A direct current micro-grid adaptive control method and system with cooperative prediction compensation

Abstract

The application discloses a kind of direct current microgrid adaptive control method and system of cooperative prediction compensation, the method includes: each DG measures local voltage and output current and receives neighbor current signal;To the neighbor current signal after communication time delay, after state estimation, molecule step forward integral prediction, global deviation correction and low-pass filtering by Luenberger state observer, output compensation current;Based on the compound error calculated after the compensation of neighbor current and local voltage and compared with adaptive threshold, meet the condition and trigger communication update, otherwise, the predicted value of event-triggered observer is used;Based on the neighbor current calculated by trigger update or prediction retention, the secondary control compensation quantity is calculated, and the droop control output is superposed to generate a modified voltage reference value, which is driven by a double-loop PI controller to drive a DC / DC converter;Wherein, the estimation error based on event-triggered observer is used to construct observer confidence, and the sensitivity of adaptive threshold is adjusted by using confidence, to balance the prediction compensation accuracy and communication frequency.

Description

Technical Field

[0001] This invention belongs to the field of distributed control technology for microgrids, and particularly relates to an adaptive control method and system for DC microgrids with collaborative predictive compensation. Background Technology

[0002] With the increasing proportion of renewable energy generation, DC microgrids, due to their high power conversion efficiency, have become an important platform for distributed generation (DG) integration. In islanded operation mode, distributed secondary control, based on a consensus protocol, achieves voltage recovery and current sharing through neighbor information exchange. The effectiveness of this control is highly dependent on the real-time performance of the communication network; however, transmission delays exist in actual communication links. Communication delays can disrupt the synchronization of the consensus protocol, leading to decreased voltage recovery accuracy, current sharing failure, and even system instability. Therefore, ensuring the performance of distributed secondary control under communication delay constraints has become one of the core challenges in the field of DC microgrid control.

[0003] Existing methods for handling communication latency can be mainly divided into two categories: passive tolerance and active compensation. Passive tolerance methods use robust control techniques to maintain system stability within a certain latency range. However, passive tolerance methods essentially sacrifice control performance for robustness, resulting in a conservative tolerance range for latency and difficulty in eliminating the impact of latency on the system. Active compensation methods, on the other hand, use prediction or estimation to address latency at its source. However, most active compensation methods are designed within a periodic communication framework and do not consider the optimal utilization of communication resources. Therefore, further research is needed on how to address the impact of latency while maintaining communication efficiency.

[0004] In terms of communication efficiency optimization, Event Triggered Control (ETC) only performs communication when the system state meets specific conditions, significantly reducing communication frequency while maintaining control performance. Static ETC uses a fixed threshold to determine triggering conditions; however, a fixed threshold triggers too frequently during transient processes such as load changes, and may lead to untimely triggering near steady state. To address this issue, Dynamic ETC introduces an internal dynamic variable to adaptively adjust the triggering threshold. However, most existing event triggering methods are based on the assumption of ideal communication, that is, that information transmission has no delay or the delay is negligible. In real-world communication delay scenarios, event triggering control maintains the old value during non-triggering periods, which has a superimposed effect with communication delay, further increasing the deviation between the neighbor information used by the controller and the actual value, thus seriously affecting control performance. Therefore, existing research has the following shortcomings: passive delay tolerance methods are highly conservative but cannot eliminate steady-state errors; most event triggering methods ignore the impact of communication delay, while delay compensation methods mostly use periodic communication, and the two lack coordinated design. Summary of the Invention

[0005] To address the problems in distributed secondary control of DC microgrids, such as communication delay leading to lag in neighbor current information, inability to eliminate steady-state current deviation, and inability to restore voltage to rated values, as well as the lack of coordinated design between existing delay compensation methods and event-triggered control, this invention provides a coordinated predictive compensation adaptive control method for DC microgrids. This method eliminates the impact of communication delay through forward predictive compensation using a state observer, achieving a coordinated design between active predictive compensation and adaptive event triggering. Furthermore, it utilizes observer confidence feedback to adjust trigger sensitivity, thereby reducing communication overhead while ensuring voltage recovery and current sharing control accuracy, achieving an automatic balance between compensation accuracy and communication frequency.

[0006] According to one aspect of the present invention, a collaborative predictive compensation adaptive control method for a DC microgrid is provided, comprising: each distributed generation unit measuring its local voltage and output current in each sampling period, and receiving neighbor current signals through a communication network; performing state estimation, forward integral prediction with numerator step size, global deviation correction, and low-pass filtering on the received neighbor current signals that arrive after communication delay using a Luenberger state observer, and outputting a compensation current; calculating the composite error of fused current deviation and voltage deviation based on the compensated neighbor current and local voltage, and comparing it with an adaptive threshold; and triggering a communication update if a preset event triggering condition is met. Otherwise, the predicted value of the event-triggered observer is used as the neighbor current signal; the secondary control compensation amount is calculated based on the neighbor current signal of the trigger update or prediction hold, and superimposed with the droop control output to generate a corrected voltage reference value. The voltage reference value is used to generate a duty cycle signal to drive the DC / DC converter via a dual-loop PI controller; wherein, the average EMA error of each neighbor channel is constructed based on the exponential moving average of the estimation error of the event-triggered observer, and the upper and lower limits of the observer confidence are evaluated based on the average EMA error and the preset confidence level to construct the observer confidence level; the sensitivity of the adaptive threshold is adjusted by feedback using the observer confidence level to balance the prediction compensation accuracy and the communication frequency.

[0007] As a further technical solution, the Luenberger state observer takes the output current of the distributed generation unit as the observed state and is constructed based on a second-order state-space free response model that considers the influence of LC filter and load impedance; the neighbor current signal that arrives after communication delay is used as the observer input to estimate the neighbor current and auxiliary state.

[0008] As a further technical solution, the molecular step forward integral prediction includes: dividing the prediction time into multiple sub-steps, using a no-input free response model to perform stepwise iterative forward integration, advancing the estimated state of the Luenberger state observer to the current moment, and obtaining the current prediction value.

[0009] As a further technical solution, the global deviation correction includes: calculating the global mean deviation using the local real-time current mean and the time-delayed current mean, and superimposing the global mean deviation onto the current value after forward integration prediction, so as to suppress the steady-state deviation caused by model mismatch.

[0010] As a further technical solution, the composite error is constructed as follows: the difference between the compensated neighbor current and the neighbor current value recorded at the most recent triggering time is defined as the current deviation, the difference between the local voltage and the rated voltage is defined as the voltage deviation, and the weighted sum of the square of the current deviation and the square of the voltage deviation is taken as the composite error.

[0011] As a further technical solution, the dynamic adjustment law of the adaptive threshold is as follows: after a trigger occurs, the adaptive threshold is increased by a preset growth coefficient to avoid continuous triggering, and a minimum trigger interval constraint is introduced to limit the minimum time interval between two consecutive triggers; during the non-trigger period, the adaptive threshold is decayed by a preset decay factor, and a lower limit of the threshold is set to maintain sensitivity to system changes.

[0012] As a further technical solution, the event triggering condition is also designed based on the neighbor inconsistency; the neighbor inconsistency is the sum of the squares of the differences between the output current of the current distributed generation unit and the compensated neighbor currents of each neighbor distributed generation unit.

[0013] As a further technical solution, the sensitivity of the adaptive threshold is adjusted by feedback using the observer confidence, including: adjusting the relative threshold coefficient of the adaptive threshold using the observer confidence as a multiplicative factor; increasing the relative threshold coefficient when the observer confidence is higher than a preset value; and decreasing the relative threshold coefficient when the observer confidence is lower than a preset value.

[0014] According to one aspect of this specification, a collaborative predictive compensation adaptive control system for a DC microgrid is provided, comprising: a data acquisition module for each distributed generation unit to measure local voltage and output current in each sampling period, and to receive neighbor current signals via a communication network; a predictive compensation module for performing state estimation, forward integral prediction with numerator step size, global deviation correction, and low-pass filtering on the received neighbor current signals that arrive after communication delay, and then outputting a compensation current; and an adaptive event triggering module for calculating the composite error of fused current deviation and voltage deviation based on the compensated neighbor current and local voltage, and comparing it with an adaptive threshold; if a preset event triggering condition is met, then... A communication update is triggered; otherwise, the predicted value of the event-triggered observer is used as the neighbor current signal. A collaborative optimization module is used to construct the average EMA error of each neighbor channel based on the exponential moving average of the estimation error of the event-triggered observer, and to construct the observer confidence level by evaluating upper and lower limits based on the average EMA error and a preset confidence level. The observer confidence level is used to adjust the sensitivity of the adaptive threshold through feedback adjustment to balance the prediction compensation accuracy and communication frequency. A secondary control output module is used to calculate the secondary control compensation amount based on the neighbor current signal held by the trigger update or prediction, and to superimpose it with the droop control output to generate a corrected voltage reference value. The voltage reference value is used to generate a duty cycle signal via a dual-loop PI controller to drive the DC / DC converter.

[0015] As a further technical solution, in the prediction compensation module, the Luenberger state observer takes the output current of the distributed generation unit as the observed state and is constructed based on a second-order state-space free response model that considers the influence of LC filters and load impedance. The neighbor current signal that arrives after communication delay is used as the observer input to estimate the neighbor current and auxiliary state. The numerator step forward integral prediction uses a no-input free response model to advance the estimated state to the current time. The global deviation correction uses the local real-time current mean and the time-delayed current mean to calculate the global mean deviation for correction.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. An active compensation scheme based on the combination of Luenberger state observer and forward prediction is designed to reconstruct the real-time value of neighbor current from the time delay signal, and suppress the steady-state deviation caused by model mismatch through global deviation correction, thereby eliminating the impact of communication delay on distributed secondary control from the source. 2. A composite error metric integrating current deviation and voltage deviation and an adaptive dynamic threshold triggering mechanism are constructed. During non-triggering periods, the local observer performs neighbor state prediction and maintenance, which effectively suppresses the superposition error of time delay and event triggering, and significantly reduces communication overhead while ensuring control accuracy. 3. By using observer confidence feedback to adjust trigger sensitivity, and by coordinating active prediction compensation with adaptive event triggering, an automatic balance between compensation accuracy and communication frequency is achieved, thereby improving the system's control efficiency and communication efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a microgrid topology diagram constructed in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the communication delay impact mechanism in an embodiment of the present invention; Figure 3 This is a flowchart of the active prediction and compensation process based on the state observer in an embodiment of the present invention; Figure 4 This is a diagram of the adaptive event triggering control mechanism in an embodiment of the present invention; Figure 5 This is a design diagram of the integrated controller in an embodiment of the present invention. Detailed Implementation

[0019] Given the current technological background, there is an urgent need to propose a novel distributed secondary control method for DC microgrids that can eliminate the impact of communication delay on consensus protocols under communication delay constraints and effectively integrate delay compensation with communication resource optimization, thereby improving the voltage recovery accuracy and current distribution regulation capability of DC microgrids in scenarios with communication delay. To this end, this invention proposes a collaborative predictive compensation adaptive control method for DC microgrids. By reconstructing the real-time values ​​of neighboring currents from the delay signal using an active compensation scheme based on a Luenberger state observer and forward prediction, and combining an adaptive event triggering mechanism that integrates current and voltage deviations with observer confidence feedback, predictive compensation and event triggering are collaboratively optimized. This effectively solves the problems of passive tolerance methods failing to eliminate steady-state errors, active compensation methods using periodic communication causing communication redundancy, and the deterioration of control performance caused by the superposition of event triggering and delay.

[0020] The terms “comprising” and “having”, and any variations thereof, in the specification, claims, and accompanying drawings of this invention are intended to cover a non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0022] This invention primarily utilizes the active prediction compensation theory of state observers and a distributed adaptive event-triggered control strategy to improve the convergence accuracy and communication efficiency of distributed secondary control in DC microgrids under communication delay constraints. It proposes a collaborative prediction compensation adaptive control method for DC microgrids. This invention fully leverages the advantages of Luenberger state observers and forward prediction in delay compensation. It eliminates the impact of delay by reconstructing the real-time values ​​of neighboring currents from the delay signal and performing global deviation correction. An adaptive event-triggered mechanism is designed for optimized communication resources, and the trigger threshold sensitivity is adjusted using observer confidence feedback, thereby achieving synergy between prediction compensation and event triggering. The control effect obtained through this invention is more precise, effectively eliminating the impact of communication delay on voltage recovery and current equalization, and significantly reducing communication overhead.

[0023] The method provided by this invention can be implemented using computer simulation software technology, and all steps can be completed in Simulink / RTlab. The microgrid topology constructed in the embodiments of this invention is as follows: Figure 1 As shown.

[0024] This invention provides a DC microgrid topology model and a distributed secondary control protocol construction, including: establishing a DC microgrid topology model composed of N distributed generation units (DGs) in islanded operation mode, with each unit connected in parallel to a common DC bus via a DC / DC converter and connected to local and common loads via line impedance; establishing an electrical dynamic model for each DG, including the dynamic relationship between filter inductor current and output voltage; employing droop control at the primary control layer, and obtaining a local voltage reference value by introducing virtual impedance; using a dual-loop PI structure of voltage outer loop and current inner loop at the local control layer to track the reference voltage; and constructing a distributed secondary control protocol, consisting of a voltage recovery channel based on the integral of local voltage deviation and a current sharing channel based on the integral of neighbor weighted current difference, which are superimposed to obtain the voltage reference value.

[0025] Specifically, this embodiment focuses on a DC microgrid operating in islanded mode. The system comprises four distributed generation units (DGs), each connected in parallel to a common DC bus via a DC / DC converter, and connected to local and common loads through line impedance.

[0026] Define L i With C i These are the filter inductor and filter capacitor for the i-th DG, respectively; i L,i V is the inductor current; i The output voltage of DG; d i For duty cycle control signal; V in,i It is a DC input voltage source; i i The output current injected by the DG into the bus.

[0027] The electrical dynamics of the i-th DG (i=1,2,3,4) can be represented as follows: .

[0028] For primary control, each DG employs droop control to achieve initial power distribution. Droop control introduces a virtual impedance, causing the output voltage of each DG to decrease linearly as its output current increases.

[0029] Define V nom Rated bus voltage; r d,i v is the droop coefficient of the i-th DG; ref,i This is the local voltage reference value after droop control processing.

[0030] The droop control law for the i-th DG is: .

[0031] Define K pv K iv The proportional-integral gain, K, constitutes the voltage loop. pi Kii The proportional-integral gain i that constitutes the current loop ref,i This is the local current reference value after droop control processing.

[0032] In the local controller of the DG, a voltage outer loop and current inner loop structure is used to track the reference voltage. The mathematical expression of the dual-loop controller is as follows: .

[0033] Next, we will perform communication topology modeling and time delay problem description, including: modeling the communication network between adjacent DGs as an undirected graph, with nodes corresponding to DG cells and edges representing bidirectional communication capabilities; defining the adjacency matrix and degree matrix, and constructing the Laplace matrix, adopting a ring communication topology; establishing a communication time delay model, assuming that there is a time delay in the transmission of information from node j to node i, and that it does not exceed the upper bound of the time delay tolerance, and analyzing the problem that the time delay causes the current-sharing channel to use the lag information, thus generating steady-state deviation.

[0034] The core of distributed secondary control lies in information sharing between adjacent DGs, and this communication structure can be described using graph theory tools.

[0035] Specifically, the communication network is defined as an undirected graph G=(V,E), where V={1,2,...,N} is the set of nodes, and each edge (i,j) in the edge set E represents the bidirectional communication capability between DG.

[0036] In the communication topology, A=[a ij ]∈R N×N This is an adjacency matrix, where each element indicates whether nodes are connected. The number of connections between each node and its neighbors is determined by its degree. It is determined that the degree of all nodes constitutes a degree matrix. The convergence analysis of the consensus protocol depends on the Laplace matrix L=DA. The four DG systems studied in this invention adopt a ring communication topology, and the adjacency matrix under this topology is: .

[0037] Define k v For voltage-controlled gain, G ii For voltage recovery weights.

[0038] While droop control achieves initial power distribution, it introduces a bus voltage deviation. The magnitude of this deviation depends on the load level and the droop coefficient. This steady-state error requires secondary control to eliminate. Secondary control can restore the bus voltage to its rated value while maintaining the power distribution ratio. The distributed secondary control protocol includes two channels: a voltage recovery channel uses the integral of the local voltage deviation to generate a compensation amount, thereby causing the voltage to converge towards its rated value. .

[0039] Define k i For current-controlled gain; c i This is the power sharing factor, which is usually set as the reciprocal of the droop factor.

[0040] The current-sharing channel adjusts the output distribution of each DG based on the integral of the weighted current difference between neighbors, so that the weighted current of all DGs tends to be consistent: ,

[0041] Therefore, the final voltage reference value of the i-th DG can be expressed as follows: .

[0042] Under ideal communication conditions (no delay, no packet loss), the above protocol can guarantee asymptotic convergence of the voltage recovery and current sharing objectives when the communication graph is connected. However, the transmission delay in actual communication links makes this ideal condition invalid, and the control performance will be affected.

[0043] The schematic diagram of the communication delay impact mechanism in this embodiment of the invention is shown below. Figure 2 As shown.

[0044] Define τ max Let τ be the upper bound for tolerable communication delay. In practical communication networks, information transmission between nodes (DGs) involves delay. Suppose that there is a communication delay τ when node j transmits information to node i. ij Then the information received by node i from neighbor j at time t can be expressed as: , When communication delay exists, the current-sharing channel becomes: .

[0045] Because of the time inconsistency between the delayed signal and the local real-time signal, the neighbor current information acquired by each DG lags behind the true value. The consensus protocol performs equalization adjustment based on the outdated information, which makes it impossible to completely eliminate the steady-state current deviation. This leads to a decrease in the accuracy of current equalization, the voltage cannot be restored to the rated value, and the system will become unstable.

[0046] Besides latency issues, the periodic communication mechanism of traditional distributed control also suffers from efficiency problems. In each sampling period, all DGs need to send information to their neighbors. This communication strategy wastes bandwidth resources, causes transmission congestion, and further worsens latency.

[0047] Based on the above analysis, the control objective of this invention is as follows: 1) Voltage Recovery: The output voltage of each DG converges to its rated value, i.e. .

[0048] 2) Current sharing: Each DG shares the load current proportionally, i.e. .

[0049] 3) Information exchange between DGs is only performed when the event triggering conditions are met, and does not generate Zeno behavior (Zeno behavior is the phenomenon of generating an unlimited number of triggers within a finite time).

[0050] 4) During the communication delay τ ij ≤τ max Under the constraints, the above voltage and current targets can still be achieved, and the closed-loop system remains stable in the Lyapunov sense.

[0051] The active prediction and compensation process based on the state observer in this embodiment of the invention is as follows: Figure 3 As shown.

[0052] Communication delay causes received neighbor information to lag behind the actual value. To reconstruct the current-moment information from the delayed signal, this invention designs an active compensation scheme based on a combination of Luenberger state observer and forward prediction. The scheme uses the DG output current as the observed state and establishes a second-order state-space free response model considering the effects of LC filters and load impedance. Using the neighbor current signal arriving after the communication delay as input, a Luenberger state observer is designed to estimate the neighbor current and auxiliary states. Based on the observer's estimation, the prediction time is divided into multiple sub-steps, and iterative forward integration is performed using the input-free free response model to advance the estimated state to the current moment and obtain the current prediction value. The global mean deviation is calculated using the local real-time current mean and the delayed current mean, and the prediction value is globally biased to suppress steady-state deviation caused by model mismatch. A first-order low-pass filter is applied to the bias-corrected compensation output to suppress high-frequency noise introduced during the prediction process, resulting in the final compensated current.

[0053] 1) State-space model of the observed system Define R load Let A be the equivalent load resistance, x(t) be the state vector, and A be the load resistance. o The system matrix is ​​denoted by C; the filter capacitor is α. f >0 is the attenuation coefficient for auxiliary states, used to describe the dynamic characteristics of the current change rate, so as to ensure that the matrix eigenvalues ​​are negative, thereby ensuring system stability.

[0054] Considering the dynamics at the output of a single DG, the current on the DC bus side is affected by the LC filter and the load impedance. Taking the output current of the DG as the observed state, the following second-order state-space free response model is established: .

[0055] Due to matrix Ao The bottom right element needs to be negative, so an auxiliary variable α is constructed. f To ensure that the eigenvalue is negative, this variable is an additional state variable artificially introduced to close the state-space model and construct a second-order free response form. It does not correspond to the actual physical quantity.

[0056] 2) Luenberger State Observer definition This is an estimate of the neighbor's current. This is an estimate of the auxiliary state. The neighbor's current signal received after a time delay. L is the observer gain vector. ob =[L1,L2] T L1 and L2 are elements of matrix Lob. This is the output estimation error.

[0057] Using the neighbor current signal that arrives after communication delay as the input of the observer, the following Luenberger state observer is designed: , Define the observer estimation error as: .

[0058] Let C0 be the output matrix, I be the identity matrix, and s be the factors in the Laplace transform. The characteristic equation of the observer error dynamics is: .

[0059] 3) Forward prediction compensation definition To predict the state vector, This is the first component of the predicted state vector at step M. This is the predicted current value. To predict initial conditions.

[0060] The observer can only estimate the system state at the corresponding time point of the time-delay signal, i.e., the state near time t-τ. To compensate for the time lag caused by the time delay τ, forward integration is performed using the system dynamic model based on the observer's estimation, advancing the estimated state to the current time. The observer's estimated state at time t is: The goal of forward prediction is to calculate the predicted state value at time t+τ. The prediction time τ is divided into M equal sub-steps, and forward integration is performed iteratively. , Within the timescale of time delay τ, the change in system input is relatively small compared to the state itself, so the forward prediction stage uses a free response with no input.

[0061] 4) Global bias correction and filtering definition This is the local real-time average current. The average value of the time-delay current. This is the compensation current after deviation correction.

[0062] Because there may be a mismatch between the observer model parameters and the actual system, and because forward prediction is based on the free response, the predicted values ​​may exhibit steady-state bias. Therefore, a global bias correction mechanism is introduced. The global mean bias b(t) is calculated using the local current: .

[0063] definition For the final compensated output current, b f is the filter coefficient, and h is the sampling time.

[0064] Finally, to suppress the high-frequency noise introduced during the prediction process, a first-order low-pass filter is applied to the compensation output: .

[0065] The adaptive event triggering control mechanism in this embodiment of the invention is as follows: Figure 4 As shown.

[0066] To address the issues of communication redundancy caused by periodic communication, an adaptive event-triggered control is designed. Communication updates are only performed when the system state changes meet the triggering conditions, and the local observer is used to predict and maintain the neighbor states during non-triggering periods.

[0067] This adaptive event-triggered control design defines the difference between the current current value of a neighbor and the value recorded at the most recent trigger moment, constructing a composite error that integrates current deviation and voltage deviation. Based on the composite error, neighbor inconsistency, and dynamic threshold variables, event triggering conditions are designed. When the composite error exceeds the threshold, a communication update is triggered; otherwise, the current value is maintained. A dynamic threshold adaptive adjustment law is designed, increasing the threshold by an increment coefficient after triggering to avoid continuous and dense triggering, and slowly decreasing it by an attenuation factor when not triggered to maintain sensitivity, with a lower bound set for the threshold. A minimum trigger interval constraint is introduced to limit the minimum time interval between two consecutive triggers.

[0068] In this embodiment, the definition is... The current current value of the neighboring unit. The neighbor current value recorded at the most recent trigger time. and Two forms characterizing the current of the same neighbor: the former is the current compensation estimate, and the latter is the true value. They are aligned at each triggering moment using the following formula: , That is, the compensation output is reset by the actual communication value at the trigger time, and predictive compensation is performed during non-trigger periods. Replacing real current The secondary control law is then input.

[0069] To comprehensively assess the necessity of communication updates, the following composite error is defined: .

[0070] Define σ i and σ v These are the weighting coefficients for current deviation and voltage deviation, respectively.

[0071] The composite error combines current deviation and voltage deviation, and this process can be expressed by the following formula: .

[0072] definition Let α be the inconsistency degree between node i and its neighbors, and η be the relative threshold coefficient. i This is a dynamic threshold variable. When the composite error exceeds the threshold, a communication update is triggered; otherwise, the current value remains unchanged.

[0073] Secondly, the event triggering conditions are defined as follows: .

[0074] Define k g The threshold growth coefficient, This is the upper bound of the threshold after confidence adjustment. The threshold is increased after triggering to avoid communication waste caused by continuous and dense triggering.

[0075] To enable the trigger threshold to adaptively adjust according to the system's operating status, the following rule is designed when a trigger occurs: .

[0076] Define β∈(0,1) as the attenuation factor, η min The threshold is set as the lower bound and is greater than 0. When not triggered, the threshold decays slowly to maintain sensitivity to system changes.

[0077] When not triggered, the following applies: .

[0078] definition and T represents the two consecutive trigger times of node i. minThis is the minimum trigger interval.

[0079] Finally, a minimum trigger interval constraint is introduced: .

[0080] In ideal event-triggered control, only a zero-order hold (ZOH) is typically used during non-triggered periods. In scenarios with communication delays, the hold value of the ZOH already contains delay errors, causing the actual error to accumulate over time.

[0081] To address this issue, the present invention provides a neighbor state prediction and retention method during non-triggered periods, comprising: a state observer with identical design structure but independently configured gain for each neighbor channel, used to predict neighbor current during non-triggered periods; when the triggering condition is not met, the neighbor current information used by the secondary controller is provided by the short-term prediction value of the observer to suppress the superposition error of time delay and event triggering.

[0082] This invention designs an independent state observer for each neighbor channel to predict neighbor currents during non-triggered periods. The structure of this observer is the same as that of the prediction-compensation observer, but the gain parameter is configured independently. , in, The observer state of the neighbor in the event triggering module; L et Adds gain to the event-triggered observer.

[0083] When the triggering condition is not met, the neighbor current information used by the secondary controller is provided by the short-term prediction value of the observer: , The time increment for prediction, i.e., the forward step size when the event-triggered observer makes short-term predictions, is used when the triggering condition is not met and the secondary controller cannot obtain the new real current from the neighbor. This state of the event-triggered observer in the formula is used. Perform a first-order extrapolation (linear prediction) to advance the current estimate. Time is used to obtain short-term predicted current.

[0084] This embodiment also provides observer confidence feedback and collaborative optimization, including: defining the exponential moving average (EMA) of the estimation error of the event-triggered observer; constructing the observer confidence based on the average EMA error of all neighboring channels and the upper and lower limits of confidence assessment; and using the observer confidence to adaptively adjust the relative threshold coefficient in the triggering condition, relaxing the triggering condition and reducing the communication frequency when the confidence is high, and tightening the triggering condition when the confidence is low, so as to achieve a balance between prediction compensation accuracy and communication frequency.

[0085] definition λ represents the instantaneous estimation error of the observer, and λ is the EMA smoothing coefficient.

[0086] The estimation accuracy of an event-triggered observer directly affects the information quality during non-triggered periods. To achieve a balance between accuracy and triggering conditions, an exponential moving average (EMA) of the observer estimation error is defined: .

[0087] definition and These are the lower and upper limits of the confidence assessment, respectively. The confidence level reflects the current reliability of the observer.

[0088] Based on the average EMA error of all neighboring channels, the observer confidence is defined as: .

[0089] Define α0 as the basic relative threshold coefficient; μ as the confidence level adjustment gain. The coefficient of the relative term determines the proportion by which the threshold is amplified as the degree of mismatch between nodes increases. The upper bound of the absolute threshold is based on this.

[0090] At this point, the observer confidence level can be used to adaptively adjust the parameters in the triggering condition: , When the confidence level is high, the triggering conditions are relaxed and the communication frequency is reduced; when the confidence level is low, the parameters fall back to the base value and the triggering conditions are tightened, thereby achieving a balance between prediction compensation accuracy and communication frequency.

[0091] The integrated controller design in this embodiment of the invention is as follows: Figure 5 As shown.

[0092] The specific control steps are as follows: Step 1: Each DG measures its local terminal voltage and output current in each sampling period, and at the same time receives information from neighboring DGs through the communication network; Step 2: For the received time-delayed current signal, use the Luenberger state observer to perform state estimation, predict the real current value at the current moment through forward integration, and output the compensation current after global bias correction and low-pass filtering. Step 3: Based on the compensated neighbor current and local voltage information, calculate the composite error and inconsistency, and compare them with the adaptive threshold. If the triggering condition is met, trigger a communication update; otherwise, use the predicted value of the event-triggered internal observer as the neighbor current information. Step 4: Calculate the compensation amount for secondary control based on the neighbor current information of trigger update or prediction hold. Step 5: Superimpose the secondary control compensation amount with the droop control output to generate the corrected voltage reference value; Step 6: The voltage-current dual-loop PI controller generates a duty cycle signal based on the reference voltage to drive the DC / DC converter.

[0093] This invention achieves effective convergence of voltage recovery and current equalization in DC microgrids under communication delay constraints through a control framework that combines active prediction compensation and adaptive event triggering. It also enables the communication frequency to be adaptively adjusted according to the observer confidence level, providing an effective technical solution for distributed control of DC microgrids in environments with communication delays and limited communication resources.

[0094] This invention also provides a collaborative predictive compensation adaptive control system for DC microgrids, applicable to DC microgrids containing multiple distributed generation units (DGs) in islanded operation mode. Each DG is connected in parallel to a common DC bus via a DC / DC converter, and the DG controllers exchange neighbor information through a communication network. The system includes a data acquisition module, a predictive compensation module, an adaptive event triggering module, a collaborative optimization module, and a secondary control output module.

[0095] The data acquisition module, configured in the local controller of each distributed generation unit, measures the local terminal voltage and output current in each sampling period using voltage and current sensors. Simultaneously, it receives current signals from neighboring distributed generation units via a communication network. The received neighbor current signals are historical signals lagging behind the current moment due to time delays in communication link transmission. This module outputs the acquired local electrical quantities and the received neighbor signals to the prediction compensation module and the adaptive event triggering module, respectively.

[0096] The prediction compensation module is used to actively compensate for the neighbor current signal that arrives after communication delay from the data acquisition module. This module integrates a Luenberger state observer and a forward integrator predictor. The Luenberger state observer uses the output current of the local distributed generation unit as the observed state, and is constructed based on a second-order state-space free response model considering the effects of LC filters and load impedance. It uses the delayed neighbor current signal as the observer input to estimate the neighbor current and auxiliary states in real time. The forward integrator predictor uses the estimated state output by the observer as the initial condition, divides the prediction time into multiple sub-steps, and uses a no-input free response model for iterative forward integration to advance the estimated state to the current time, obtaining the current prediction value at the current moment. This module also includes a global deviation correction unit and a low-pass filter. The global deviation correction unit calculates the global mean deviation using the local real-time current mean and the delayed current mean, and adds this deviation to the current value after forward integration prediction to suppress steady-state deviation caused by model mismatch. The low-pass filter performs a first-order low-pass filter on the deviation-corrected signal to suppress high-frequency noise introduced during the prediction process, ultimately outputting the compensated current.

[0097] The adaptive event triggering module determines whether a communication update needs to be triggered based on the compensated neighbor current and local voltage information. This module first calculates the composite error: the difference between the compensated neighbor current and the neighbor current value recorded at the most recent trigger moment is defined as the current deviation, and the difference between the local voltage and the rated voltage is defined as the voltage deviation. The squares of the current deviation and the voltage deviation are weighted and summed to obtain the composite error. The module also calculates the neighbor inconsistency, which is the sum of the squares of the differences between the current output current of the current distributed generation unit and the compensated neighbor currents of each neighboring distributed generation unit. The module internally stores an adaptive threshold and has a dynamic adjustment unit. When the composite error exceeds the adaptive threshold, a preset event triggering condition is met, and the module triggers a communication update, sending the current local information to the neighbors; otherwise, communication is not triggered, and the predicted value from the event triggering observer is used as the neighbor current signal for subsequent control. After triggering, the dynamic adjustment unit increases the adaptive threshold by a preset growth factor and introduces a minimum trigger interval constraint to limit the time interval between two consecutive triggers; during the non-triggering period, the adaptive threshold is slowly decayed by a preset attenuation factor, and a lower threshold is set to maintain sensitivity to system changes.

[0098] The collaborative optimization module automatically balances prediction compensation accuracy with communication frequency. This module calculates the instantaneous estimation error of the event-triggered observer in each neighbor channel in real time, uses an exponential moving average algorithm to obtain the average estimation error for each channel, and then averages the average estimation errors of all neighbor channels. Combined with preset confidence levels, upper and lower limits are evaluated to construct the observer's confidence score. This confidence score reflects the reliability of the current event-triggered observer. The collaborative optimization module uses the confidence score as a multiplicative factor to adjust the relative threshold coefficient of the adaptive threshold in real time: when the confidence score is higher than the preset value, the relative threshold coefficient is increased, relaxing the triggering conditions to reduce the communication frequency; when the confidence score is lower than the preset value, the relative threshold coefficient is decreased, tightening the triggering conditions to increase the communication frequency. Thus, unnecessary communication is reduced when the observer's prediction is accurate, and communication is increased to obtain more accurate neighbor information when prediction reliability decreases, achieving a dynamic balance between compensation accuracy and communication frequency.

[0099] The secondary control output module is used to calculate the compensation amount for distributed secondary control based on the neighbor current signal (triggered update or predicted hold) output by the adaptive event triggering module. This module includes a voltage recovery channel and a current sharing channel: the voltage recovery channel generates the voltage compensation amount based on the integral of the local voltage deviation, and the current sharing channel generates the current compensation amount based on the integral of the neighbor weighted current difference. The outputs of the two channels are superimposed to obtain the secondary control compensation amount. This compensation amount is superimposed with the output of the primary droop control to generate a corrected voltage reference value. This voltage reference value is tracked by a cascaded voltage outer loop-current inner loop dual-loop PI controller. The dual-loop PI controller outputs a duty cycle signal to drive the DC / DC converter of each distributed generation (DG), thereby regulating the local output voltage and output current.

[0100] Each of the above modules is deployed in the local digital controller of each distributed generation unit. Data interaction between modules is completed through the internal bus of the controller, and the DG controllers exchange information through an event-triggered communication network. The entire system forms a closed-loop control loop of "local measurement - time delay compensation - trigger judgment - secondary control - PWM drive", which realizes the restoration of DC bus voltage to rated value and the proportional distribution of output current of each DG under the constraint of communication time delay.

[0101] In summary, the adaptive control method for DC microgrids with collaborative predictive compensation provided by the present invention is suitable for solving problems such as decreased convergence accuracy, difficulty in eliminating steady-state error, and separation design of predictive compensation and event triggering in the distributed secondary control of DC microgrids under communication delay constraints. First, a delay compensation scheme based on a Luenberger state observer combined with forward prediction is designed. The transmitting end state is reconstructed using neighbor current signals arriving after communication delay, and the current value at the current moment is predicted by forward integration using a system dynamic model. A global bias correction mechanism is then used to suppress steady-state bias caused by model mismatch, and low-pass filtering is applied to suppress high-frequency noise. Second, a composite error metric integrating current and voltage biases and an adaptive dynamic threshold triggering mechanism are constructed. Communication updates are performed only when triggering conditions are met; during non-triggering periods, the local observer predicts and holds the neighbor currents to suppress the superposition of delay and event triggering errors. Then, an exponential moving average of the observer estimation error is defined, and an observer confidence level is constructed based on this. The sensitivity of the triggering threshold is adaptively adjusted using confidence feedback to achieve an automatic balance between compensation accuracy and communication frequency. Finally, proactive prediction compensation and adaptive event triggering are designed collaboratively to unify delay compensation and communication resource optimization. This invention can effectively eliminate the impact of communication delay on distributed secondary control, and significantly reduce communication overhead while ensuring the accuracy of voltage recovery and current sharing control, thus providing technical support for distributed control of DC microgrids with communication delay.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A collaborative predictive compensation adaptive control method for DC microgrids, characterized in that, include: Each distributed generation unit measures its local voltage and output current in each sampling period and receives neighbor current signals through a communication network. The received neighbor current signal, which arrives after a communication delay, is evaluated by the Luenberger state observer. After state estimation, molecular step forward integral prediction, global bias correction and low-pass filtering, the compensation current is output. The composite error of fused current deviation and voltage deviation is calculated based on the compensated neighbor current and local voltage, and compared with the adaptive threshold. If the preset event triggering conditions are met, a communication update is triggered; otherwise, the predicted value of the event-triggered observer is used as the neighbor current signal. The secondary control compensation amount is calculated based on the neighbor current signal that is triggered to update or predicted to hold, and is superimposed with the droop control output to generate a corrected voltage reference value. The voltage reference value is used to generate a duty cycle signal via a dual-loop PI controller to drive the DC / DC converter. Specifically, the average EMA error of each neighboring channel is constructed based on the exponential moving average of the estimation error of the event-triggered observer, and the observer confidence is constructed by evaluating the upper and lower limits based on the average EMA error and the preset confidence. The sensitivity of the adaptive threshold is adjusted by feedback using the observer confidence to balance the prediction compensation accuracy and the communication frequency.

2. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The Luenberger state observer takes the output current of the distributed generation unit as the observed state and is constructed based on a second-order state-space free response model that considers the influence of LC filters and load impedance. The neighbor current signal that arrives after communication delay is used as the observer input to estimate the neighbor current and auxiliary state.

3. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The molecular step forward integral prediction includes: dividing the prediction time into multiple sub-steps, using a no-input free response model to perform stepwise iterative forward integration, advancing the estimated state of the Luenberger state observer to the current time, and obtaining the current prediction value.

4. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The global bias correction includes: calculating the global mean bias using the local real-time current mean and the time-delayed current mean, and superimposing the global mean bias onto the current value after forward integration prediction, so as to suppress the steady-state bias caused by model mismatch.

5. The adaptive control method for a DC microgrid with collaborative predictive compensation according to claim 1, characterized in that, The composite error is constructed as follows: the difference between the compensated neighbor current and the neighbor current value recorded at the most recent triggering time is defined as the current deviation, the difference between the local voltage and the rated voltage is defined as the voltage deviation, and the weighted sum of the square of the current deviation and the square of the voltage deviation is taken as the composite error.

6. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The dynamic adjustment rule of the adaptive threshold is as follows: after a trigger occurs, the adaptive threshold is increased by a preset growth coefficient to avoid continuous triggering, and a minimum trigger interval constraint is introduced to limit the minimum time interval between two consecutive triggers; during the non-trigger period, the adaptive threshold is decayed by a preset decay factor, and a lower limit of the threshold is set to maintain sensitivity to system changes.

7. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The event triggering condition is also designed based on neighbor inconsistency; the neighbor inconsistency is the sum of the squares of the differences between the output current of the current distributed generation unit and the compensated neighbor currents of each neighbor distributed generation unit.

8. The adaptive control method for DC microgrids with collaborative predictive compensation according to claim 1, characterized in that, The sensitivity of the adaptive threshold is adjusted by feedback using the observer confidence, including: adjusting the relative threshold coefficient of the adaptive threshold using the observer confidence as a multiplicative factor; increasing the relative threshold coefficient when the observer confidence is higher than a preset value; and decreasing the relative threshold coefficient when the observer confidence is lower than a preset value.

9. A collaborative predictive compensation adaptive control system for a DC microgrid, characterized in that, include: The data acquisition module is used by each distributed generation unit to measure the local voltage and output current in each sampling period, and to receive the neighbor current signal through the communication network; The prediction and compensation module is used to perform state estimation, molecular step forward integration prediction, global bias correction and low-pass filtering on the received neighbor current signal that arrives after communication delay, and output the compensation current. An adaptive event triggering module is used to calculate the composite error of fused current deviation and voltage deviation based on the compensated neighbor current and local voltage, and compare it with an adaptive threshold. If the preset event triggering conditions are met, a communication update is triggered; otherwise, the predicted value of the event-triggered observer is used as the neighbor current signal. The collaborative optimization module is used to construct the average EMA error of each neighbor channel based on the exponential moving average of the estimation error of the event-triggered observer, and to construct the observer confidence based on the upper and lower limits of the average EMA error and the preset confidence level. The sensitivity of the adaptive threshold is adjusted by feedback using the observer confidence level in order to balance the prediction compensation accuracy and communication frequency. The secondary control output module is used to calculate the secondary control compensation amount based on the neighbor current signal that is triggered for update or predicted for hold, and superimpose it with the droop control output to generate a corrected voltage reference value. The voltage reference value is used to generate a duty cycle signal via a dual-loop PI controller to drive the DC / DC converter.

10. The adaptive control system for a DC microgrid with collaborative predictive compensation according to claim 9, characterized in that, In the prediction and compensation module, the Luenberger state observer takes the output current of the distributed generation unit as the observed state and is constructed based on a second-order state-space free response model that considers the influence of LC filters and load impedance. The neighbor current signal that arrives after communication delay is used as the observer input to estimate the neighbor current and auxiliary state. The molecular step forward integral prediction uses a no-input free response model to advance the estimated state to the current time. The global deviation correction is performed by calculating the global mean deviation using the local real-time current mean and the time-delayed current mean.

Images