in the collaborative control chain;

[0119] The purpose of this mechanism is to establish an ideal reference surface that follows the conservation laws of the power grid and is not affected by historical lag, so as to achieve a reliable measurement of deviations from actual operation.

[0120] In a preferred embodiment of the present invention, the parameterization injection module is used to map the delay factor to a time shift operator acting on the node state time series, map the drift factor to a colored noise matrix with the same dimension as the node state variable, map the bias factor to a step bias function acting on the corresponding node state variable, and perform discrete-time state update calculation based on the mapping results to generate simulation state data.

[0121] This embodiment provides a parameterized injection mechanism for constructing control layer anomaly images. Specifically, after the aforementioned ideal benchmark is established, if the system only observes the difference between the real data and the ideal benchmark, it may still be unable to determine what specific control layer anomaly the deviation feature corresponds to. Therefore, this embodiment further transforms common control layer problems into reusable disturbance templates, enabling the system to generate several simulation state trajectories with engineering significance.

[0122] In the specific control logic processing steps, the delay factor is mapped to a time shift effect, which mainly corresponds to problems such as communication link congestion, message queuing, wireless relay congestion, or edge controller processing lag. In physical terms, it often causes the adjustment actions of adjacent nodes to be staggered instead of simultaneous.

[0123] The drift factor is mapped to a disturbance that accumulates at a low frequency over time. It is used to describe sensor temperature drift, sampling reference offset, or measurement link aging. Its typical characteristic is not an instantaneous jump, but a long-term bias in a certain direction with low-frequency fluctuations. The bias factor is mapped to a step effect. It is suitable for expressing the situation where the state variable of a certain node is suddenly raised or lowered. For example, the consistency variable is abnormally rewritten, causing the subsequent control actions to deviate as a whole after a certain moment.

[0124] The system superimposes these factors onto the ideal state and updates them step by step in discrete time to form multiple types of simulation state data; when performing discrete time state update calculations, the system processes the data frame by frame according to the time step.

[0125] To ensure the clarity of the computational logic and the uniqueness of feature generation, let the sequence of the evolution of an ideal state variable in a certain dimension over time be denoted as . The corresponding generated simulation state time series is ,in This represents the current discrete time step; for the delay factor, it is mapped to a time shift step constant. The system shifts the ideal state time series backward by a specified step size; that is, the delayed simulation state update rule is:

[0126]

[0127] For the drift factor, the system accumulates a sequence of colored noise components generated by the autoregressive moving average model at the ideal state value of the current step size. The drift simulation state update rule is as follows:

[0128]

[0129] For the bias factor, the system at the set mutation time Then, a preset amplitude is superimposed on the corresponding ideal state constant. The step change, i.e., the bias simulation state update rule, is: when the time step is less than the abrupt change time... When, the calculation formula is:

[0130]

[0131] When the time step is greater than or equal to the mutation moment When, the calculation formula is:

[0132]

[0133] These disturbance terms, when superimposed on the ideal state according to the corresponding rules, together constitute a simulation trajectory with specific control anomaly characteristics; a simplified scenario can be described as follows: if the frequency adjustment message of node N2 is delayed by one analysis step, the frequency recovery action of N2 and its successor neighbors will be shifted back as a whole.

[0134] If the measured value of node N3 continues to rise, the coordinated quantity it publishes will deviate more and more from the actual load; if the virtual voltage command of N1 is stepped up, the relevant reactive power support chain will be biased as a whole after that moment.

[0135] In the system's fault tolerance mechanism, if a certain type of disturbance factor lacks support from equipment history or operation and maintenance experience, the system will not forcibly generate such an image, but will retain other known disturbance templates to prevent misleading matching with unreliable templates.

[0136] If multiple disturbance factors may coexist, such as the superposition of communication delay and sensor drift, the system can generate a combined simulation trajectory, but sets an upper limit on the number of combinations and prioritizes retaining the template that is most consistent with the current message anomaly characteristics.

[0137] If the simulation state exceeds the device's allowed operating boundary within a local window, such as showing a power support intention that is clearly inconsistent with the actual power supply capacity, the system will automatically trim the trajectory and not use it as a valid source of theoretical residuals.

[0138] During continuous high humidity in the hospital, the temperature sensing module of a certain energy storage converter aged, and the wireless bridging link between hospital buildings was congested during the typhoon. The system injected two types of factors, chronic measurement drift and message time shift, into the ideal benchmark to form two sets of simulation state data.

[0139] Comparing the actual fluctuations, it was found that the actual frequency recovery process is more like a relay-style delayed adjustment action rather than a slow drift of the measured value. This indicates that the communication layer coordination problem is more worthy of attention than simply sensor aging.

[0140] The purpose of this mechanism is to transform abstract control layer anomalies into comparable engineering trajectories, thereby enabling targeted characterization of the causes of anomalies, rather than remaining at a level where fluctuations exist but their causes are unclear.

[0141] In a preferred embodiment of the present invention, the dual-track residual generation module is specifically used to: subtract the time-aligned voltage amplitude and frequency from the corresponding ideal state variables, subtract the time-aligned active power and reactive power from the corresponding ideal power reference values, and combine them to generate a real residual vector; and subtract the corresponding state variables in the simulation state data from the ideal state variables and the ideal power reference values ​​according to the same physical state parameters, and combine them to generate a theoretical residual vector.

[0142] This embodiment provides a dual-track residual generation mechanism; specifically, given the existing ideal benchmark and simulation state data, the system does not directly compare the original waveform, but projects the actual operation and the simulation operation onto a common coordinate system of the degree of deviation from the ideal state.

[0143] The reason for this is that there are many original dimensions in the hospital microgrid, and the absolute values ​​of voltage, frequency, active power and reactive power span multiple orders of magnitude. Direct comparison may be risked by a high-amplitude quantity masking other key changes.

[0144] In the specific control logic processing steps, the real residual vector is composed of the voltage amplitude, frequency, active power and reactive power after time alignment, which are compared with the corresponding ideal state or ideal reference value. Its essence is to quantify the deviation amplitude and deviation direction of the real system relative to the ideal model.

[0145] The theoretical residual vector is obtained by comparing the simulation state data with the same ideal benchmark. Its essence is to characterize the typical evolution characteristics of the system deviating from the ideal model if a certain type of control layer anomaly exists.

[0146] To illustrate with a simplified scenario, suppose that within a certain window, the frequency and active power of node N1 deviate significantly, while the voltage and reactive power of node N2 deviate significantly. In this case, the actual residual will retain this multi-dimensional, cross-node deviation characteristic.

[0147] If the delayed injection simulation also produces a trajectory that starts with a frequency shift in N1 and then propagates to the voltage support imbalance in N2, the theoretical residual will exhibit similar structural characteristics; the subsequent comparison of the system is not of the original values, but whether these deviation characteristics originate from the same source.

[0148] In the system's fault tolerance mechanism, if a certain type of measurement is unavailable under the current equipment configuration, such as individual nodes only being able to stably provide active power and frequency, the system can generate a dimension-reduced residual, perform a dual-track comparison only on the available dimensions, and mark the dimension integrity in the results; if there is a significant risk of dimensional imbalance in a certain dimension, the system can first perform engineering normalization processing.

[0149] The specific logic of the normalization process is as follows: extract the maximum absolute deviation threshold of the residual dimension within the set historical normal operation phase as the benchmark scale, divide the currently generated original residual value by the benchmark scale, and map it into a dimensionless ratio sequence so that small frequency deviations are not completely submerged by large power changes, and ensure that each physical quantity is processed by the coupled decision module under equal weight.

[0150] If the actual residual is extremely large and the theoretical residual cannot be explained, the system does not forcibly classify it as a control layer fault, but marks it as an out-of-model event candidate and submits it for subsequent review, such as investigating sudden short circuits, protection actions, or equipment hardware failures.

[0151] During the peak nighttime operation period of the hospital's imaging center, the active power demand near node N1 increases rapidly, while node N2 provides reactive power support, causing fluctuations in both frequency and voltage.

[0152] The system first generates the actual residuals, showing that the active power deviation near the load connection point comes first, followed by the voltage deviation of the adjacent support node; at the same time, a certain type of delayed injection simulation generated by the system also shows a similar cross-node sequential relationship, so the theoretical residuals and the actual residuals can be matched in the subsequent process.

[0153] If another type of bias injection simulation shows that the state variables of a certain node suddenly increase as a whole, but the actual residual does not have a corresponding abrupt change, then the theoretical trajectory will be naturally excluded;

[0154] The purpose of this mechanism is to map real-world operation and abnormal simulation to the same deviation space, thereby achieving comparable expressions across dimensions, nodes, and causes.

[0155] In a preferred embodiment of the present invention, the coupling decision module is used to employ a multidimensional dynamic time warping algorithm to calculate the matching value between the actual residual vector and the theoretical residual vector in the time dimension and the node association structure dimension, and generate a state evaluation result based on the matching value.

[0156] This embodiment provides a residual matching decision mechanism oriented towards spatiotemporal coupling; specifically, although the aforementioned dual-track residual has transformed data from different sources into a comparable form, in a distributed control scenario, anomalies do not always occur strictly synchronously.

[0157] A certain collaborative failure often manifests as an abnormal state deviation occurring first at one node, and then spreading to other associated nodes along the communication topology or physical line connection relationship; therefore, if a static node-by-node comparison method is used at the same absolute moment, it is easy to miss the true association determination due to the propagation lag characteristic.

[0158] In the specific control logic processing steps, the role of the multidimensional dynamic time warping algorithm is to allow the actual residual and the theoretical residual to have a limited discrete alignment misalignment in the time dimension, while requiring that the evolution process of this alignment misalignment still conforms to the physical or communication association structure of the node.

[0159] In other words, the system not only compares the waveform similarity of the residual sequences in the time domain, but also compares the consistency between the propagation timing of state deviations in the node network topology and the diffusion path.

[0160] Specifically, the execution steps of the multidimensional dynamic time warping algorithm are as follows: the actual residual vector and the theoretical residual vector are spatially aligned and sorted according to the node topological adjacency matrix to form their respective multidimensional spatial state matrices at the time step.

[0161] Calculate the actual residuals and theoretical residuals at any two different time steps. and The Euclidean distance between the spatial state matrices is denoted as the single-step distance. The cumulative distance matrix is ​​constructed using a dynamic programming algorithm, with a set time step. and The corresponding cumulative distance is Using recursive rules:

[0162]

[0163] in, , and These represent the cumulative distances of adjacent preceding step length node pairs. A regularized mapping path with the minimum total distance from the starting point to the ending point is searched in the cumulative distance matrix. The normalized path metric parameter is obtained by dividing the sum of the cumulative distances of all nodes on the path by the total number of planning steps of the path. The reciprocal of the metric parameter is then taken to convert it into the final matching value, thereby quantitatively evaluating the synergistic consistency of the two residual trajectories in terms of temporal misalignment and spatial diffusion mechanisms.

[0164] For example, in a local subnet, there is a collaborative interaction chain from N1 to N2 and then to N3. If the theoretical residual shows an abnormal deviation in frequency, it should first appear at node N1, propagate to node N2, and eventually affect node N3. Although the actual residual may be ahead or behind by microseconds in absolute acquisition time, its propagation sequence still follows the same diffusion topology direction, so a higher matching value can be obtained by calculation.

[0165] Conversely, if the actual residuals first appear at edge nodes that have no topological connection with the theoretical propagation starting point, or if the propagation time sequence conflicts with the network connection relationship, then even if the local time series waveform values ​​are similar, they will not be identified as belonging to the same type of control layer anomaly mechanism.

[0166] In the system's fault tolerance mechanism, if the network structure associated with a node undergoes reconstruction changes within the current analysis window, the system first segments the residual sequence according to the topology switching time node, and then performs spatiotemporal coupling matching independently for each segment to avoid mixing data from different network topologies and causing misjudgment.

[0167] If the duration of the actual residual feature is significantly shorter than the theoretical residual reference feature time, for example, it corresponds to only a very brief transient load impact, the system allows local time segment alignment but will reduce the confidence interval of the matching result to prevent occasional high-frequency spikes from being over-interpreted as complete control layer failure events.

[0168] If multiple theoretical residual templates all obtain similar matching values, the system does not directly output a single deterministic conclusion, but retains the priority ranking sequence of candidate templates and waits for subsequent sampling windows to supplement more observational data evidence;

[0169] After the compressor unit of the oxygen plant near the hospital's emergency building starts up, the local network near node N1 first experiences a frequency drop, the energy storage node N2 responds and intervenes, and then the diesel generator node N3 slowly takes over the support through a damping response.

[0170] If this phenomenon is part of a normal load increase process, the high-amplitude area of ​​the actual residual often spreads radially along the physical load access location; if this phenomenon is caused by an abnormal control layer coordination mechanism, the deviation evolution of the actual residual tends to follow the communication propagation path of the control signaling message and spread topologically.

[0171] The system performs a spatiotemporal coupling similarity comparison operation between the extracted real residual trajectory and multiple types of theoretical residual trajectories. It finds that the characteristics of a type of theoretical trajectory where communication link delay causes adjacent nodes to successively undergo hysteresis adjustment response show a high degree of spatial consistency with the cross-node diffusion order of the real residual. Therefore, it outputs a higher matching value that represents a highly suspected fault.

[0172] The purpose of this mechanism is to capture and quantify the unique temporal local misalignment but topologically homogeneous physical characteristics of the abnormal evolution process of distributed control systems, thereby achieving a more robust anomaly detection capability than the traditional absolute time static threshold comparison method, which is more tolerant of small random delays.

[0173] In a preferred embodiment of the present invention, the coupling decision module presets a first matching threshold and a second matching threshold for the associated spatiotemporal matching dimension, and the first matching threshold is greater than the second matching threshold; the first matching threshold is determined by the lower boundary of the matching value calculated by injecting a historical known communication layer fault sample set in the offline state of the system; the second matching threshold is determined by the upper boundary of the matching value calculated by the system under the normal source-load disturbance sample set;

[0174] When the matching value is greater than the first matching threshold, the control layer collaborative fault result is output; when the matching value is less than the second matching threshold, the environmental fluctuation result is output; when the matching value is greater than or equal to the second threshold and less than or equal to the first threshold, the result to be reviewed is output.

[0175] This embodiment provides a hierarchical result output mechanism. Specifically, a single matching value cannot directly meet the judgment conditions for reliable operation in engineering because the control operation of the hospital microgrid is safety sensitive. If the judgment threshold is set too low, normal load fluctuations may be mistaken for faults and falsely trigger reconfiguration. If the judgment threshold is set too high, it may lead to the omission of real collaborative faults. A lower threshold corresponds to a situation where the actual deviation is significantly inconsistent with the theoretical abnormal structural characteristics. In this case, it is more likely to be a normal environmental fluctuation, such as cloud cover, load access, or periodic switching of temperature control equipment.

[0176] The reason for setting up a verification zone between the two is that in distributed control scenarios, there are often blurred boundary features such as short-term topology switching, local missing measurements, and superposition of disturbances. If a binary division is forcibly applied, it is easy to give an unstable conclusion under the critical state.

[0177] Simplified scenario description: If the best match of a window is greater than the first threshold, then directly enter the control layer exception process; if it is less than the second threshold, then maintain the existing strategy; if it falls between the two thresholds, it means that there is a certain similarity but the confidence is not sufficient, and further sample collection is required.

[0178] In the system's fault tolerance mechanism, if the high and low thresholds are no longer suitable for the current system due to seasonal changes, equipment aging, or topology reconstruction, the system can periodically update the threshold levels based on long-term operation and maintenance records. However, the updates should be performed offline or through configuration verification to avoid frequent online drift that could cause instability in the judgment criteria.

[0179] If the matching values ​​corresponding to multiple theoretical templates are simultaneously in the verification area, the system will prioritize the template that is more consistent with the abnormal characteristics of the message and continue to observe it. If the actual operation is during a major event, such as an emergency switch-off of the entire hospital or the protection action has just ended, the system can temporarily raise the threshold for entering the fault conclusion to avoid excessive control during the system recovery period.

[0180] During typhoon nights, the load on hospital negative pressure wards and refrigeration systems fluctuated repeatedly, and photovoltaic output essentially disappeared; within a certain analysis window, the actual residual and the theoretical trajectory of link delay type have certain similarities, but are also mixed with obvious random fluctuations of source load, and the matching result is in the middle range.

[0181] At this point, the system does not immediately identify a communication layer fault, nor does it simply regard it as a natural fluctuation. Instead, it outputs a result to be verified, leaving room for the next stage of enhanced sampling and recalculation. Only if multiple windows subsequently display high matching will it proceed with stronger processing actions.

[0182] The purpose of this mechanism is to establish a tiered decision-making strategy that meets engineering safety requirements, thereby achieving a robust output that allows for rapid handling when confidence is high, careful observation when confidence is low, and delayed review when confidence is insufficient.

[0183] In a preferred embodiment of the present invention, the adaptive coordination module is used to perform communication link reconstruction or control parameter resetting when the state assessment result is a control layer coordination failure result; to maintain the current coordination strategy when the state assessment result is an environmental fluctuation result; and to trigger the data acquisition module to increase sampling density, trigger the timing alignment module to update the time alignment result, and trigger the coupling decision module to recalculate the matching value when the state assessment result is a result to be reviewed. The control commands are used for node power allocation correction, frequency adjustment correction, or voltage adjustment correction.

[0184] This embodiment provides an adaptive collaborative processing mechanism that is linked to the judgment result; specifically, in a scenario where power supply cannot be interrupted, such as a hospital microgrid, identification is only the first half, and the real key is how to process it after identification.

[0185] If the system takes the same action for all anomalies, such as reconstructing the communication link or adjusting the control parameters in all cases, it will not only increase operational disturbances, but may also introduce additional instability into the otherwise normal source load fluctuations; therefore, this embodiment makes the handling strategy correspond one-to-one with the state assessment results.

[0186] In the specific control logic processing steps, when the result is a control layer coordination failure, the adaptive coordination module can prioritize the reconfiguration of the communication link, such as switching to a backup wireless relay, rerouting through a fiber optic ring network, or bypassing the abnormal edge controller.

[0187] Control parameter resetting can also be performed, such as adjusting the droop coefficient, restoring the damping settings of the virtual synchronous generator, or redistributing the main support nodes to block abnormal propagation paths;

[0188] When the result is environmental fluctuations, the system maintains the current collaborative strategy, only allowing nodes to complete active, reactive, voltage and frequency regulation according to the existing decentralized control law, without making additional reconfiguration, so as to avoid excessive intervention in normal fluctuations.

[0189] When the result is pending verification, the system enters a cautious enhanced observation mode: increasing the sampling density of key nodes, shortening the alignment window of adjacent messages, recalculating the matching value, and deciding whether to upgrade the handling based on the recalculation result;

[0190] The control commands are applied to node power allocation correction, frequency regulation correction, and voltage regulation correction. Among them, power allocation correction is more inclined to active power support path adjustment, frequency regulation correction is more inclined to fast frequency stabilization strategy, and voltage regulation correction is more inclined to reactive power support and bus voltage maintenance.

[0191] In the system's fault tolerance mechanism, if the system is determined to be due to a control layer coordination failure, but the backup communication link is also unavailable, then the system first uses control parameter resetting and local autonomous limiting to maintain stability, and then waits for the link to recover.

[0192] If environmental fluctuations continue to increase and approach the equipment safety boundary, even if a control layer fault is not currently identified, the system can still implement conservative power limits without changing the nature of the coordination mechanism to prevent battery or diesel engine overload.

[0193] If the newly added data still contradicts the data during the review period, the system will not infinitely loop the recalculation, but will mark the relevant nodes as key inspection targets, while keeping the entire network operating under the most conservative and feasible strategy.

[0194] On the night the typhoon made landfall, the hospital's oxygen production station experienced a sudden increase in load, and the system initially provided results pending verification. The adaptive coordination module then commanded the energy storage node, diesel engine node, and adjacent busbars of the oxygen production station to increase sampling density and shorten the alignment period of consistency messages.

[0195] After recalculation, the actual residual is highly consistent with the theoretical trajectory of frequency cooperative chain delay propagation. The system then upgrades the result to a control layer cooperative failure, automatically switches to the backup communication path, and fine-tunes the frequency support parameters of the energy storage node so that it can undertake a more direct frequency stabilization task before communication is restored.

[0196] Conversely, if the recalculation reveals that the residual structure mainly changes with the load location rather than spreading with the message chain, then the existing strategy is maintained, and each node is only allowed to complete the power supply guarantee according to the predetermined power allocation.

[0197] The purpose of this mechanism is to transform the identification results into differentiated and executable on-site control actions, thereby enabling rapid suppression of real faults, restrained intervention in normal fluctuations, and safe verification of gray zone states.

[0198] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A microgrid collaborative control system based on distributed control, characterized in that, include: The data acquisition module is used to collect the voltage amplitude, frequency, active power, reactive power, node connection relationship information, node load data at a preset start time, and node consistency status messages with timestamps from distributed nodes. The timing alignment module is used to perform time alignment of the voltage amplitude, frequency, active power, and reactive power based on the timestamp; The ideal benchmark construction module is used to determine the current topology based on the node connection relationship information, determine the initial load based on the node load data, and solve the ideal state variables and ideal power reference values ​​according to the preset control equations. The parameter injection module is used to introduce delay factors, drift factors, and bias factors that characterize the heterogeneous communication hysteresis and physical measurement errors of the microgrid into the ideal state variables and the ideal power reference values, and generate simulation state data through time-domain iterative calculation. The dual-track residual generation module is used to generate real residual vectors and theoretical residual vectors based on the time-aligned voltage amplitude, frequency, active power, reactive power, ideal state variables, ideal power reference values, and simulation state data, according to the correspondence of state parameters with the same physical meaning and dimensions. The coupling decision module is used to calculate the matching value between the actual residual vector and the theoretical residual vector by combining the consistency status messages between the nodes and output the status evaluation result; The adaptive coordination module is used to generate control commands based on the state assessment results and send them to the corresponding distributed node underlying controllers via a peer-to-peer communication network link.

2. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The inter-node consistency status message also includes a virtual voltage adjustment command sequence and a virtual frequency adjustment command sequence; The inter-node consistency status message is used to characterize the peer-to-peer collaborative control process between adjacent distributed nodes, and is used by the coupled decision module to verify the collaborative consistency between distributed nodes.

3. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The preset control equations include droop control equations and / or virtual synchronous generator equations; The droop control equation is used for droop control nodes, and the virtual synchronous generator equation is used for virtual synchronous generator nodes. The ideal state variables include one or more of the following: node voltage magnitude, node frequency, and node phase angle.

4. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The ideal reference construction module is used to reconstruct the ideal state variables and the ideal power reference values ​​based on preset Kirchhoff constraints, under conditions of zero communication delay, zero sensor error, and zero line parameter mutation.

5. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The parameterized injection module is used to map the delay factor into a time-shift operator that acts on the node state time series. The drift factor is mapped to a colored noise matrix with the same dimension as the node state variables. The bias factor is mapped to a step bias function that acts on the corresponding node state variable. The discrete-time state update calculation is then performed based on the mapping result to generate the simulation state data.

6. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The dual-track residual generation module is specifically used for: The time-aligned voltage amplitude and frequency are subtracted from the corresponding ideal state variables, and the time-aligned active power and reactive power are subtracted from the corresponding ideal power reference values. These are then combined to generate the actual residual vector. The theoretical residual vector is generated by subtracting the corresponding state variables in the simulation state data from the ideal state variables and the ideal power reference values ​​using the same physical state parameters.

7. The microgrid collaborative control system based on distributed control according to claim 1, characterized in that, The coupling decision module is used to employ a multidimensional dynamic time warping algorithm to calculate the matching value between the actual residual vector and the theoretical residual vector in the time dimension and the node association structure dimension, and to generate the state evaluation result based on the matching value.

8. The microgrid collaborative control system based on distributed control according to claim 7, characterized in that, The coupling decision module is preset with a first matching threshold and a second matching threshold for the associated spatiotemporal matching dimension, and the first matching threshold is greater than the second matching threshold; The first matching threshold is determined by the lower boundary of the matching value calculated from the historical known communication layer fault sample set injected into the system in the offline state; the second matching threshold is determined by the upper boundary of the matching value calculated by the system under the normal source-load disturbance sample set. When the matching value is greater than the first matching threshold, the control layer collaborative fault result is output. When the matching value is less than the second matching threshold, the environmental fluctuation result is output. When the matching value is greater than or equal to the second matching threshold and less than or equal to the first matching threshold, the result to be reviewed is output.

9. The microgrid collaborative control system based on distributed control according to claim 8, characterized in that, The adaptive coordination module is used to perform communication link reconstruction or control parameter resetting when the state evaluation result is the control layer coordination failure result; When the state assessment result is the environmental fluctuation result, the current collaborative strategy shall be maintained; When the state evaluation result is the result to be reviewed, the data acquisition module is triggered to increase sampling density, the time alignment module is triggered to update the time alignment result, and the coupling decision module is triggered to recalculate the matching value. The control commands are used for node power allocation correction, frequency regulation correction, or voltage regulation correction.

Images