A method and system for distributed energy storage system centerless collaborative control based on event-triggered consensus algorithm

Abstract

The application provides a kind of based on event trigger consistency algorithm distributed energy storage system centerless collaborative control method and system, the method comprises: constructing centerless distributed architecture, each energy storage unit is interacted with neighborhood information by sparse topology;Establish power response and SOC model, design dynamic event trigger mechanism based on local power deviation, only activate communication when deviation exceeds threshold value;Build local objective function that fuses consistency term and SOC equalization term, solve collaborative control law using distributed event trigger optimization algorithm based on Nesterov acceleration, execute under event trigger, realize total power tracking grid instruction and distribute task according to remaining capacity proportion.This method does not need central server, can realize high reliability collaborative control only by local and neighborhood information, effectively delay system aging, suitable for communication limited rural microgrid scene.

Description

Technical Field

[0001] This invention relates to the field of power system automatic control technology, and in particular to a decentralized collaborative control method and system for distributed energy storage systems based on an event-triggered consensus algorithm. Background Technology

[0002] Existing distributed energy storage systems primarily employ centralized control architectures or periodic distributed consensus algorithms for collaborative control. Centralized control relies on a central controller to aggregate global state information and make unified decisions, which carries the risk of single-point failure; if the central node fails, the entire system will become uncontrollable. Furthermore, it suffers from long communication links, poor real-time performance, and an exponentially increasing computational burden on the central node with the number of energy storage units, limiting scalability. It also fails to consider the issue of unbalanced SOC among energy storage units, leading to overcharging and over-discharging of individual units, accelerating the overall aging of the system. While periodic distributed consensus algorithms eliminate the central node, they employ a fixed-period communication mechanism, forcing units to exchange information at high frequencies regardless of system state changes, resulting in significant waste of communication resources. Simultaneously, they do not incorporate SOC status into power allocation strategies, failing to automatically adjust charging and discharging tasks according to remaining capacity proportions, exacerbating aging differences between units. Moreover, existing methods lack adaptability to scenarios with limited communication bandwidth, making reliable operation difficult in weak communication environments such as rural microgrids, and failing to achieve truly decentralized, highly reliable, and low-communication-overhead collaborative control. Summary of the Invention

[0003] Objective: This invention aims to address the single-point-of-failure risk, excessive communication burden, and accelerated system aging caused by unbalanced State of Charge (SOC) in traditional centralized energy storage control architectures. By deploying a local controller in each energy storage unit, a decentralized distributed control architecture is constructed, utilizing only sparse neighborhood communication to achieve multi-unit collaboration. A dynamic event-triggered mechanism based on local power deviation is designed, remaining silent during non-triggered periods to reduce communication and computational overhead. A local objective function integrating consistency and SOC balancing terms is constructed, and a Nesterov-accelerated distributed event-triggered optimization algorithm is used to solve the collaborative control law. This achieves total power tracking of grid dispatch commands while automatically allocating charging and discharging tasks according to the remaining capacity ratio, effectively delaying overall system aging. This invention improves the reliability, economy, and plug-and-play capability of distributed energy storage systems, reduces dependence on central servers, and promotes the engineering application of decentralized collaborative control in communication-constrained microgrid scenarios.

[0004] This invention provides a decentralized collaborative control method for distributed energy storage systems based on an event-triggered consensus algorithm, comprising the following steps:

[0005] Step 1: Construct a decentralized distributed control architecture, where each energy storage unit interacts with neighboring units through a sparse communication topology.

[0006] Step 2: Establish a joint dynamic model of time-varying parameters for each energy storage unit, taking into account aging characteristics, including a time-varying power response model and a time-varying state of charge (SOC) model.

[0007] Step 3: Design a dynamic event triggering mechanism based on local power deviation to determine the communication triggering time between each energy storage unit and its neighboring units;

[0008] Step 4: Construct a local objective function with three objectives: power distribution consistency, SOC balance, and system aging delay.

[0009] Step 5: Solve the distributed cooperative control law based on the local objective function using a Nesterov-accelerated distributed event-triggered optimization algorithm;

[0010] Step 6: Execute the distributed collaborative control law under the event triggering mechanism to achieve decentralized collaborative control.

[0011] Step 2 includes:

[0012] Based on the number of cycle aging times of the i-th energy storage unit Compared with the current state of charge at time t Establish time-varying rated capacity model, time-varying charge / discharge efficiency model, and time-varying time constant model:

[0013] ,

[0014] in, , These are the SOC deviation penalty coefficient and the temperature deviation penalty coefficient, respectively. The operating temperature of the energy storage unit is collected by a temperature sensor; For reference temperature; Let i be the number of aging cycles for the i-th energy storage unit; , , Let represent the initial rated capacity of the i-th energy storage unit, the initial charge / discharge efficiency of the i-th energy storage unit, and the initial time constant of the i-th energy storage unit, respectively. , , Let represent the time-varying rated capacity of the i-th energy storage unit, the time-varying charge-discharge efficiency of the i-th energy storage unit, and the time-varying time constant of the i-th energy storage unit, respectively. , , Let represent the capacity decay coefficient, efficiency degradation coefficient, and internal resistance growth coefficient of the i-th energy storage unit, respectively.

[0015] Based on time-varying parameters, a time-varying power response model is established:

[0016] ,

[0017] in, It is a time constant. The actual output power collected by the local sensor. The control input is set at time t; Let be the first derivative of the actual output power of the i-th energy storage unit with respect to time;

[0018] Establish a time-varying SOC model that considers the coupling between aging and temperature:

[0019] ,

[0020] in, The state of charge is collected by local sensors. The sign function for charging and discharging direction. The self-discharge rate; and the actual output power. Power Constraint State of charge Constrained by SOC ;

[0021] in, , These represent the lower limit of the power constraint and the upper limit of the power constraint for the i-th energy storage unit, respectively. , These represent the lower limit of the state of charge constraint and the upper limit of the state of charge constraint for the i-th energy storage unit, respectively.

[0022] Step 3 includes:

[0023] Define local power measurement error:

[0024] ,

[0025] in, Due to local power measurement error, The sampling power at the last trigger time;

[0026] Construct dynamic event triggering functions :

[0027] ,

[0028] in, is the rate of change weighting coefficient, representing the sensitivity of sudden changes in control power in the triggering criterion, with a value range of [0.1, 0.5]. The SOC balance weight coefficient represents the sensitivity of controlling SOC deviation in the triggering criterion, and its value ranges from [0.2, 1]. The trigger coefficient represents the overall control of the trigger frequency by adjusting the dynamic threshold, and its value ranges from [0.1, 0.5]. For dynamic thresholds, This is a static threshold. For SOC deviation from equilibrium, This represents the rate of change of local power measurement error.

[0029] when When the current energy storage unit is determined to have entered the triggered state, it broadcasts the actual output power to neighboring units. With state of charge and update control input ;when When the current energy storage unit is determined to enter the hold state, the control input is maintained. constant.

[0030] Step 4 includes:

[0031] The local objective function is:

[0032] ,

[0033] in, For the neighborhood node set, Here, represents the communication weight, indicating the strength of the consistency influence of neighboring unit j on unit i, and its value ranges from [0,1]. For power grid dispatch instructions, is the SOC equilibrium weight coefficient, which controls the priority of the SOC equilibrium objective in the optimization, and its value ranges from [1,3]. To control the smoothing coefficient, suppress sudden changes in control quantity, and reduce converter switching losses, the value range is [0.1, 0.8]. Let be the local objective function of the i-th energy storage unit. Let i be the actual output power of the i-th energy storage unit. The actual output power of neighboring cell j Let i be the rated capacity of the i-th energy storage unit. Let k be the rated capacity of the energy storage unit. This represents the state of charge of the k-th energy storage unit.

[0034] Step 5 includes the following steps:

[0035] Step 5-1: Initialize the actual output power of each energy storage unit. State of charge Neighborhood node set Trigger coefficient With dynamic threshold ;

[0036] Step 5-2: Each energy storage unit collects its current actual output power through local sensors. With state of charge And based on the dynamic event trigger function Determine whether to receive status information broadcast by neighboring units;

[0037] Step 5-3, construct auxiliary points based on the Nesterov momentum extrapolation mechanism:

[0038] ,

[0039] in, Nesterov auxiliary point, This is the momentum accumulation term from the previous iteration. The momentum coefficient;

[0040] Step 5-4, at the Nesterov auxiliary point Calculate the gradient of the local objective function at the given location:

[0041] ,

[0042] in, Let j be a Nesterov auxiliary point of neighborhood unit j, and let j be a set of neighborhood nodes. The unit in the broadcast is obtained at the trigger time; The local objective function of the i-th energy storage unit At Nesterov auxiliary point gradient at;

[0043] Step 5-5, update the momentum accumulation term:

[0044] ,

[0045] in This represents the momentum accumulation term at time t;

[0046] Steps 5-6: Calculate the coupling consistency power correction. :

[0047] ,

[0048] in, This is the consistency gain, used to couple the neighborhood auxiliary point bias to the power correction;

[0049] Steps 5-7: Update the control input based on the power correction and project it to the feasible region:

[0050] ,

[0051] ,

[0052] in, For temporary control input, For projection operators;

[0053] Steps 5-8: Calculate the convergence criterion. If the convergence criterion is met, terminate the iteration and output the distributed cooperative control law. Otherwise, Return to step 5-2.

[0054] In steps 5-8, the convergence criterion is: ,

[0055] ,

[0056] in The convergence threshold is typically set to 0.01. The power allocation convergence threshold is set to 0.05.

[0057] Step 6 includes:

[0058] At the triggering time, when the dynamic event triggers the function At that time, each energy storage unit will solve for the control input. The time constant applied to the power converter based on the power response model Obtain the actual output power response And based on the charge and discharge efficiency in the SOC model With rated capacity Update state of charge:

[0059] ,

[0060] The updated actual output power is then broadcast to neighboring units. With state of charge During non-triggered periods, each energy storage unit maintains its momentum accumulation term. Nesterov auxiliary point and control input The power output remains unchanged, maintained through the power response model, and based on the charge / discharge efficiency in the SOC model. With rated capacity Continuously update local charge status Determine whether the power tracking error and the SOC equalization error meet the system's convergence conditions. If not, trigger the optimization algorithm to solve the problem again.

[0061] The present invention also provides a decentralized collaborative control system for a distributed energy storage system based on an event-triggered consensus algorithm implemented according to the method, comprising:

[0062] The power grid dispatching and communication interface module includes a power grid dispatching center server, a remote communication interface, and a dispatching instruction parsing unit, which is used to generate and issue power grid dispatching instructions;

[0063] A distributed energy storage unit cluster includes multiple energy storage units. Each energy storage unit includes an energy storage body, a DC / AC power converter, a voltage sensor, a current sensor, a temperature sensor, and a SOC estimation unit, which are used to perform charging and discharging and collect status information.

[0064] The local controller group includes two or more local controllers, each of which is deployed in an energy storage unit. It includes an event triggering and determination unit, a consistency coordination unit, a Nesterov accelerated optimization solution unit, a PWM generation unit, and an embedded processor, which are used to solve and execute the distributed cooperative control law.

[0065] Sparse communication networks, including CAN bus communication interfaces, Ethernet communication interfaces, and neighborhood communication links, are used to connect adjacent local controllers to achieve decentralized information exchange.

[0066] The voltage sensor, current sensor, temperature sensor, and SOC estimation unit are connected to the local controller to provide feedback on the actual output power and state of charge; the PWM generation unit is connected to the DC / AC power converter to output PWM switching signals; and the local controller exchanges Nesterov auxiliary points through the neighborhood communication link.

[0067] The present invention also provides an electronic device, including a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method.

[0068] The present invention also provides a storage medium storing a computer program or instructions that, when the computer program or instructions are run on a computer, execute the steps of the method described.

[0069] Beneficial effects: 1. High power point tracking accuracy and fast convergence speed: Within a 24-hour scheduling cycle, the average tracking error between the total output power and the grid dispatch command is only 44.04kW, and the final value converges to 26.00kW. Compared with the control group without Nesterov acceleration, the average tracking error is reduced by 26.1%, and the final value is reduced by 59.9%. The Nesterov momentum extrapolation mechanism effectively suppresses the oscillation phenomenon in distributed optimization by introducing exponentially weighted accumulation of historical gradient information, and still achieves fast convergence under sparse communication conditions.

[0070] 2. Significant SOC balancing effect, effectively delaying system aging: The initial SOC standard deviation of the 10 heterogeneous energy storage units was 0.1845, and all converged to below 0.05 within 3.5 hours, achieving a convergence rate of 100%. During the steady-state phase, the deviation between the power allocation ratio of each unit and the rated capacity ratio was controlled within the range of 0.0019~0.0167, and the coefficient of variation of the cumulative charge and discharge capacity over 24 hours was only 4.4%. It automatically realized the allocation of charge and discharge tasks according to the remaining capacity ratio, effectively balancing the aging rate of each unit and avoiding the overall system lifespan degradation caused by overcharging and over-discharging of individual units.

[0071] 3. Significantly reduced communication burden, suitable for weak communication environments: The dynamic event triggering mechanism reduces the number of communications within 24 hours from 950 times of periodic communication to 437 times, reducing the communication burden by 54.0%. Each unit only triggers communication when the power deviation exceeds the adaptive threshold, remaining silent during other times. This significantly reduces communication resource consumption while ensuring control accuracy, making it particularly suitable for scenarios with limited communication bandwidth, such as rural microgrids.

[0072] 4. High reliability and strong plug-and-play capability of the decentralized architecture: The system abandons the traditional central controller, and each energy storage unit only achieves coordination through sparse neighborhood communication, eliminating the risk of single point of failure. Under the condition of significant initial SOC differences (0.35~0.90), the 10 heterogeneous units can autonomously achieve power point tracking and SOC equalization without the coordination of a central node, verifying the robust adaptability of the proposed decentralized architecture to unit heterogeneity and topology changes. Attached Figure Description

[0073] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0074] Figure 1 This is a schematic diagram of the structure of the present invention.

[0075] Figure 2 This is a flowchart of the algorithm used in this invention.

[0076] Figure 3 This is a graph showing power point tracking and SOC equalization.

[0077] Figure 4 A graph illustrating the sparsity of event-triggered communication.

[0078] Figure 5 The graph shows a comparison of the convergence performance of the three methods.

[0079] Figure 6 This is a capacity-weighted power allocation and aging balance diagram. Detailed Implementation

[0080] This embodiment provides a decentralized collaborative control method for a distributed energy storage system based on an event-triggered consensus algorithm, including:

[0081] A decentralized distributed control architecture is constructed, in which each energy storage unit interacts with neighboring units through a sparse communication topology.

[0082] Establish a joint dynamic model of time-varying parameters for each energy storage unit, taking into account aging characteristics, including a time-varying power response model and a time-varying state of charge (SOC) model.

[0083] Design a dynamic event triggering mechanism based on local power deviation to determine the communication triggering time between each energy storage unit and neighboring units;

[0084] Construct a local objective function with three objectives: power distribution consistency, SOC balance, and system aging delay.

[0085] The distributed cooperative control law is solved based on the local objective function using a Nesterov-accelerated distributed event-triggered optimization algorithm.

[0086] The distributed collaborative control law is executed under the event-triggered mechanism to achieve decentralized collaborative control.

[0087] Furthermore, the power response model and state of charge (SOC) model for each energy storage unit are established, including:

[0088] Based on the number of cycle aging times of each energy storage unit With current state of charge Establish time-varying rated capacity model, time-varying charge / discharge efficiency model, and time-varying time constant model:

[0089] ,

[0090] in, , , These are the time-varying rated capacity model, the time-varying charge-discharge efficiency model, and the time-varying time constant model, respectively. , , These are the initial rated capacity, initial charge / discharge efficiency, and initial time constant, respectively. , , These are the capacity decay coefficient, efficiency degradation coefficient, and internal resistance growth coefficient, respectively. , These are the SOC deviation penalty coefficient and the temperature deviation penalty coefficient, respectively. The current state of charge; The operating temperature of the energy storage unit is collected by the temperature sensor. For reference temperature; The number of aging cycles for each energy storage unit;

[0091] Based on time-varying parameters, a time-varying power response model is established:

[0092] ,

[0093] in, It is a time constant. The actual output power collected by the local sensor. For control input;

[0094] Establish a time-varying SOC model that considers the coupling between aging and temperature:

[0095] ,

[0096] in, The state of charge is collected by local sensors. The sign function for charging and discharging direction. The self-discharge rate; and the actual output power. Power Constraint State of charge Constrained by SOC ;

[0097] Furthermore, the design of a dynamic event triggering mechanism based on local power deviation includes:

[0098] Define local power measurement error:

[0099] ,

[0100] in, Due to local power measurement error, This represents the sampling power at the last trigger moment.

[0101] Construct dynamic event triggering functions:

[0102] ,

[0103] in, This is a function that triggers dynamic events. is the rate of change weighting coefficient, representing the sensitivity of sudden changes in control power in the triggering criterion, with a value range of [0.1, 0.5]. The SOC balance weight coefficient represents the sensitivity of controlling SOC deviation in the triggering criterion, and its value ranges from [0.2, 1]. The trigger coefficient represents the overall control of the trigger frequency by adjusting the dynamic threshold, and its value ranges from [0.1, 0.5]. For dynamic thresholds, This is a static threshold. For SOC deviation from equilibrium, This represents the rate of change of local power measurement error.

[0104] when When the current energy storage unit is determined to have entered the triggered state, it broadcasts the actual output power to neighboring units. With state of charge and update control input ;when When the current energy storage unit is determined to enter the hold state, the control input is maintained. constant.

[0105] Furthermore, constructing the local objective function includes:

[0106] The local objective function is:

[0107] ,

[0108] in, For the neighborhood node set, Here, represents the communication weight, indicating the strength of the consistency influence of neighboring unit j on unit i, and its value ranges from [0,1]. For power grid dispatch instructions, is the SOC equilibrium weight coefficient, which controls the priority of the SOC equilibrium objective in the optimization, and its value ranges from [1,3]. To control the smoothness coefficient, suppress sudden changes in control quantity, and reduce converter switching losses, the value range is [0.1, 0.8].

[0109] Furthermore, based on the local objective function, the distributed cooperative control law is solved using a Nesterov-accelerated distributed event-triggered optimization algorithm, including the following steps:

[0110] 1) Initialize the actual output power of each energy storage unit State of charge Neighborhood node set Trigger coefficient With dynamic threshold .

[0111] 2) Each energy storage unit collects its current actual output power through local sensors. With state of charge And based on the dynamic event trigger function Determine whether to receive status information broadcast by neighboring units;

[0112] 3) Constructing auxiliary points based on the Nesterov momentum extrapolation mechanism:

[0113] ,

[0114] in, The Nesterov auxiliary point, This is the momentum accumulation term from the previous iteration. The momentum coefficient;

[0115] 4) At the auxiliary point Calculate the gradient of the local objective function at the given location:

[0116] ,

[0117] in, For the Nesterov auxiliary point of neighborhood unit j, the set of neighborhood nodes is... The unit in the broadcast is obtained at the trigger time;

[0118] 5) Update the momentum accumulation term:

[0119] ;

[0120] 6) Calculate the coupling consistency power correction:

[0121] ,

[0122] in, This is the consistency gain, used to couple the neighborhood auxiliary point bias to the power correction;

[0123] 7) Update the control input based on the power correction amount and project it to the feasible region:

[0124] ,

[0125] ,

[0126] in, For temporary control input, This is the projection operator.

[0127] 8) Calculate the convergence criterion:

[0128] ,

[0129] ,

[0130] If satisfied, the iteration terminates and the distributed cooperative control law is output. Otherwise, , return 2).

[0131] Furthermore, executing the distributed cooperative control law under the event-triggered mechanism includes:

[0132] At the triggering time, when the dynamic event triggers the function At that time, each energy storage unit will solve for the control input. The time constant applied to the power converter based on the power response model Obtain the actual output power response And based on the charge and discharge efficiency in the SOC model With rated capacity Update state of charge:

[0133] ,

[0134] The updated actual output power is then broadcast to neighboring units. With state of charge During non-triggered periods, each energy storage unit maintains its momentum accumulation term. Nesterov auxiliary point and control input The power output remains unchanged, maintained through the power response model, and based on the charge / discharge efficiency in the SOC model. With rated capacity Continuously update local charge status Determine whether the power tracking error and the SOC equalization error meet the system's convergence conditions. If not, trigger the optimization algorithm to solve the problem again.

[0135] Furthermore, the decentralized distributed control architecture includes:

[0136] The power grid dispatching and communication interface module includes a power grid dispatching center server, a remote communication interface, and a dispatching instruction parsing unit, which is used to generate and issue power grid dispatching instructions;

[0137] A distributed energy storage unit cluster includes multiple energy storage units. Each energy storage unit includes an energy storage body, a DC / AC power converter, a voltage sensor, a current sensor, a temperature sensor, and a SOC estimation unit, which are used to perform charging and discharging and collect status information.

[0138] The local controller group includes two or more local controllers, each of which is deployed in an energy storage unit. It includes an event triggering and determination unit, a consistency coordination unit, a Nesterov accelerated optimization solution unit, a PWM generation unit, and an embedded processor, which are used to solve and execute the distributed cooperative control law.

[0139] Sparse communication networks, including CAN bus communication interfaces, Ethernet communication interfaces, and neighborhood communication links, are used to connect adjacent local controllers to achieve decentralized information exchange.

[0140] The voltage sensor, current sensor, temperature sensor, and SOC estimation unit are connected to the local controller to provide feedback on the actual output power and state of charge; the PWM generation unit is connected to the DC / AC power converter to output PWM switching signals; and the local controller exchanges Nesterov auxiliary points through the neighborhood communication link.

[0141] Figure 1 The schematic diagram of the present invention illustrates the four main modules of the decentralized distributed control architecture: power grid dispatch and communication interface module, distributed energy storage unit group, local controller group and sparse communication network. Each module is connected by signal lines, reflecting the topological characteristics of having no central node and only neighborhood communication.

[0142] Ten heterogeneous lithium iron phosphate energy storage units are deployed in a rural microgrid. Each unit has a different rated capacity and an initial state of charge (SOC) ranging from 0.35 to 0.90, with a standard deviation of 0.1845. Each unit is equipped with an energy storage device, a DC / AC power converter, a voltage sensor, a current sensor, a temperature sensor, and an SOC estimation unit. The local controller uses an STM32F407 processor. Adjacent units establish sparse communication links via a CAN bus, forming a [missing information - likely a specific network structure]. Figure 1 The diagram shows a decentralized architecture.

[0143] During system initialization, the parameters of each local controller are set as follows: Take 0.0002, Take 0.0001, Take 0.0003; Take 0.5, Take 0.2; Set the value to 0.001. Event trigger parameter settings: Take 0.3, Take 0.6, Take 0.2, Set the value to 0.05. Optimize parameter settings: Take 2.0, Take 0.3.

[0144] During operation, each unit samples its local actual output power and state of charge at a 1-second sampling period, and determines whether to enter the triggered state based on a dynamic event triggering function. Upon triggering, the state is broadcast to the neighborhood and the control input is updated; in the non-triggered state, the control law remains unchanged from the previous moment. After Nesterov acceleration optimization, the control law is applied to the power converter.

[0145] 24-hour operation results are as follows Figures 3-6 As shown. Figure 3This indicates that the total output power is highly consistent with the grid's dual-peak load command, with an average tracking error of 44.04kW and a final value that converges to 26.00kW; the SOC of each unit converges from an initial standard deviation of 0.1845 to below 0.05 within 3.5 hours, with an equilibrium convergence rate of 100%. Figure 4 The results show that the dynamic event triggering mechanism reduces the number of communications within 24 hours from 950 times in traditional periodic communications to 437 times, a reduction of 54.0%. The triggering times of each unit are not uniformly distributed, and are only densely triggered at the load peak-valley switching points. Figure 5 The comparison results show that the power tracking error of this method is reduced by 26.1% compared with the control group without Nesterov acceleration, the final value is reduced by 59.9%, and the soc convergence speed is improved by 28.6%. Figure 6 The deviation between the power allocation ratio of each unit and the rated capacity ratio during the steady-state phase was verified to be controlled within 0.0019~0.0167, and the coefficient of variation of the cumulative charge and discharge in 24 hours was only 4.4%, automatically realizing the allocation of tasks according to the remaining capacity ratio.

[0146] This invention provides a decentralized collaborative control method and system for distributed energy storage systems based on an event-triggered consensus algorithm. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A decentralized collaborative control method for a distributed energy storage system based on an event-triggered consensus algorithm, characterized in that, Includes the following steps: Step 1: Construct a decentralized distributed control architecture, where each energy storage unit interacts with neighboring units through a sparse communication topology. Step 2: Establish a joint dynamic model of time-varying parameters for each energy storage unit, taking into account aging characteristics, including a time-varying power response model and a time-varying state of charge (SOC) model. Step 3: Design a dynamic event triggering mechanism based on local power deviation to determine the communication triggering time between each energy storage unit and its neighboring units; Step 4: Construct a local objective function with three objectives: power distribution consistency, SOC balance, and system aging delay. Step 5: Solve the distributed cooperative control law based on the local objective function using a Nesterov-accelerated distributed event-triggered optimization algorithm; Step 6: Execute the distributed collaborative control law under the event triggering mechanism to achieve decentralized collaborative control.

2. The method according to claim 1, characterized in that, Step 2 includes: Based on the number of cycle aging times of the i-th energy storage unit Compared with the current state of charge at time t Establish time-varying rated capacity model, time-varying charge / discharge efficiency model, and time-varying time constant model: ， in, , These are the SOC deviation penalty coefficient and the temperature deviation penalty coefficient, respectively. The operating temperature of the energy storage unit is collected by a temperature sensor; For reference temperature; Let be the number of aging cycles for the i-th energy storage unit; , , Let represent the initial rated capacity of the i-th energy storage unit, the initial charge / discharge efficiency of the i-th energy storage unit, and the initial time constant of the i-th energy storage unit, respectively. , , Let represent the time-varying rated capacity of the i-th energy storage unit, the time-varying charge-discharge efficiency of the i-th energy storage unit, and the time-varying time constant of the i-th energy storage unit, respectively. , , Let represent the capacity decay coefficient, efficiency degradation coefficient, and internal resistance growth coefficient of the i-th energy storage unit, respectively. Based on time-varying parameters, a time-varying power response model is established: ， in, It is a time constant. The actual output power collected by the local sensor. The control input is set at time t; Let be the first derivative of the actual output power of the i-th energy storage unit with respect to time; Establish a time-varying SOC model that considers the coupling between aging and temperature: ， in, The state of charge is collected by local sensors. The sign function for charging and discharging direction. The self-discharge rate; and the actual output power. Power Constraint State of charge Constrained by SOC ; in, , These represent the lower limit of the power constraint and the upper limit of the power constraint for the i-th energy storage unit, respectively. , These represent the lower limit of the state of charge constraint and the upper limit of the state of charge constraint for the i-th energy storage unit, respectively.

3. The method according to claim 2, characterized in that, Step 3 includes: Define local power measurement error: ， in, Due to local power measurement error, The sampling power at the last trigger time; Construct dynamic event triggering functions : ， in, The rate of change weighting coefficient. The SOC equilibrium weight coefficient, For trigger coefficient, For dynamic thresholds, This is a static threshold. For SOC deviation from equilibrium, This represents the rate of change of local power measurement error. when When the current energy storage unit is determined to have entered the triggered state, it broadcasts the actual output power to neighboring units. With state of charge and update control input ;when When the current energy storage unit is determined to enter the hold state, the control input is maintained. constant.

4. The method according to claim 3, characterized in that, Step 4 includes: The local objective function is: ， in, For the neighborhood node set, For communication weights, For power grid dispatch instructions, The SOC equilibrium weight coefficient, To control the smoothing coefficient; Let be the local objective function of the i-th energy storage unit. Let i be the actual output power of the i-th energy storage unit. The actual output power of neighboring cell j Let i be the rated capacity of the i-th energy storage unit. Let k be the rated capacity of the energy storage unit. This represents the state of charge of the k-th energy storage unit.

5. The method according to claim 4, characterized in that, Step 5 includes the following steps: Step 5-1: Initialize the actual output power of each energy storage unit. State of charge Neighborhood node set Trigger coefficient With dynamic threshold ; Step 5-2: Each energy storage unit collects its current actual output power through local sensors. With state of charge And based on the dynamic event trigger function Determine whether to receive status information broadcast by neighboring units; Step 5-3, construct auxiliary points based on the Nesterov momentum extrapolation mechanism: ， in, Nesterov auxiliary point, This is the momentum accumulation term from the previous iteration. The momentum coefficient; Step 5-4, at the Nesterov auxiliary point Calculate the gradient of the local objective function at the given location: ， in, Let j be a Nesterov auxiliary point of neighborhood unit j, and let j be a set of neighborhood nodes. The unit in the middle is broadcast at the trigger time; The local objective function for the i-th energy storage unit At Nesterov auxiliary point gradient at; Step 5-5, update the momentum accumulation term: ， in This represents the momentum accumulation term at time t; Steps 5-6: Calculate the coupling consistency power correction. : ， in, For consistency gain; Steps 5-7: Update the control input based on the power correction and project it to the feasible region: ， ， in, For temporary control input, For projection operators; Steps 5-8: Calculate the convergence criterion. If the convergence criterion is met, terminate the iteration and output the distributed cooperative control law. Otherwise, Return to step 5-2.

6. The method according to claim 5, characterized in that, In steps 5-8, the convergence criterion is: , ， in This is the convergence threshold.

7. The method according to claim 6, characterized in that, Step 6 includes: At the triggering time, when the dynamic event triggers the function At that time, each energy storage unit will solve for the control input. The time constant applied to the power converter based on the power response model Obtain the actual output power response And based on the charge and discharge efficiency in the SOC model With rated capacity Update state of charge: ， The updated actual output power is then broadcast to neighboring units. With state of charge During non-triggered periods, each energy storage unit maintains its momentum accumulation term. Nesterov auxiliary point and control input The power output remains unchanged, maintained through the power response model, and based on the charge / discharge efficiency in the SOC model. With rated capacity Continuously update local charge status Determine whether the power tracking error and the SOC equalization error meet the system's convergence conditions. If not, trigger the optimization algorithm to solve the problem again.

8. A decentralized collaborative control system for a distributed energy storage system based on an event-triggered consensus algorithm, implemented according to any one of claims 1 to 7, characterized in that, include: The power grid dispatching and communication interface module includes a power grid dispatching center server, a remote communication interface, and a dispatching instruction parsing unit, which is used to generate and issue power grid dispatching instructions; A distributed energy storage unit cluster includes multiple energy storage units. Each energy storage unit includes an energy storage body, a DC / AC power converter, a voltage sensor, a current sensor, a temperature sensor, and a SOC estimation unit, which are used to perform charging and discharging and collect status information. The local controller group includes two or more local controllers, each of which is deployed in an energy storage unit. It includes an event triggering and determination unit, a consistency coordination unit, a Nesterov accelerated optimization solution unit, a PWM generation unit, and an embedded processor, which are used to solve and execute the distributed cooperative control law. Sparse communication networks, including CAN bus communication interfaces, Ethernet communication interfaces, and neighborhood communication links, are used to connect adjacent local controllers to achieve decentralized information exchange. The voltage sensor, current sensor, temperature sensor, and SOC estimation unit are connected to the local controller to provide feedback on the actual output power and state of charge; the PWM generation unit is connected to the DC / AC power converter to output PWM switching signals; and the local controller exchanges Nesterov auxiliary points through the neighborhood communication link.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, It stores a computer program or instructions that, when run on a computer, perform the steps of the method as described in any one of claims 1 to 7.

Images