Micro-grid distributed data acquisition control system considering communication delay

Abstract

The application relates to the technical field of micro-grid control, and particularly discloses a micro-grid distributed data acquisition control system considering communication delay, which is applied to an alternating-current micro-grid containing multiple distributed inverters, each of which is configured with a local control unit; the local control unit sequentially comprises, from top to bottom, an adaptive high-frequency data acquisition layer, which is used for acquiring a real-time operation state of the micro-grid, extracting local high-frequency electrical parameters, and dynamically adjusting a data acquisition frequency and a communication period according to the real-time operation state. The application scheme effectively reduces the negative influence of communication delay and equipment parameter drift on the cooperative control instruction, reduces the cooperative error and reactive current circulation among the inverters, and improves the dynamic adjustment capability of the micro-grid system under the condition of a complex communication network.

Description

Technical Field

[0001] This invention relates to the field of microgrid control technology, and in particular to a distributed data acquisition and control system for microgrids that takes into account communication latency. Background Technology

[0002] Existing AC microgrids typically consist of multiple distributed inverters. Their decentralized collaborative control relies on communication networks to exchange power status data between adjacent nodes to regulate system frequency and voltage. In real-world industrial or remote deployment environments, these communication networks are susceptible to interference, leading to dynamically changing delays and packet loss in data transmission.

[0003] In addition, existing control systems mostly use fixed data acquisition frequencies and fixed communication cycles for data interaction, without considering adaptive adjustments based on changes in power grid operating conditions and communication network quality. This results in the inefficient use of limited communication bandwidth or the inability to transmit some critical operating data in a timely manner.

[0004] Because the power state data of adjacent nodes acquired by the control system has a time lag, and the physical filtering parameters of the distributed inverter power supply will objectively drift with changes in the operating environment, it is difficult for a state observer with fixed parameters to accurately extrapolate the state of nodes with communication delays. When the distributed consensus algorithm uses lagging data that has a time lag and has not been corrected for physical parameters for calculation, it will output biased frequency and voltage compensation commands. This situation will cause power distribution errors and reactive power circulating currents when the distributed inverter power supply performs droop control, affecting the overall coordinated regulation performance and operational stability of the microgrid. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the object of this invention is to propose a distributed data acquisition and control system for microgrids that takes into account communication delays, so as to improve the dynamic adjustment capability of microgrid systems under complex communication network conditions.

[0006] To achieve the above objectives, a first aspect of the present invention proposes a distributed data acquisition and control system for microgrids that takes into account communication delays. This system is applied to an AC microgrid containing multiple distributed inverters, each of which is equipped with a local control unit. The local control unit comprises, from top to bottom:

[0007] An adaptive high-frequency data acquisition layer is used to acquire the real-time operating status of the microgrid, extract local high-frequency electrical parameters, and dynamically adjust the data acquisition frequency and communication cycle according to the real-time operating status.

[0008] A multi-mode communication link dynamic switching layer is used to monitor the quality status of multiple communication links and dynamically select the best link to transmit the local high-frequency electrical parameters.

[0009] The dual adaptive asynchronous time-delay state observation layer is used to combine the local high-frequency electrical parameters with the real-time communication link status to identify the filtering parameters of the distributed inverter power supply online in order to achieve adaptive adjustment of the state observer model and reconstruct the current delay-free real power state vector of the adjacent distributed inverter power supply.

[0010] A distributed consensus decision layer is used to perform distributed consensus calculations based on the delay-free true power state vector, generating frequency compensation terms and AC bus voltage amplitude compensation terms.

[0011] The pure analog adaptive droop execution layer is used to inject the frequency compensation term and AC bus voltage amplitude compensation term into the physical droop control loop of the distributed inverter power supply in the form of analog signals, so as to complete the closed-loop dynamic adjustment of the microgrid.

[0012] To achieve the above objectives, a second aspect of the present invention provides a distributed data acquisition and control method for microgrids that takes into account communication delays, applied to an AC microgrid containing multiple distributed inverters, comprising the following steps performed by a local control unit configured at each of the distributed inverters:

[0013] The system acquires the real-time operating status of the microgrid, extracts local high-frequency electrical parameters, and dynamically adjusts the data acquisition frequency and communication cycle based on the real-time operating status.

[0014] Monitor the quality status of multiple communication links and dynamically select the best link to transmit the local high-frequency electrical parameters;

[0015] By combining the local high-frequency electrical parameters with the real-time communication link status, the filtering parameters of the distributed inverter power supply are identified online to achieve adaptive adjustment of the state observer model, and the current zero-delay real power state vector of the adjacent distributed inverter power supply is reconstructed.

[0016] Distributed consensus calculation is performed based on the aforementioned delay-free true power state vector to generate frequency compensation term and AC bus voltage amplitude compensation term.

[0017] The frequency compensation term and the AC bus voltage amplitude compensation term are injected into the physical droop control loop of the distributed inverter power supply in the form of analog signals to complete the closed-loop dynamic regulation of the microgrid.

[0018] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described microgrid distributed data acquisition and control method taking into account communication latency.

[0019] In the actual operation of a microgrid, the technical solution of this invention improves the reliability of underlying electrical data transmission by coordinating an adaptive high-frequency data acquisition layer and a multi-mode communication link dynamic switching layer, dynamically adjusting the data interaction frequency and selecting the optimal communication path based on the real-time state of the power grid and the link quality. In the face of physical communication delays, the dual adaptive asynchronous time-delay state observation layer dynamically corrects the observation model by combining online identified filtering parameters, restoring the current delay-free true power state vector of adjacent inverter power sources. Based on this delay-free true data, the distributed consensus decision layer calculates a timely compensation term, and finally, the pure analog adaptive drooping execution layer injects the compensation term into the physical control loop in the form of an analog signal.

[0020] This scheme effectively reduces the negative impact of communication delay and equipment parameter drift on coordinated control commands, reduces coordination errors and reactive power circulation between inverters, and improves the dynamic adjustment capability of microgrid systems under complex communication network conditions. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the implementation of the microgrid distributed data acquisition and control system considering communication delay provided by the present invention;

[0022] Figure 2 This is a Pareto front distribution data map of multi-objective optimization of multi-mode communication link quality in a microgrid distributed data acquisition and control system considering communication delay, provided by the present invention.

[0023] Figure 3 This is the nonlinear response surface plot of parameter transmission weights based on Pearson correlation coefficient in the microgrid distributed data acquisition and control system considering communication delay provided by the present invention.

[0024] Figure 4 This is a convergence curve of online identification of LC filter parameters of distributed inverter power supply in the microgrid distributed data acquisition and control system considering communication delay provided by the present invention;

[0025] Figure 5 This is a comparison diagram of the multi-node power sharing transient response under the finite-time consensus algorithm in the microgrid distributed data acquisition and control system considering communication delay provided by the present invention.

[0026] Figure 6This is an inverse Sigmoid curve of the dynamic trust weight under different sensitivity coefficients in the microgrid distributed data acquisition and control system considering communication delay provided by the present invention.

[0027] Figure 7 This is a comparison diagram of clock drift parasitic phase cross-coupling and reverse rotation decoupling phasors in the microgrid distributed data acquisition and control system considering communication delay provided by the present invention;

[0028] Figure 8 This is a flowchart illustrating the distributed data acquisition and control method for microgrids that takes into account communication delay provided by the present invention.

[0029] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0030] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0031] The following description, with reference to the accompanying drawings, describes a microgrid distributed data acquisition control system, method, and electronic device that takes into account communication delays.

[0032] Example 1:

[0033] like Figure 1 As shown, this embodiment provides a distributed data acquisition and control system for microgrids that takes into account communication latency. The system is applied to an AC microgrid containing multiple distributed inverters. Each distributed inverter is equipped with a local control unit, which, from top to bottom, includes: an adaptive high-frequency data acquisition layer, a multi-mode communication link dynamic switching layer, a dual adaptive asynchronous delay state observation layer, a distributed consensus decision layer, and a pure analog adaptive droop execution layer.

[0034] In actual industrial settings of AC microgrids, distributed inverters are typically connected in parallel to a common AC bus to provide continuous power to local loads or the main grid. The local control unit, serving as the computational and command center of the entire system, employs a top-down data and command flow processing mechanism.

[0035] The top-level adaptive high-frequency data acquisition layer is used to acquire the real-time operating status of the microgrid, extract local high-frequency electrical parameters, and dynamically adjust the data acquisition frequency and communication cycle according to the real-time operating status. After acquiring the data, the multi-mode communication link dynamic switching layer is used to monitor the quality status of multiple communication links and dynamically select the best link to transmit the local high-frequency electrical parameters.

[0036] Subsequently, the data stream enters the dual adaptive asynchronous time-delay state observation layer, which combines the local high-frequency electrical parameters with the real-time communication link status to identify the filtering parameters of the distributed inverter power supply online in order to achieve adaptive adjustment of the state observer model and reconstruct the current delay-free real power state vector of the adjacent distributed inverter power supply.

[0037] After acquiring the real data without delay, the distributed consensus decision layer is used to perform distributed consensus calculation based on the real power state vector without delay, and generate frequency compensation term and AC bus voltage amplitude compensation term.

[0038] Finally, a pure analog adaptive droop control layer is used to inject the frequency compensation term and AC bus voltage amplitude compensation term into the physical droop control loop of the distributed inverter power supply in the form of analog signals, completing the closed-loop dynamic regulation of the microgrid. This five-layer architecture design logically forms a complete closed loop, effectively avoiding the single-point failure risk brought about by traditional centralized control.

[0039] For example, the specific hardware architecture and basic signal processing flow of the adaptive high-frequency data acquisition layer are fully described here. This layer physically includes a high-precision Hall voltage sensor, a Hall current sensor, signal conditioning and filtering circuitry, and an analog-to-digital converter with multi-channel synchronous sampling capability.

[0040] To avoid serious control consequences caused by ambiguous technical parameter definitions, this document clearly distinguishes the electrical parameter sequences at different stages based on engineering specifications: Definition This is a real-time discrete current sequence directly sampled from the sensor port via an analog-to-digital converter. This sequence includes the fundamental component, higher harmonic components, and system noise. This is the fundamental current phasor extracted after digital low-pass filtering and coordinate transformation, used for actual power calculation and droop control.

[0041] The adaptive high-frequency data acquisition layer acquires discrete sequences of three-phase AC voltage and real-time current. Then, the phase angle is extracted using a phase-locked loop, and the electrical quantities in the stationary three-phase coordinate system are transformed into DC quantities in the synchronously rotating dq coordinate system through Parker transformation. This allows the extraction of local high-frequency electrical parameters containing active and reactive currents, and the fundamental phasor of the actual current is then calculated. And the corresponding active power and reactive power.

[0042] Optionally, in order to achieve effective data transmission under limited communication bandwidth, the adaptive high-frequency data acquisition layer automatically switches the data acquisition frequency and communication cycle between steady-state operation mode, load fluctuation mode and fault transient mode by real-time monitoring of local active power change rate and AC bus voltage deviation.

[0043] In actual operation scenarios, if the power grid is in a period of stable load, the system automatically enters the steady-state operation mode. At this time, the sampling frequency is reduced and the communication cycle is extended, and only the average active and reactive power with a longer integration time is transmitted to reduce the processor load. When the starting of an induction motor or the input of a large nonlinear load is detected, causing the local active power change rate or AC bus voltage deviation to exceed the first-level set threshold, the system switches to the load fluctuation mode. At this time, the data acquisition frequency is increased and the communication cycle is shortened, and the fundamental phasors of dq axis voltage and current are transmitted to ensure dynamic response capability.

[0044] When a short circuit or topology change causes the rate of change or deviation to exceed a higher set threshold in the second level, the system enters a fault transient mode, performing extremely high-frequency sampling and dense communication to ensure complete capture of fault data. Simultaneously, a parameter-weighted transmission mechanism based on the Pearson correlation coefficient is employed to dynamically adjust the transmission weights of each electrical parameter. The transmission weight update rule is as follows:

[0045] ;

[0046] in, For the updated transmission weights, As the initial weights, To adjust the factor, The Pearson correlation coefficient between the local high-frequency electrical parameters and the preset fault status label.

[0047] In the above formula, the preset fault state labels are a baseline sequence characterizing key features at the time of historical fault occurrences. The Pearson correlation coefficient is calculated. The system can quantitatively assess the correlation between a currently collected electrical parameter and potential fault characteristics.

[0048] When the correlation coefficient of a certain parameter is high, the factor can be adjusted. Amplify its initial weights This mechanism ensures that the most relevant key feature data is delivered to neighboring nodes preferentially and with higher priority in the communication queue; conversely, less relevant conventional parameters are kept at a lower weight or are transmitted later. This ensures that, even in congested communication channels, the most valuable engineering features are delivered to neighboring nodes reliably and preferentially.

[0049] like Figure 3 The figure shows the nonlinear response surface data of parameter transmission weights based on the Pearson correlation coefficient. This figure reflects the system's ability to dynamically adjust the transmission priority of underlying electrical parameters and the limitations of hardware response when facing different fault characteristic correlations in the adaptive high-frequency data acquisition layer.

[0050] In the figure, the horizontal axis at the bottom represents the Pearson correlation coefficient, in dimensionless units; the vertical axis at the bottom represents the dynamic adjustment factor, in dimensionless units; and the vertical axis represents the transmission weight update value, in dimensionless units.

[0051] The graph is presented as a three-dimensional surface that gradually rises from the lower left to the upper right. The gradient of the surface color intuitively reflects the magnitude of the values. The smooth transition from the dark blue area to the dark red area at the top represents the increase in the gradient of the transmission weight resource allocation.

[0052] In scenarios with mild load fluctuations, when the Pearson correlation coefficient is in the low range of 0.1 to 0.4 and the dynamic adjustment factor remains around 2.0, the transmission weight update value rises slowly to below 1.8, located in the blue to light blue area of ​​the surface, indicating that the system assigns a lower transmission priority to conventional parameters in order to save communication bandwidth.

[0053] When a topological change in the power grid causes a sharp increase in the correlation between a certain electrical parameter and a preset fault state label—for example, when the Pearson correlation coefficient approaches 0.8 and the dynamic adjustment factor increases to 4.0—the transmission weight update value rapidly climbs to above 4.2, entering the orange to red region of the surface. This significant increase in spatial curvature demonstrates that the parameter-weighted transmission mechanism can non-linearly amplify the transmission weight of key feature data, ensuring that the most engineering-valuable data is prioritized for delivery when communication channels are congested.

[0054] However, observing the extreme high-value regions of the response surface in the horizontal coordinate range of 0.9 to 1.0 and the vertical coordinate range of 4.5 to 5.0 reveals that the dark red surface edges exhibit a flattened truncation distortion phenomenon. After the transmission weight update value reaches 5.0, it no longer maintains the original linear or exponential growth trend, but instead shows a plateau period with slight quantitative fluctuations.

[0055] This image feature reveals a technical flaw in the current local control unit, which has computational limit protection or bandwidth allocation saturation within the digital processor. When multiple highly correlated parameters are concurrently applied, the system is limited by the capacity of the underlying registers and cannot continue to allocate higher distinguishing weights. This indicates that the dynamic weight allocation mechanism still has room for further improvement in the upper limit boundary constraints and anti-saturation algorithms under extreme communication congestion conditions.

[0056] It is important to note that in the harsh electromagnetic environment of industrial sites, a single communication standard is highly susceptible to interference and interruption. Therefore, the multi-mode communication link dynamic switching layer monitors the transmission delay of each communication link in real time. Packet loss rate Signal-to-noise ratio and available bandwidth And calculate the overall quality score based on the following formula.

[0057] ;

[0058] in, For the maximum allowable delay, To the maximum allowable packet loss rate, For maximum signal-to-noise ratio, For the maximum available bandwidth, , , , These are preset weighting coefficients, and their sum is 1.

[0059] In terms of hardware implementation, this layer integrates a ZigBee radio frequency module based on the IEEE 802.15.4 standard, a LoRa module based on spread spectrum technology, and an orthogonal frequency division multiplexing power line carrier module. During each communication control cycle, the system continuously collects four status indicators for these three physical links.

[0060] By using the normalized weighted calculation of the above formula, communication indicators with different dimensions can be transformed into a dimensionless scalar between 0 and 1. This provides an intuitive and quantitative reflection of the actual availability of each link.

[0061] When the main communication link in use is interfered with by external factors, its overall quality score is affected. When the threshold for link switching is lower than the preset threshold, a switching process is initiated to establish a backup communication link and then disconnect the current communication link. This redundant switching process requires the control unit to first activate the backup link with the highest score in the background handshake, confirm that the data is bidirectionally connected on the backup link, and then release the resources of the original damaged link, thereby avoiding the sudden and massive data loss caused by hard switching.

[0062] like Figure 2 The figure presents frontier distribution data for multi-objective optimization of multi-mode communication link quality. It illustrates the technical effectiveness and limitations of the multi-mode communication link dynamic switching layer in optimizing multiple redundant physical hardware links when the local control unit faces electromagnetic interference from the distribution network at different levels.

[0063] The horizontal axis in the graph represents transmission delay in milliseconds; the vertical axis represents packet loss rate in percentage.

[0064] The graph shows three characteristic curves of different colors distributed from bottom to top, which correspond to the non-dominated solution sets obtained by multi-objective balance calculations under three typical operating conditions.

[0065] The blue curve at the bottom represents the non-dominated solution set under mild interference conditions, corresponding to the scenario where the system is operating normally and the load is stable. At this time, the transmission is mainly carried out by the wireless radio frequency link. Its waveform transformation shows extremely smooth and low-level distribution characteristics. Within the transmission delay range of 5 to 15 milliseconds, the packet loss rate is stably suppressed to within 1.2 to 1.5 percent. This indicates that the multimode communication link dynamic switching layer can accurately lock the best link with high timeliness and high integrity for the transmission of local high-frequency electrical parameters when the channel environment is good.

[0066] The orange curve in the middle represents the non-dominated solution set under moderate interference conditions, corresponding to the scenario where the introduction of a large nonlinear load in the system causes local load fluctuations. At this time, the system mainly switches to the spread spectrum communication link with stronger anti-interference capability. Its waveform shifts to the upper right, and the curve slides within the range of 15 to 45 milliseconds of transmission delay. The corresponding packet loss rate rises to 2.5 to 3.5 percentage points. This quantitatively proves that when the external disturbance increases, the multi-mode communication link dynamic switching layer sacrifices some real-time response bandwidth to obtain information reliability in a multi-objective trade-off control decision process.

[0067] The red curve at the top represents the non-inferior solution set under severe fault transient conditions, corresponding to extreme scenarios where the system experiences short-circuit faults or severe topological changes leading to a surge in strong electromagnetic noise. At this time, the system fully activates the power line carrier module with physical cable barriers to relay transmission, and its waveform changes show obvious rightward extension and rise.

[0068] However, observing the tail region of this heavy power transient curve reveals that when the transmission delay is forced to lengthen to 120 milliseconds, the packet loss rate curve shows a certain degree of nonlinear upward sloping, approaching 5.5%. This waveform characteristic reveals the technical defects of power line carrier in channel capacity saturation and transmission congestion when encountering concentrated interference from extremely high frequency transient harmonic clusters. This indicates that there is still room for further optimization and algorithm upgrades in the delay compensation and network filtering of this scheme under heavy interference.

[0069] This multi-objective frontier distribution map, through objective data waveform comparison, fully demonstrates that the multi-mode communication link dynamic switching layer can autonomously find the optimal technical balance between the two opposing objective dimensions of transmission delay and packet loss rate in real time based on the link comprehensive score under the ever-changing industrial disturbance environment, thereby providing high-timeliness and high-reliability data transmission support for the subsequent observer and decision-making layer.

[0070] Specifically, the multi-mode communication link dynamic switching layer also includes a link failure fallback unit. When all communication links are detected to be interrupted, a local islanding operation mode is triggered, and the system disconnects from the communication network, switching to autonomous and independent control based on local historical data and the short-term predicted power state vector of the dual adaptive asynchronous delay state observer. In scenarios where extreme natural disasters or severe network attacks cause global communication paralysis, relying on communication consensus algorithms will not be able to obtain external input.

[0071] At this point, the link failure fallback unit will block the external data receiving port, extract historical power data from the last few cycles before the communication interruption in the local memory, and extrapolate to generate a short-term predicted power state vector using the prediction function of an autoregressive model or Kalman filter. The local control unit maintains basic droop characteristics based on this predicted vector, sacrificing some global current sharing accuracy to ensure the basic power supply and voltage and frequency stability of the microgrid in the absence of communication, and to prevent system collapse and shutdown.

[0072] Optionally, under the condition that the microgrid is operating normally and there is a certain communication delay, the dual adaptive asynchronous time delay state observation layer undertakes the core tasks of model correction and state reconstruction.

[0073] This layer contains an online inverter parameter identification unit. During long-term operation of the distributed inverter power supply, the inductance of its internal filter inductor will decrease due to core saturation and temperature rise, and the capacitance of the filter capacitor will also shift due to dielectric aging.

[0074] Continuing to use the factory-specified parameters for state observation will lead to severe extrapolation divergence. Therefore, the online inverter parameter identification unit uses the recursive least squares method to identify parameters, update the difference equation parameters in the reduced-order state equations of the distributed inverter power supply, and inversely calculate the parameters at the discrete time step. Actual filter inductance value and capacitance value :

[0075] ;

[0076] ;

[0077] in, The system sampling period is to , to All of these are the parameters of the difference equation obtained through identification.

[0078] During actual processor operation, the control unit collects the inverter bridge output voltage, grid current, and filter capacitor voltage to construct an input-output data matrix in the discrete-time domain. Using the recursive least squares method, the parameter vector is continuously adjusted within each sampling period to minimize the sum of squared errors between the model's predicted output and the actual sampled output. After convergence, the parameters are extracted. to , to Substituting these parameters into the formula above allows for the extraction of the actual physical parameters at the current moment purely through software, without the need for additional hardware sensors. and .

[0079] like Figure 4 The figure shows the convergence curve data of online identification of filter parameters of distributed inverter power supply. It illustrates the technical effect and steady-state limitations of the dual adaptive asynchronous time delay state observation layer in adaptively tracking the aging of external physical entities through internal algorithm models when facing the evolution of physical parameters of internal physical components due to heating or partial discharge.

[0080] The horizontal axis in the figure represents the number of iterations; the left vertical axis represents the filter inductor identification value in millihenries; and the right vertical axis represents the filter capacitor identification value in microfarads.

[0081] The graph contains two undulating curves of different colors, with the blue curve representing the filter inductor identification curve and the orange curve representing the filter capacitor identification curve.

[0082] In the initial identification phase, within the interval of 0 to 300 iterations, the blue curve exhibited a significant waveform overshoot, with the value climbing to over 3 millihenries before rapidly declining under the influence of the error decay law; the orange curve, on the other hand, showed a smooth exponential climbing waveform change.

[0083] After 800 iterations, both curves broke free from transient fluctuations and entered a relatively stable convergence range. The identified values ​​of the filter inductor and filter capacitor stabilized around 2 millihenries and 50 microfarads, respectively. This waveform convergence characteristic objectively proves that the online inverter parameter identification unit can extract the true physical parameters at the current moment through iterative least-squares algorithm without adding additional hardware sensors, thus enabling the internal model of the state observer to reflect the true dynamic characteristics of the physical entity.

[0084] However, observation of the steady-state range of 1000 to 2000 iterations reveals that the two identification curves do not remain horizontal straight lines, but are continuously accompanied by subtle high-frequency spikes and random fluctuations. In particular, there is significant steady-state quantization noise between the filter capacitor identification values ​​of 48 μF and 52 μF. This image feature exposes the fact that the current identification algorithm lacks a sufficient low-pass smoothing filtering mechanism in the steady-state stage when facing strong electromagnetic interference and analog-to-digital converter sampling noise in industrial environments. This indicates that the online identification parameter extraction scheme still has room for further upgrades and optimizations in terms of steady-state noise robustness and the dynamic adjustment accuracy of the algorithm's forgetting factor.

[0085] For example, when the calculated actual filter inductance value or capacitance value When the deviation from the preset nominal inductance or capacitance value exceeds a preset proportion, the dual adaptive asynchronous delay state observation layer automatically updates the state matrix of the built-in state observer. and input matrix This enables dual adaptive adjustment of the parameters of the state observer model.

[0086] In modern control theory, the core of a state observer lies in its system state equations. In the microgrid inverter model, the state matrix... The elements are directly derived from the actual filter inductance value. Compared with the actual filter capacitor value The reciprocal of the matrix and its combination constitute the matrix. When the parameters experience objective physical drift and the deviation reaches the trigger condition, the system recalculates online and overwrites the original matrix. This process constitutes the first layer of adaptive adjustment, namely, internal model adaptation based on the evolution of physical object parameters, which ensures that the observer has the ability to reflect the true dynamic characteristics of the physical entity.

[0087] It is also important to note that, assuming the internal model is accurate, the time lag issue of external data needs to be addressed. The dual adaptive asynchronous delay state observation layer simultaneously adjusts the forward inference step size based on the real-time transmission delay and real-time packet loss rate of the communication link, and reconstructs the adjacent distributed inverter power supplies through forward inference calculations. At the present moment The delay-free true power state vector :

[0088] ;

[0089] ;

[0090] in, For integration variables, For the adjacent distributed inverter power supply At a historical moment Power data transmitted, From a historical moment up to the current moment The change in state This is the input vector collected locally.

[0091] In this computational phase, the system constitutes a second layer of adaptive adjustment, namely, time-compensated adaptive adjustment based on the evolution of network environment parameters. When the local control unit is at the current moment... Upon receiving a data packet with a timestamp, the system parses out the historical moment when the data packet was packaged and sent by a neighboring node. The difference between these two points in time is the communication delay.

[0092] The system utilizes the updated state matrix Constructing a matrix exponential function As a state transition matrix, and in the time interval The integral is calculated using definite integrals. In engineering implementation, this integral is usually solved numerically using higher-order Runge-Kutta methods or Taylor series expansions.

[0093] Through this integral deduction, the state change caused by the time lag during this delay period can be calculated. Compare it with outdated historical data. By superimposing, the actual operating state of adjacent nodes at this moment can be calculated and reconstructed in advance within the local processor, thereby effectively compensating for the phase shift misalignment caused by communication delay on the time axis.

[0094] Specifically, the distributed consensus decision layer is used to perform distributed consensus calculations based on the delay-free true power state vector, generating frequency compensation terms and AC bus voltage amplitude compensation terms. The distributed consensus decision layer uses a finite-time consensus algorithm to calculate the frequency compensation terms and AC bus voltage amplitude compensation terms, and its control law function has the following form:

[0095] ;

[0096] in, For the first Compensation control variables for a distributed inverter power supply. To control the gain, For the preset index, For the first A set of neighboring distributed inverters that are communicating normally. Adjacent distributed inverter power supply The delay-free true power state vector, This represents the local, latency-free true power state vector. In traditional asymptotic consensus algorithms, the convergence speed of the state difference decreases exponentially as the difference decreases, resulting in a long time required for the system to reach steady-state equilibrium. This embodiment introduces a nonlinear sign function. With fractional exponent (Usually the value is between 0 and 1), which constructs a finite-time consistency control law.

[0097] The aforementioned control law function performs element-wise algebraic operations independently on each component within the state vector (i.e., the active and reactive power sequences). When the local state vector... A certain component and the reconstructed neighbor state vector When the corresponding components deviate, the nonlinear control law can maintain sufficient control strength even with small deviations, forcing all distributed inverters in the entire microgrid system to achieve power sharing and voltage-frequency synchronization within a finite time frame that has been rigorously proven mathematically. The calculated output... These are mapped to frequency compensation terms and AC bus voltage amplitude compensation terms, respectively, to correct the set reference values ​​of the underlying controller.

[0098] like Figure 5 The figure shows comparative data of the transient response of power sharing among multiple nodes under the finite-time consensus algorithm. It illustrates the differences in convergence speed and the limitations of dynamic adjustment when the distributed consensus decision layer calls different control laws to perform distributed consensus calculation after receiving the reconstructed zero-delay real power state vector.

[0099] The horizontal axis in the figure represents the simulation time in seconds; the vertical axis represents the output active power in kilowatts.

[0100] The graph contains six curves representing the output power waveform changes of different distributed inverters. The three gray dashed lines represent the node transient response using the traditional asymptotic consensus algorithm, while the other three solid lines (blue, orange, and yellow) represent the node transient response using the finite-time consensus algorithm of this scheme.

[0101] Within the simulation time interval of 0 to 5 seconds, the system's set steady-state power distribution target value was 25 kilowatts. It can be observed that the three gray dashed lines exhibit a slow, exponentially decaying waveform transformation. The convergence speed of the state difference decreases as the difference decreases, barely approaching the steady-state target value after about 4 seconds. This directly demonstrates the technical shortcomings of traditional asymptotic consensus algorithms, which require a long time for the system to reach absolute equilibrium.

[0102] In contrast, the three solid colored lines in this scheme, under the action of the nonlinear control law, can maintain sufficient control strength even when the deviation is small, forcing all nodes in the system to quickly converge to the 25 kW equalization baseline in about 1.5 seconds, which greatly shortens the transition time required for all distributed inverters in the entire microgrid system to achieve power equalization and voltage frequency synchronization.

[0103] However, observing the solid colored line in the steady-state range of 1.5 to 5 seconds reveals that the waveform does not remain perfectly flat. Instead, it exhibits small-amplitude, continuous high-frequency flutter around the 25 kW baseline. This graphical characteristic objectively exposes the technical flaw in this scheme: while the introduction of a nonlinear sign function and fractional exponent accelerates transient convergence, it is prone to causing minor system oscillations near the steady-state point due to the discrete execution dead zone of the underlying digital controller. This indicates that the control strategy still has room for further optimization and upgrading in terms of steady-state smoothness suppression and the adaptive decay mechanism of nonlinear control gain.

[0104] Optionally, the distributed consensus decision layer is also used to actively detect abnormal adjacent distributed inverters: by periodically sending heartbeat signals to adjacent distributed inverters, paging detection is performed when no heartbeat response is received, and when the target adjacent distributed inverter is determined to be abnormal, it is automatically removed from the communication topology, so that the remaining distributed inverters can re-form an undirected and connected sparse communication topology; when the abnormal adjacent distributed inverter is detected to have recovered, it is automatically reconnected to the control loop through a handshake protocol.

[0105] In practical sparse communication networks, physical downtime of distributed inverter power supplies or long-term network outages are objective risks. The heartbeat signal is a short byte message with the source device identifier and timestamp.

[0106] If the local control unit fails to detect the heartbeat of a neighboring distributed inverter for several consecutive communication cycles, it initiates a high-frequency directional paging message. If no echo is received, it can be determined that a substantial anomaly has occurred in the neighboring distributed inverter.

[0107] At this point, the local control unit resets the connection weight corresponding to the power supply device to zero in the adjacency matrix, and the algorithm's summation set... It will dynamically shrink. Graph theory analysis shows that as long as the reconstructed remaining communication topology still maintains a connected graph structure containing spanning trees, the finite-time consensus algorithm can still guarantee the global stability of the remaining inverter power system. The handshake protocol during recovery ensures that the states of reconnected adjacent distributed inverter power sources can transition smoothly, preventing abrupt topology changes from causing grid oscillations.

[0108] For example, the instruction is issued from the digital layer, and the pure analog adaptive droop execution layer is used to inject the frequency compensation term and the AC bus voltage amplitude compensation term into the physical droop control loop of the distributed inverter in the form of analog signals to complete the closed-loop dynamic adjustment of the microgrid.

[0109] Traditional microgrid droop control is usually implemented in code within a digital signal processor, but digital control inevitably introduces sample-and-hold delay, computation execution delay, and pulse width modulation update delay.

[0110] To achieve the ultimate dynamic response bandwidth, the frequency-active droop control and voltage-reactive droop control in the pure analog adaptive droop execution layer are both implemented using pure analog hardware circuits. The pure analog adaptive droop execution layer uses digitally calculated frequency compensation terms. AC bus voltage amplitude compensation item Direct physical injection droop control loop:

[0111] ;

[0112] ;

[0113] in, and These are the adjusted frequency and voltage commands, respectively. and These are the rated frequency and rated voltage, respectively. and This refers to the actual active and reactive power. and Rated active and reactive power, and This is the droop coefficient.

[0114] In terms of hardware structure, a high-precision digital-to-analog converter (DAC) converts the digital domain... and It is converted into a DC analog voltage signal. This analog signal enters an adder circuit constructed from a precision operational amplifier. Simultaneously, the actual output power... and The signal is obtained by directly multiplying and filtering the sensor signal using an analog multiplier. The operational amplifier network is scaled and polarity-reversed according to the coefficient relationships in the above formula, incorporating the analog active power deviation, rated frequency reference, and voltage-form frequency compensation terms. The analog voltages are directly added at the analog adder nodes to generate a continuous frequency command analog voltage. .

[0115] Similarly, generate voltage commands to simulate voltage. These two purely analog instructions are directly fed into the subsequent analog pulse width modulation generator (such as an analog PWM chip based on triangular wave comparison) to generate switching signals that drive the insulated gate bipolar transistor (IGBT) in real time. Since the response time of the analog circuit is only limited by the slew rate and parasitic capacitance of the amplifier, its adjustment delay is typically on the order of nanoseconds, achieving drooping execution with extremely low latency.

[0116] Based on existing technology analysis, in traditional microgrid secondary control and collaborative control systems, control algorithms often assume that the communication network is ideal and has no delay, and that the inverter's own physical parameters are constant. When the system is deployed in a physical environment with a large actual span and strong electromagnetic interference, simple software control is prone to outputting incorrect compensation commands due to communication data lag, resulting in huge circulating current losses between inverters, and in severe cases, causing frequency collapse of the entire power grid.

[0117] This embodiment breaks through this limitation with a five-layer progressive hardware and software integration architecture: from bottom to top, the pure simulation execution layer ensures the transient real-time performance of the underlying response; the adaptive high-frequency acquisition and multi-mode communication switching layer improves the arrival rate and timeliness of information from the data source and transmission path.

[0118] At the core algorithm level, the dual adaptive asynchronous delay state observation layer simultaneously identifies and compensates for internal physical parameter aging and external network communication delays online, restoring distorted and lagging data to a real state with current timeliness; finally, in conjunction with the finite-time consistency algorithm, robust collaboration is achieved in non-ideal real-world environments.

[0119] The overall technical solution forms an anti-interference and adaptive closed-loop control link from hardware perception, network transmission, mathematical reconstruction, logical decision-making to underlying physical execution. This significantly reduces the control deviation caused by communication delay and hardware parameter drift, and effectively improves the power supply quality and operational stability of distributed microgrid systems in complex engineering application scenarios.

[0120] Example 2:

[0121] In conventional microgrid collaborative control architectures, the convergence and stability of consensus algorithms highly depend on the authenticity and reliability of the data exchanged between nodes. However, when distributed inverters are exposed to outdoor or industrial sites lacking strict physical isolation, their communication links are vulnerable to malicious intrusion. Attackers may hijack the communication modules of adjacent nodes to continuously inject carefully disguised fake data into the system network, a technique known as fake data injection attack.

[0122] Meanwhile, harsh operating environments may also cause sudden hard failures in the voltage or current sensors of adjacent nodes, resulting in the transmission of severely distorted erroneous state data. If the local control unit directly accepts this data, which contains deceptive or abrupt errors, received from the communication network, the dual adaptive asynchronous time-delay state observation layer will treat it as real historical data for forward extrapolation. This will lead to the distributed consensus decision layer calculating frequency compensation terms and AC bus voltage amplitude compensation terms with severely deviated directions and amplitudes.

[0123] When this error compensation command is executed by the pure analog adaptive drooping execution layer, it will directly cause serious safety accidents such as the output voltage of the underlying inverter exceeding the limit, system frequency collapse, or large-area reactive power circulation.

[0124] Therefore, in the distributed data acquisition and control system for microgrids that takes into account communication delay provided in this embodiment, the distributed consensus decision layer is also equipped with a data credibility assessment module based on spatiotemporal physical coupling; the data credibility assessment module based on spatiotemporal physical coupling is used to perform dynamic trust assessment and isolation filtering on the delay-free real power state vector before performing distributed consensus calculation.

[0125] This module is located between state observation and consistency decision-making in the logical architecture, serving as the core security verification module before data flows to the physical control loop.

[0126] Optionally, the core verification logic of the data credibility assessment module based on spatiotemporal physical coupling does not rely on traditional cryptographic encryption protocols at the communication network level, but is based on the immutable electrical connectivity and physical laws within the AC microgrid. Specific execution steps include:

[0127] Based on the preset global admittance matrix and the real-time electrical power collected by the local AC bus, Kirchhoff physical space constraint equations are constructed to calculate the adjacent distributed inverter power sources. The physical reference state vector at the current moment .

[0128] In practical microgrid engineering applications, although the distributed inverters are spatially dispersed, they are all connected to a common AC bus or regional node via distribution lines with specific impedance characteristics in terms of electrical topology. A pre-defined global admittance matrix objectively reflects the line conductance, susceptance, and spatial topology information of the entire microgrid system at stable operating frequencies. This parameter set is determined by the inherent properties of the grid's physical devices and is typically stored in the non-volatile memory of the local control unit, making it difficult for attackers to alter these physical entity attributes by tampering with communication network data.

[0129] The local control unit uses high-precision local sensors to acquire the three-phase voltage of the local AC bus and the current phasor injected by the local inverter in real time, safely, and without delay. According to Kirchhoff's current law and voltage law, the voltage and injected current at any node in the power grid are strictly constrained by the global admittance matrix.

[0130] Therefore, the local control unit can utilize its own reliable local data acquisition and prior admittance topology parameters to deduce, from a purely physical spatial constraint dimension, the active and reactive power states that its neighboring nodes should output at this moment by solving nonlinear power flow equations or linearized DC power flow approximation equations. This deduction process tightly couples the communication network space with the physical electrical space, generating the physical reference state vector. As an objective physical reference benchmark, it provides an independent and objective standard for subsequently judging the authenticity of communication data, free from interference by network attacks.

[0131] For example, after obtaining the mirror reference data calculated based on purely physical laws, the system enters the numerical cross-validation stage. The data reliability assessment module based on spatiotemporal physical coupling needs to calculate the physical reference state vector. The adjacent distributed inverter power supply reconstructed by the dual adaptive asynchronous time-delay state observation layer Delay-free true power state vector Spatiotemporal deviation residuals between Its defining formula is:

[0132] ;

[0133] In the formula: It represents the state parameters of the information flow dimension, namely the power states of adjacent nodes obtained from the network communication link and after time delay compensation by the state observer; This represents the state parameters of the physical flow dimension, that is, the power state independently calculated by the local control unit using the grid spatial topology and Kirchhoff's laws. The L2 norm operator for vectors is used to calculate geometric distance in Euclidean space. The calculated spatiotemporal bias residual is a non-negative scalar value.

[0134] When a microgrid is in a normal, attack-free, and fault-free operating state, due to line parameter measurement errors, the inherent accuracy limits of sensors, and the objective existence of environmental noise, and There will inevitably be slight numerical differences between them, and the spatiotemporal deviation residuals at this time... It will fluctuate within a low background noise range.

[0135] However, if a neighboring node's sensor experiences a short circuit, open circuit, or other hard fault, or if its communication control module is maliciously tampered with and begins broadcasting false power data that violates physical laws, the data will be reconstructed by the observer. This will deviate significantly from its actual physical state. At this point, because the actual voltage change of the local AC bus will immediately and accurately reflect the true injected power change of that physical node, the locally calculated... It can maintain tracking of the real physical world. This disconnect between spurious data in the information domain and the real state in the physical domain will cause the distance in the above equation to increase rapidly, resulting in a significant increase in the spatiotemporal deviation residual. A surge occurred.

[0136] Through this L2 distance calculation, the system successfully quantifies hidden network data tampering behavior or hardware failures into an intuitive and monitorable mathematical feature.

[0137] Specifically, in order to apply the quantified residual characteristics to the subsequent closed-loop control strategy, and to avoid system misjudgment and control oscillations caused by single-measurement noise, the data reliability assessment module needs to construct a smooth filtering mechanism with adaptive damping characteristics. Based on the spatiotemporal deviation residual... Construct a dynamic trust decay function to calculate the local control unit's trust level for each adjacent distributed inverter power supply in real time. Dynamic trust weight :

[0138]

[0139] In the formula: To characterize the local control unit at the current moment The scalar weights for the degree of trust in the data transmitted by neighboring nodes are strictly constrained by a nonlinear function to a closed interval between 0 and 1. It is a natural exponential function; This is the preset confidence sensitivity adjustment coefficient; This is a safety deviation threshold set based on historical normal steady-state data. Mathematically, this decay function exhibits typical nonlinear inverse Sigmoid curve characteristics, a curve shape that highly aligns with the fault-tolerance requirements in industrial control.

[0140] In the actual commissioning and operation of microgrids, engineers first collect long-term residual data of the system under healthy conditions and then use statistical calculations to determine the background noise distribution and safety deviation threshold. It is typically set to a number of standard deviations (e.g., three standard deviations) of the historical normal steady-state background noise mean. This is used when calculating the spatiotemporal deviation residuals in real time. Much smaller than the safety deviation threshold When the exponential term approaches 0, the dynamic trust weight... It remains in a high-level region approaching 1. This indicates that the local control system fully trusts the data from the neighboring node, considering data fluctuations to be within a reasonable range of physical errors. Confidence sensitivity adjustment coefficient. This determines the slope or severity of the decrease in confidence when the residual crosses the safety deviation threshold.

[0141] In application scenarios where the power grid faces strong electromagnetic interference and normal data fluctuates significantly. The value is usually set relatively small to make the attenuation curve flatter, thereby enhancing the system's tolerance to occasional noise; however, in aerospace or precision manufacturing microgrids where control precision and safety requirements are extremely high, The value will be increased accordingly, so that once the residual exceeds the safety threshold, the trust weight will decay rapidly. This trust allocation mechanism based on nonlinear exponential decay effectively avoids the control system splitting jitter caused by high-frequency fluctuations of parameters near the threshold, compared to a simple step function, thus ensuring the dynamic smoothness of the closed-loop system during the transition process.

[0142] like Figure 6 The figure shows the inverse distribution curves of dynamic trust weights under different sensitivity coefficients. This figure reflects the technical effect of the data credibility assessment module based on spatiotemporal physical coupling in performing dynamic trust assessment and isolation filtering on the real power state vectors of adjacent nodes without delay before performing distributed consensus calculation, as well as the limitations of the underlying digital computation.

[0143] The horizontal axis in the figure represents the spatiotemporal bias residual, with units of dimensionless; the vertical axis represents the dynamic trust weight, with units of dimensionless.

[0144] The graph contains three curves that exhibit a non-linear exponential decay trend from left to right: the blue curve represents the low sensitivity decay curve, the orange curve represents the medium sensitivity decay curve, and the yellow curve represents the high sensitivity decay curve.

[0145] When the spatiotemporal deviation residual of the horizontal axis is less than the safety deviation threshold of 2.0 set based on historical normal steady-state data, the dynamic trust weights of the three curves are all maintained in a high-level region approaching 1.0, indicating that the local control system fully trusts the data of adjacent nodes and considers the current data fluctuation to be within a reasonable range of physical errors.

[0146] When the residual crosses the safety deviation threshold of 2.0 to the right, the blue curve shows a relatively gentle downward slope, reflecting the system's high tolerance for occasional noise when facing strong electromagnetic interference; while the yellow curve shows a precipitous drop in waveform, reflecting the system's ability to quickly and flexibly suppress abnormal data in application scenarios with extremely high safety requirements. This nonlinear allocation mechanism effectively avoids the splitting jitter of the control system and ensures the dynamic smoothness during the transition process.

[0147] When the horizontal axis reaches the extreme value alarm threshold of 4.0, the system directly performs hard isolation at the logic control level, forcibly assigning the dynamic trust weights corresponding to the three curves to 0.

[0148] However, observing the waveform transformation of the yellow curve in the steeply descending section reveals that the curve does not remain continuously smooth, but rather exhibits a slight step-like quantization jitter. Furthermore, when the residual just crosses the extreme value alarm threshold of 4.0, the weight waveform does not instantly return to zero vertically, but retains a brief non-zero value phase of approximately 0.05. This image feature objectively exposes the truncation error caused by the fixed-point arithmetic word length limitation of the digital processor inside the local control unit during nonlinear operations, as well as the hard cutoff instruction delay defect caused by the sampling execution cycle of the control chip. This indicates that the trust evaluation algorithm still has room for further optimization and upgrading in terms of the accuracy of ultra-high frequency transient network disconnection response and the resistance to saturation overflow in the underlying hardware code.

[0149] It is also important to note the dynamic trust weight. The calculation process only completes the security posture assessment; its true defensive effectiveness needs to be reflected in the execution phase of the distributed consensus algorithm. When the distributed consensus decision layer solves the frequency compensation term and the AC bus voltage amplitude compensation term, the dynamic trust weight is used... This adaptive decay coefficient is multiplied into the corresponding adjacent node state difference term.

[0150] In a typical distributed current sharing or frequency recovery algorithm, the local control variable usually depends on the sum of the differences between the local state and the states of each neighboring node. Introducing this summation term means that when the data of a neighboring node is evaluated by the system as having decreased credibility, i.e., the residual becomes larger, it leads to a decrease in the weight. As the data decays from 1 to 0, the proportion of the data difference at that node in the total control compensation calculation is proportionally reduced. This mechanism is known in engineering as "soft isolation" or "flexible suppression." It allows the system to reduce its control power instead of immediately taking extreme measures to disconnect the network when it suspects data problems, thereby limiting the damage caused by potential attacks while ensuring a certain level of coordination.

[0151] However, to prevent extremely malicious tampering or severe hardware failures, the system must possess reliable boundary protection capabilities. Therefore, the control strategy further specifies that when the spatiotemporal deviation residual... When the value exceeds the preset extreme value alarm threshold, the data reliability assessment module will assign the dynamic trust weight to the corresponding adjacent distributed inverter power supply. The value is forcibly assigned to zero, and by definition, the extreme value alarm threshold is required to be greater than the safety deviation threshold. .

[0152] The extreme value alarm threshold represents the safety limit that the microgrid physical system can withstand. When the residual exceeds this absolute high threshold, it indicates that the corresponding adjacent node has completely lost the authenticity of the data or has experienced an irreversible serious fault. At this time, the system bypasses the smooth decay region of the Sigmoid curve and directly performs "hard isolation" at the logic control level. That is, by forcibly setting the weight to zero, any influence of the node's information on the local control loop is completely cut off at the data input end of the consensus algorithm. After being hard isolated, the local inverter will no longer respond to any synchronization or voltage equalization requests from the faulty node, effectively preventing the cascading propagation of erroneous commands.

[0153] Currently, most defenses against microgrid data injection attacks focus on information technology, employing sophisticated asymmetric encryption, digital signatures, and blockchain ledgers. These traditional security measures not only significantly increase the computational burden on the microgrid's underlying control chips, leading to increased latency, but also become less effective once attackers gain physical access to the nodes and control the legitimate key system. These network-layer encryption methods are insufficient to prevent malicious data tampering transmitted through legitimate channels.

[0154] The data credibility assessment scheme based on spatiotemporal physical coupling proposed in this embodiment expands the single dimension of information security protection by introducing the inviolable physical Kirchhoff constraints in AC power grid operation as a cross-validation benchmark. It compares the spatiotemporal residuals of the information flow state derived by the state observer with the physical flow state derived based on the admittance matrix, and combines an adaptive Sigmoid nonlinear trust decay mechanism and an extreme value hard isolation strategy.

[0155] This integrated solution requires no additional hardware sensors. Through multi-dimensional data fusion calculations at the logic level, it successfully endows the microgrid's underlying controller with the ability to proactively identify and filter maliciously tampered (or false) data in harsh network environments. Its practical technical effect lies in its ability to effectively prevent the spread of erroneous information across nodes when the microgrid encounters high-intensity covert network attacks or large-scale sensor failures. It ensures the correctness of local AC bus voltage amplitude compensation and frequency regulation, thereby effectively maintaining the control robustness and power supply security of the distributed microgrid system under complex industrial application conditions.

[0156] Example 3:

[0157] In practical applications of microgrids deployed over long distances or wide areas, each distributed inverter is independently controlled by its local digital signal processor (DSP) or field-programmable gate array (FPGA). These digital control chips all rely on internal independent quartz crystal oscillators to provide the system operating clock and sampling reference.

[0158] However, due to manufacturing process tolerances, drastic temperature fluctuations in industrial environments, and long-term aging of crystals, physical hardware clock drift inevitably occurs in quartz crystal oscillators at different nodes. In the collaborative control system of AC microgrids, the phasor calculation of AC voltage and current is highly dependent on the synchronous rotating coordinate system (dq coordinate system). The rotation angle of this coordinate system is obtained by time integration of the rated angular frequency of the microgrid based on the local clock.

[0159] When the underlying hardware clocks of different distributed nodes lose synchronization, this tiny time deviation is amplified under high-frequency sampling, and thus mapped as a parasitic phase deviation in the dq coordinate system. Without correction, the receiving node will misinterpret the coordinate system misalignment caused by clock drift in adjacent nodes as a real physical imbalance of active or reactive power in the power grid. This results in a compensation command with severely erroneous direction and amplitude, causing reactive circulating current losses between distributed inverters and reducing the operational stability of the microgrid.

[0160] To address the coordinate system out-of-synchronization problem caused by the aforementioned underlying microelectronic characteristics, the distributed clock drift phase compensation module is used to eliminate the fundamental phasor phase deviation of the dq coordinate system of adjacent nodes caused by the asynchronous sampling clocks at the underlying level.

[0161] Specifically, the execution process of the distributed clock drift phase compensation module first relies on the accurate measurement of time deviation between distributed nodes. Without adding external dedicated synchronization hardware lines, the system extracts the bidirectional data packet interaction mechanism from the local node to the adjacent distributed inverter power supply based on the dynamic switching layer of the multi-mode communication link. Four network timestamps during the sending and receiving of data frames: local sending time Reception time of adjacent nodes Adjacent node reply sending time and local reception time This allows for the calculation of the relationship between the local node and the adjacent distributed inverter power supply. Relative clock offset between Its solution model is as follows:

[0162] ;

[0163] In the above timestamp extraction and calculation process, to avoid random software interrupt delays (i.e., software jitter) introduced by the microcontroller operating system's task scheduling mechanism, four network timestamps... , , , The capture is all done in the Media Access Control (MAC) layer of the communication interface through a dedicated hardware timer unit.

[0164] The moment the local node's communication controller detects that the start frame delimiter of a synchronization data frame has been sent to the physical medium, the hardware timer immediately latches the current local clock count value and records it as the local transmission time. When the data frame reaches an adjacent distributed inverter via a space link or power line. At the moment the MAC layer hardware of the adjacent node detects the frame delimiter, it latches its local clock count value and records it as the adjacent node's reception time. .

[0165] Subsequently, after processing, the neighboring nodes generate a reply data frame with timestamp information, and record the reply sending time of the neighboring nodes at the moment it is sent to the physical medium. Finally, the local node records the local reception time the instant it receives the reply data frame. .

[0166] The derivation of the above formula is based on the engineering assumption that the propagation delay of the round-trip communication link is symmetrical. Variables This includes the one-way propagation delay and the actual clock offset between the two nodes; while the variable This includes the return propagation delay and the actual clock offset of opposite polarity. By subtracting the two equations and dividing by a constant 2, the round-trip propagation delay of the communication link is mathematically canceled out, thus accurately separating and calculating the relative clock offset at the nanosecond to microsecond level at the pure software protocol level. This value objectively reflects the relationship between adjacent nodes within the current measurement period. The underlying hardware clock compared to the local node Whether the reference clock is ahead or behind, and the specific time difference range.

[0167] For example, after obtaining the clock offset parameter in the time domain, the system needs to convert it to the phasor domain of the AC power system in order to intervene in electrical characteristics. The distributed clock drift phase compensation module will calculate the relative clock offset. This is mapped to the parasitic phase shift angle at the fundamental angular frequency of the AC wave. The definition of its mapping operation is:

[0168] ;

[0169] in, This is the reference rated frequency for the AC bus of the microgrid. In actual microgrid operation scenarios, the reference rated frequency of the AC bus... The frequency is typically set to the power grid frequency to establish the system's standard time period. For example, at a frequency of 50 Hz, an AC signal takes 20 milliseconds to complete one full electrical angular cycle (i.e., twice the radius of pi). This means that any tiny time shift on the time axis will be projected proportionally as a rotation angle in the spatial coordinate system according to the angular frequency of the power grid.

[0170] Since the state acquisition of adjacent nodes is performed under their own local clock reference, when the node calculates the projection of its output current or voltage in the dq rotating coordinate system, the rotation transformation angle it uses is different from that of the local node. The rotational transformation angle has already produced a physical drift.

[0171] Without mapping, the received active and reactive data will carry hidden coupling errors. For example, the active component, which was originally distributed purely on the d-axis, will be partially projected onto the q-axis due to an incorrect rotation of the coordinate system, forming a false reactive component; similarly, the reactive component on the q-axis will also be partially projected onto the d-axis. This phenomenon is called parasitic coordinate system cross-coupling in control engineering.

[0172] Using the above mapping formula, the system abstracts the hardware time error. Multiply by angular frequency coefficient The geometric rotation angle corresponding to this parasitic cross-coupling phenomenon in the phasor domain, i.e., the parasitic phase shift angle, was precisely quantified. This perspective quantifies the severity of the "skew" in the data transmitted by neighboring nodes from a local viewpoint due to clock inaccuracies, providing an accurate correction benchmark for the subsequent construction of a mathematical reverse compensation mechanism.

[0173] Optionally, after completing the quantization calculation of the parasitic phase offset angle, the system needs to perform a mathematical-means physical restoration of the received distorted data. Before the dual adaptive asynchronous delay state observation layer reconstructs the state vector, the parasitic phase offset angle is used. Constructing a two-dimensional coordinate rotation transformation matrix The received initial phasor states of adjacent nodes are compensated by inverse rotation. The constructed two-dimensional coordinate rotation transformation matrix and its application form are as follows:

[0174] ;

[0175] From the perspective of linear algebra and motor control theory, the dq rotating coordinate system constitutes an orthogonal two-dimensional planar vector space, where the d-axis typically represents the active power component and the q-axis represents the reactive power component. (Two-dimensional coordinate rotation transformation matrix) It is a standard direction cosine matrix.

[0176] Since the data sent from adjacent nodes carries a positive parasitic phase offset, in order to bring it back to the same time and spatial phase reference of the local control unit, a geometric rotation operation with equal angular amplitude but opposite direction must be performed on it.

[0177] Cosine terms in matrix elements Used to extract the self-projection components of the phasor on the original coordinate axes, while the sine term... Its corresponding term with the negative sign This is used to eliminate cross-coupling components caused by coordinate axis tilt. Through this matrix construction, the system establishes a virtual phase corrector at the digital logic level, which can accurately rotate any two-dimensional input vector with a known phase deviation towards the target orientation by a given angle while keeping its magnitude unchanged.

[0178] This matrix operation module, as an independent operator within the microgrid's underlying control chip, is periodically called by a high-priority timer to cope with the dynamic drift of the quartz crystal oscillator frequency caused by changes in ambient temperature, ensuring that the two-dimensional coordinate rotation transformation matrix can be adaptively updated as temperature and time change.

[0179] It is also important to note that once the 2D coordinate rotation transformation matrix is ​​constructed, it must be applied at the correct data flow nodes. The system performs vector multiplication to reconstruct the phasor state after clock synchronization correction. Its compensation equation is:

[0180] ;

[0181] in, For the uncompensated adjacent distributed inverter power supply At a historical moment The original power state phasor is transmitted, at which point the clock-synchronized corrected phasor state is... The logical data will be used to replace the equivalent in Embodiment 1 It serves as the sole reference input to the dual adaptive asynchronous time delay state observation layer for forward time extrapolation calculation.

[0182] In the data processing pipeline, the uncompensated raw power state phasor This is raw data just extracted from the dynamic switching layer of the multi-mode communication link. Although it contains the active and reactive power status information of the node, its reference frame is distorted. The system will use a two-dimensional coordinate rotation transformation matrix. Multiply the original quantity phase in the form of a one-dimensional column vector on the right.

[0183] In a physical sense, this multiplication operation achieves the reverse decoupling of parasitic phase deviations in the two-dimensional coordinate system, forcibly mapping the power phasors carrying distortion errors sent by adjacent nodes back to the standard synchronization coordinate system of the local control unit.

[0184] like Figure 7 The figure shows the comparison data of clock drift parasitic phase cross-coupling and reverse rotation decoupling phasors. This figure illustrates the technical effect and hardware limitations of the distributed clock drift phase compensation module in eliminating the phase deviation of the fundamental phasor of the synchronous rotating coordinate system of adjacent nodes when the underlying sampling clock is not synchronized.

[0185] The horizontal axis in the figure represents the active power component, with the unit being kilowatts; the vertical axis represents the reactive power component, with the unit being kilovars.

[0186] The diagram contains three directional vector lines of different colors, representing the power phasor states in different reference frames.

[0187] The gray dashed vector located in the first quadrant with a small tilt angle represents the true power phasor reference in the standard synchronous coordinate system recognized by the local control unit, with an active power component of 30 kW and a reactive power component of 10 kvar.

[0188] The blue solid vector with a large tilt angle located above the gray dashed line represents the uncompensated raw received distortion quantity. Due to the loss of synchronization of the underlying hardware clocks of adjacent nodes, the time deviation is mapped as a parasitic phase offset angle at the AC fundamental angular frequency, causing the active power component of this blue vector to be erroneously reduced to about 23 kW, while the reactive power component falsely rises to about 22 kV. This parasitic cross-coupling phenomenon will cause reactive circulating current losses between microgrid inverters.

[0189] The orange solid line vector located near the gray dashed line and with an approximate tilt angle represents the clock synchronization correction phasor restored after geometric rotation operation. The system uses the calculated relative clock offset to construct a two-dimensional coordinate rotation transformation matrix to perform reverse rotation compensation on the blue distortion vector, so that it rotates towards the target orientation while maintaining its own magnitude, significantly eliminating the cross-coupling components caused by the tilt of the coordinate axis.

[0190] However, observing the endpoints of the orange solid vector and the gray dashed vector reveals that the orange vector does not coincide with the gray reference vector; a slight spatial deviation of approximately 1 kW and 2 kvar remains between them. This image feature objectively exposes that when performing purely software-level analytical geometry and matrix algebra operations, this scheme is limited by the truncation error of the underlying digital signal processor's floating-point operations and the inherent resolution limitations of the network timestamp capture timer. This prevents the scheme from eliminating the clock drift correction error at the microelectronic hardware level, indicating that the coordinate system reverse decoupling algorithm still has room for further engineering optimization in terms of extreme microsecond-level time synchronization accuracy and high-frequency phase compensation resolution.

[0191] Corrected phasor states It has eliminated all false phase interference caused by the asynchronous underlying hardware clock, and the power component it reflects truly represents the electrical output state of the physical node at that time.

[0192] Subsequently, the system no longer uses the original data, but only inputs this set of high-quality phasor states, which have been corrected by clock synchronization, into the subsequent dual adaptive asynchronous delay state observation layer. After receiving this clock-synchronized input, the state observation layer performs forward time extrapolation calculations specifically for the delay caused by the communication transmission process.

[0193] Through this logical connection between modules, the system architecture effectively decouples and independently compensates for the phase deviation caused by hardware clock errors and the time lag caused by communication network congestion, avoiding the risk of multiple errors intertwining and causing the control algorithm to diverge.

[0194] Based on the analysis of existing technologies, the current technical solutions for handling the synchronization problem of large-scale or distributed microgrids typically rely on installing a GPS or BeiDou satellite navigation system timing receiver in each inverter control box to obtain a unified pulse-per-second (PPS) signal.

[0195] However, this approach increases hardware costs and faces the risk of satellite signal loss in underground substations, tunnels, or densely populated urban canyons with high-rise buildings. Another existing technology uses dedicated fiber optic or Ethernet synchronization cables based on the IEEE 1588 protocol, but this goes against the distributed nature of microgrid systems, which are "plug and play" and "sparse communication," and greatly increases the complexity of power grid construction and maintenance.

[0196] In contrast, the distributed clock drift phase compensation scheme proposed in this embodiment does not add any microsecond-level high-frequency clock lines or external satellite antennas. The overall technical solution captures the underlying MAC layer timestamp by reusing the basic interactive messages of existing multimode communication networks, and completes the correction of clock drift at the microelectronic hardware level in the AC phasor domain using purely software-level analytical geometry and matrix algebra operations.

[0197] This scheme transforms time synchronization from a physical hardware constraint into algorithmic parameter compensation, enabling the local control unit to accurately extract the true active and reactive power states under a unified time reference from network data carrying parasitic errors. Its practical application benefits include blocking false power allocation errors caused by clock misalignment at the source of the control chain, significantly reducing parasitic reactive power circulating losses between microgrid inverters due to phase measurement misalignment, and effectively improving the power supply quality and long-term operational reliability of distributed power systems under complex conditions without an external time reference.

[0198] Example 4:

[0199] like Figure 8 As shown in the figure, this embodiment four provides a microgrid distributed data acquisition and control method that takes into account communication delay. The method corresponds to the aforementioned system in terms of execution logic and technical architecture, and is mainly applied to AC microgrids containing multiple distributed inverter power sources.

[0200] In existing AC microgrid engineering applications, their distributed collaborative control heavily relies on the underlying communication network for power status data exchange between adjacent nodes. However, due to strong electromagnetic interference and physical barriers in actual industrial sites or wide-area deployment environments, communication networks are highly susceptible to interference, leading to dynamic delays and packet loss during data transmission.

[0201] In addition, most existing control methods use a fixed data acquisition frequency and a fixed communication network transmission and reception cycle for interaction, without considering adaptive adjustments based on fluctuations in the real-time operating conditions of the power grid and the quality of the communication network. This results in the inefficient use of limited communication bandwidth or the lag of some critical operating data during transient periods.

[0202] Meanwhile, because the filtering parameters of the high-power physical components inside the distributed inverter will objectively drift with increasing operating temperature and service life, it will be difficult to accurately predict the node state containing communication delays if a state observer with fixed parameters is used in the control algorithm. When the distributed consensus algorithm uses distorted data with time lag and without physical parameter correction for calculation, it will output biased frequency and voltage compensation commands, causing serious power distribution errors and huge reactive power circulating current losses when the distributed inverter performs droop control, endangering the overall coordinated regulation performance and operational stability of the microgrid.

[0203] To address the aforementioned technical problems, this method is primarily executed by a local control unit configured at each of the distributed inverter power supplies. In actual physical hardware, this local control unit typically comprises a heterogeneous digital control board based on a digital signal processor (DSP) and a field-programmable gate array (FPGA), and it specifically executes the following control steps:

[0204] First, the data sensing and adaptive scheduling steps are executed. The control system acquires the real-time operating status of the microgrid, extracts local high-frequency electrical parameters, and dynamically adjusts the data acquisition frequency and communication cycle based on the real-time operating status.

[0205] In the actual underlying hardware of a microgrid, this step relies on high-precision voltage transformers (PTs) configured on the AC bus side and wideband Hall current sensors configured on the inverter output side for physical quantity sensing. After passing through a low-pass anti-aliasing filter conditioning circuit, the analog electrical signal is sent to the analog-to-digital converter (ADC) peripheral module of the digital signal processor for discretization sampling.

[0206] The control method monitors the voltage fluctuation rate of the AC bus and the local active power change rate of the inverter output in real time based on the program logic preset in memory. When the algorithm determines that the microgrid is in a steady-state load range, the system controls the ADC peripheral to enter a frequency reduction sampling mode and extends the message transmission interval of the local communication controller to save processor computational overhead and network bandwidth. When the system determines that the power grid is experiencing transient and sudden operating conditions such as the starting of a large-capacity induction motor or the connection of a large nonlinear load, the ADC is quickly configured to a high-frequency continuous sampling mode, extracting the high-frequency dq rotating coordinate system fundamental phasor as the local high-frequency electrical parameters, while simultaneously significantly shortening the packet transmission cycle of network data packets. This step, through the dynamic linkage between the software algorithm and the hardware sampling peripheral, ensures that limited communication network resources are always tilted towards data representing the key dynamic characteristics of the power grid.

[0207] Secondly, dynamic routing and switching steps for the hardware links are executed. The control system monitors the quality status of multiple communication links and dynamically selects the optimal link to transmit the local high-frequency electrical parameters. In actual operating microgrid containers or distributed distribution cabinets, nodes are typically redundantly deployed with various physical communication hardware such as ZigBee wireless radio frequency modules based on the IEEE 802.15.4 standard, LoRa long-range spread spectrum modules, or power line carrier (PLC) modems. The Media Access Control (MAC) layer software of the local control unit periodically sends probe messages to the communication modules of adjacent nodes to calculate the transmission delay, packet loss rate, signal-to-noise ratio, and available channel bandwidth of each physical hardware communication link in the current period.

[0208] When the primary hardware link handling data transmission (such as an externally exposed ZigBee antenna module) is subjected to sudden high-frequency electromagnetic noise interference in the industrial environment, causing its overall transmission quality score to drop below the safety threshold, the control method will trigger the automatic switching logic of the underlying physical driver. This will wake up and activate the backup hardware link with the highest score (such as a PLC module attached to a physical cable) to relay the data transmission. This dynamic selection of the optimal link ensures the safe transmission of the collected local high-frequency electrical parameters to the adjacent distributed inverter power supply, guaranteeing the reliability of cross-node interaction of underlying electrical data at the physical transmission level.

[0209] Third, the system performs parameter identification and state delay reconstruction based on physical characteristics. The control system combines the local high-frequency electrical parameters with the real-time communication link status to identify the filtering parameters of the distributed inverter power supply online, enabling adaptive adjustment of the state observer model and reconstructing the current delay-free true power state vector of adjacent distributed inverter power supplies.

[0210] During long-term high-power operation of inverter hardware, the high-power LC filter configured at its grid-connected output is prone to physical parameter changes. For example, the inductance of the filter inductor will decrease due to the heating of the magnetic core, and the capacity of the filter film capacitor will decay due to partial discharge.

[0211] To overcome this hardware aging phenomenon, the digital controller allocates a matrix operation space in the random access memory (RAM), and uses the voltage and current vectors in the local high-frequency electrical parameters to calculate the actual inductance and capacitance of the hardware LC filter at the current moment through a recursive least squares algorithm in a loop.

[0212] Subsequently, the digital controller updates these actual physical parameters into the state matrix of the state observer model built into the memory, completing the internal algorithm model's adaptive tracking of the aging of external physical entities. Simultaneously, the digital controller parses the timestamps from the headers of received data packets from adjacent nodes, calculated by hardware timers, to determine the actual communication time lag.

[0213] The algorithm uses the updated state observer model to perform forward inference calculations with numerical integration on the time axis to compensate for the time difference caused by communication delay during the physical propagation of data packets. It restores and reconstructs the current delay-free true power state vector of adjacent distributed inverters within the microprocessor from network data with historical delay errors.

[0214] Fourth, the multi-node collaborative consensus instruction calculation step is executed. The control system performs distributed consensus calculation based on the aforementioned delay-free true power state vector, generating frequency compensation terms and AC bus voltage amplitude compensation terms. After acquiring the state data of adjacent nodes stripped of communication delay errors, the central processing unit (CPU) of the local control unit calls the finite-time consensus algorithm subroutine.

[0215] This algorithm uses algebraic operations to compare the current actual power state of the local network with the reconstructed, delay-free true power state vectors of each adjacent node, calculating the power output deviation between nodes in the distributed network topology. The algorithm calculates convergence components for frequency regulation for active power deviation and for voltage regulation for reactive power deviation. These calculation results are discretized in the floating-point unit within the controller and assigned as frequency compensation terms and AC bus voltage amplitude compensation terms, respectively, providing high-precision digital instructions for subsequent adjustments to the underlying control objectives of the physical inverter.

[0216] Finally, the physical simulation injection step of the control command is executed. The control system injects the frequency compensation term and the AC bus voltage amplitude compensation term into the physical droop control loop of the distributed inverter power supply in the form of analog signals, completing the closed-loop dynamic regulation of the microgrid.

[0217] To avoid the computation cycle delay caused by the execution of code by the digital processor and the update delay of the digital pulse width modulation (PWM) module, this method uses a pure analog hardware loop to complete the cross-domain conversion in the final instruction execution stage. The digital control board converts the frequency compensation term and AC bus voltage amplitude compensation term calculated in memory into the corresponding DC analog voltage level in real time through a high-speed, high-precision digital-to-analog converter (DAC) chip.

[0218] These analog voltage signals are transmitted directly to the high-frequency precision operational amplifier network on the bottom driver board of the inverter power supply through isolation barriers and copper traces. The operational amplifier network, according to the resistor-capacitor ratio pre-configured on the circuit board (i.e., the hardware-defined droop factor), weights and sums the analog compensation signal with the reference voltage and frequency signals acquired by the system to generate continuously changing control command levels.

[0219] The integrated voltage level is directly fed into an analog comparator, where it is compared with a high-frequency triangular carrier wave to generate continuous PWM drive pulses. These pulses then physically control the switching on and off of high-power semiconductor devices such as insulated-gate bipolar transistors (IGBTs) or silicon carbide (SiC). Through this direct signal injection from the analog circuit, the system completes closed-loop dynamic regulation of the microgrid bus voltage and frequency within a nanosecond response time.

[0220] Considering the actual industrial application scenarios of microgrids, the method provided in this embodiment closely integrates the timeliness of the upper-layer digital algorithm with the physical evolution characteristics of the lower-layer electrical hardware. By dynamically scheduling the hardware sampling module and communication network hardware, the timeliness of the interaction of the lower-layer data under disturbed environments is improved; by identifying the parameter drift of the actual physical LC filter online and combining it with the forward time extrapolation algorithm, the dual errors caused by hardware aging and network communication delay, which are common in long-distance distribution networks, are effectively overcome, restoring the distorted historical lagging data to the true power state with current timeliness; finally, the compensation command calculated based on the real, delay-free data is directly injected into the physical analog control loop through digital-to-analog conversion and operational amplifier circuits.

[0221] This method effectively reduces the negative impact of communication delays and hardware parameter drift on coordinated control commands, reduces the internal reactive power circulation of distributed inverters during parallel operation, and improves the dynamic adjustment capability and power supply stability of AC microgrid systems under non-ideal operating conditions such as communication congestion and equipment aging.

[0222] Example 5:

[0223] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0224] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0225] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0226] Bus 102 may include a path for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0227] The memory 103 stores a computer program corresponding to the microgrid distributed data acquisition and control method considering communication delay in the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0228] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0229] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A distributed data acquisition and control system for microgrids that takes into account communication latency, applied to an AC microgrid containing multiple distributed inverter power sources, characterized in that, Each of the distributed inverters is equipped with a local control unit, which, from top to bottom, comprises: An adaptive high-frequency data acquisition layer is used to acquire the real-time operating status of the microgrid, extract local high-frequency electrical parameters, and dynamically adjust the data acquisition frequency and communication cycle according to the real-time operating status. A multi-mode communication link dynamic switching layer is used to monitor the quality status of multiple communication links and dynamically select the best link to transmit the local high-frequency electrical parameters. The dual adaptive asynchronous time-delay state observation layer is used to combine the local high-frequency electrical parameters with the real-time communication link status to identify the filtering parameters of the distributed inverter power supply online in order to achieve adaptive adjustment of the state observer model and reconstruct the current delay-free real power state vector of the adjacent distributed inverter power supply. A distributed consensus decision layer is used to perform distributed consensus calculations based on the delay-free true power state vector, generating frequency compensation terms and AC bus voltage amplitude compensation terms. A pure analog adaptive droop execution layer is used to inject the frequency compensation term and AC bus voltage amplitude compensation term into the physical droop control loop of the distributed inverter power supply in the form of analog signals to complete the closed-loop dynamic adjustment of the microgrid. The dual adaptive asynchronous time-delay state observation layer includes an online inverter parameter identification unit. This unit identifies parameters based on the recursive least squares method to update the difference equation parameters in the reduced-order state equation of the distributed inverter power supply and inversely calculates the parameters at the discrete time step. Actual filter inductance value and capacitance value : ； ； in, The system sampling period is to , to All of these are the parameters of the difference equation obtained through identification; When the calculated actual filter inductance value or capacitance value When the deviation from the preset nominal inductance or capacitance value exceeds a preset proportion, the dual adaptive asynchronous delay state observation layer automatically updates the state matrix of the built-in state observer. and input matrix This enables dual adaptive adjustment of the parameters of the state observer model; The dual adaptive asynchronous delay state observation layer simultaneously adjusts the forward inference step size based on the real-time transmission delay and real-time packet loss rate of the communication link, and reconstructs the adjacent distributed inverter power supplies through forward inference calculation. At the present moment The delay-free true power state vector : ； ； in, For integration variables, For the adjacent distributed inverter power supply At a historical moment Power data transmitted, From a historical moment up to the current moment The change in state This is the input vector collected locally.

2. The microgrid distributed data acquisition and control system taking into account communication delay as described in claim 1, characterized in that, The adaptive high-frequency data acquisition layer automatically switches the data acquisition frequency and communication cycle between steady-state operation mode, load fluctuation mode and fault transient mode by monitoring the local active power change rate and AC bus voltage deviation in real time. Simultaneously, a parameter-weighted transmission mechanism based on the Pearson correlation coefficient is adopted to dynamically adjust the transmission weights of each electrical parameter. The transmission weight update rule is as follows: ； in, For the updated transmission weights, As the initial weights, To adjust the factor, The Pearson correlation coefficient between the local high-frequency electrical parameters and the preset fault status label.

3. The microgrid distributed data acquisition and control system taking into account communication delay as described in claim 1, characterized in that, The multi-mode communication link dynamic switching layer monitors the transmission delay of each communication link in real time. Packet loss rate Signal-to-noise ratio and available bandwidth And calculate the overall quality score based on the following formula. : ； in, For the maximum allowable delay, To the maximum allowable packet loss rate, For maximum signal-to-noise ratio, For the maximum available bandwidth, , , , These are preset weighting coefficients, and their sum is 1; When the overall quality score When the threshold for link switching is lower than the preset threshold, a switching process is initiated to establish a backup communication link and then disconnect the current communication link.

4. The microgrid distributed data acquisition and control system taking into account communication delay according to claim 1, characterized in that, The decentralized consensus decision layer uses a finite-time consensus algorithm to solve the frequency compensation term and the AC bus voltage amplitude compensation term, and its control law function has the following form: ； in, For the first Compensation control variables for a distributed inverter power supply. To control the gain, For the preset index, For the first A set of neighboring distributed inverters that are communicating normally. Adjacent distributed inverter power supply The delay-free true power state vector, This is the local, current, latency-free true power state vector.

5. The microgrid distributed data acquisition and control system taking into account communication delay according to claim 1, characterized in that, The decentralized consensus decision layer is also used to proactively detect anomalies in adjacent distributed inverter power sources: By periodically sending heartbeat signals to adjacent distributed inverters, paging detection is performed when no heartbeat response is received. When the target adjacent distributed inverter is determined to be abnormal, it is automatically removed from the communication topology, so that the remaining distributed inverters can re-form an undirected and connected sparse communication topology. When an abnormality is detected and a neighboring distributed inverter power supply recovers, it is automatically reconnected to the control loop via a handshake protocol.

6. The microgrid distributed data acquisition and control system considering communication delay according to claim 1, characterized in that, The frequency-active droop control and voltage-reactive droop control in the pure analog adaptive droop execution layer are both implemented by pure analog hardware circuits. The pure analog adaptive droop execution layer uses digitally calculated frequency compensation terms. AC bus voltage amplitude compensation item Direct physical injection droop control loop: ； ； in, and These are the adjusted frequency and voltage commands, respectively. and These are the rated frequency and rated voltage, respectively. and This refers to the actual active and reactive power. and Rated active and reactive power, and This is the droop coefficient.

7. The microgrid distributed data acquisition and control system considering communication delay according to claim 1, characterized in that, The decentralized consensus decision layer is also equipped with a data credibility assessment module based on spatiotemporal physical coupling; The data credibility assessment module based on spatiotemporal-physical coupling is used to perform dynamic trust assessment and isolation filtering on the latency-free real power state vector before distributed consensus computation. The specific execution steps include: Based on the preset global admittance matrix and the real-time electrical quantities collected by the local AC bus, Kirchhoff's physical space constraint equations are constructed to deduce the values ​​of adjacent distributed inverter power supplies. The physical reference state vector at the current moment ; Calculate the physical reference state vector The adjacent distributed inverter power supply reconstructed by the dual adaptive asynchronous time-delay state observation layer Delay-free true power state vector Spatiotemporal deviation residuals between Its defining formula is: ； Based on the spatiotemporal deviation residual Construct a dynamic trust decay function to calculate the local control unit's trust level for each adjacent distributed inverter power supply in real time. Dynamic trust weight : ； in, The preset confidence sensitivity adjustment coefficient, A safety deviation threshold set based on historical normal steady-state data; When the decentralized consensus decision layer solves for the frequency compensation term and the AC bus voltage amplitude compensation term, the dynamic trust weight is used. This adaptive decay coefficient is multiplied into the corresponding adjacent node state difference term; When the spatiotemporal deviation residual When the value exceeds the preset extreme value alarm threshold, the data reliability assessment module will assign the dynamic trust weight to the corresponding adjacent distributed inverter power supply. The value is forcibly assigned to zero, and the extreme value alarm threshold is greater than the safety deviation threshold. .

Images