Energy storage system distributed collaborative control and optimization method and system and storage medium

Abstract

This disclosure relates to the field of power system energy storage technology, and provides a distributed collaborative control and optimization method, system, and storage medium for energy storage systems. The method includes: constructing a distributed communication network, treating each energy storage unit as an intelligent node, with each node interacting with neighboring nodes through the distributed communication network; for each energy storage unit, based on real-time collected operating data, evaluating at least one of the following: state of charge, health, and safety, to obtain a state evaluation result; each energy storage unit, based on its own state evaluation result, combined with information obtained from neighboring nodes and the total power demand of the system, performing collaborative calculations through a distributed optimization algorithm, aiming to optimize the overall system operating indicators, solving for and executing the charging and discharging power commands of each energy storage unit; each energy storage unit dynamically adjusting control parameters according to the deviation between the actual operating state and the expected target, so that the system continuously maintains the optimal operating state and achieves global optimal control.

Description

Technical Field

[0001] This disclosure relates to the field of power system energy storage technology, specifically to a distributed collaborative control and optimization method, system, and storage medium for energy storage systems. Background Technology

[0002] With the large-scale development and grid connection of new energy sources (wind power, solar power, etc.), the intermittency, volatility, and randomness of their output have brought enormous challenges to the safe and stable operation of the power system. Energy storage systems, as a key technology for smoothing out the fluctuations of new energy sources and improving the flexibility of the power grid, have been rapidly developed.

[0003] In related technologies, the control methods of energy storage systems are mainly divided into centralized control and distributed control. Centralized control uses a central controller to collect the operating information of all energy storage units and formulate control strategies to distribute to each energy storage unit. Although it has a simple structure and high control accuracy, it has obvious drawbacks such as low reliability, information transmission delay, and significantly increased bandwidth pressure. Distributed control, on the other hand, has low information interaction efficiency between multiple energy storage units during collaborative operation, making it difficult to achieve globally optimal control results. Summary of the Invention

[0004] The purpose of this disclosure is to provide a method, system, and storage medium for distributed collaborative control and optimization of energy storage systems to solve problems in related technologies.

[0005] This disclosure provides a distributed collaborative control and optimization method for an energy storage system, applicable to a distributed energy storage system comprising multiple energy storage units. The method includes: A distributed communication network is constructed, with each energy storage unit acting as an intelligent node, and each node interacting with its neighboring nodes through the distributed communication network. For each energy storage unit, based on real-time collected operating data, at least one of the following is assessed: state of charge, health status, and safety status, to obtain a status assessment result; Based on its own state assessment results, combined with information obtained from adjacent nodes and the total power demand of the system, each energy storage unit performs collaborative calculations through a distributed optimization algorithm to solve and execute the charging and discharging power commands of each energy storage unit with the goal of optimizing the overall system operation indicators. Each energy storage unit dynamically adjusts its control parameters based on the deviation between its actual operating status and the expected target.

[0006] Optionally, each node interacts with neighboring nodes through the distributed communication network, including: Each agent node sends its own state assessment results and local power reference values ​​to its neighboring nodes; Each agent node receives the state assessment results and local power reference values ​​sent by its neighboring nodes.

[0007] Optionally, the distributed communication network is a mesh network without a central node and is configured with a communication redundancy mechanism; the communication redundancy mechanism includes: Configure at least two physical or logical links using different communication protocols for each agent node; Real-time monitoring of the communication quality of the primary link; When the communication quality of the primary link is lower than a preset threshold, the communication path is automatically switched to the backup link.

[0008] Optionally, the state of charge is estimated based on voltage and current data using a Kalman filter algorithm combined with the ampere-hour integral method; The health status is evaluated based on at least one of the following: capacity decay rate, internal resistance change rate, and historical cycle life data, combined with a neural network algorithm. The safety status is determined based on whether the temperature data exceeds the limit and whether there are abnormal fluctuations in the voltage and current data.

[0009] Optionally, a distributed optimization algorithm is used for collaborative computation to optimize the overall system performance indicators, solving for the charging and discharging power commands of each energy storage unit, including: An objective function is established with the optimization objectives of minimizing system operating costs, minimizing energy storage unit lifespan loss, and meeting the total system power demand. The charging and discharging power capacity and state of charge range of each energy storage unit are used as constraints. The objective function is iteratively solved using the alternating direction multiplier method, consensus algorithm, or particle swarm optimization algorithm to obtain the charging and discharging power commands for each energy storage unit.

[0010] Optionally, the system operating cost includes charging and discharging loss cost, maintenance cost, and capacity decay cost; the lifespan loss of the energy storage unit is quantified by the fluctuation range of its state of charge and the charging and discharging rate.

[0011] Optionally, the dynamic adjustment control parameters include: Monitor the deviation between the actual charging and discharging power and changes in state of charge and the corresponding expected target values; When the absolute value or duration of the deviation exceeds a set threshold, parameter adjustment is triggered. By applying reinforcement learning algorithms, the weight coefficients or convergence factors of the distributed optimization algorithm are adjusted based on historical running data and the current deviation.

[0012] Optionally, the method further includes fault diagnosis and fault-tolerant control, specifically including: Based on the aforementioned operating data and communication status, diagnose energy storage unit faults or communication link faults. If the fault is due to abnormal sensor data, the abnormal data will be corrected by a multi-source data fusion algorithm, and the power allocation weight of the corresponding energy storage unit will be reduced. If a unit experiences a severe failure or communication interruption, the faulty unit is isolated from the collaborative computing network, and the collaborative computing steps are re-executed to assign new charge / discharge power commands to the remaining normal units.

[0013] This disclosure also provides a distributed energy storage system, including: Multiple energy storage units, A distributed communication network connects multiple energy storage units that act as intelligent agent nodes, enabling information exchange between the nodes. Each of the aforementioned intelligent agent nodes is configured as follows: Based on the real-time operating data of the corresponding energy storage unit, at least one of its state of charge, health status and safety status is evaluated to obtain the status assessment result; Based on its own state assessment results, combined with information obtained from neighboring nodes and the total power demand of the system, a distributed optimization algorithm is used for collaborative calculation to optimize the overall system operation indicators, and to solve and execute the charging and discharging power commands of each energy storage unit. The control parameters are dynamically adjusted based on the deviation between the actual operating status of the corresponding energy storage unit and the expected target.

[0014] This disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described distributed collaborative control and optimization method for energy storage systems.

[0015] Through the above technical solutions, a distributed communication network based on multiple agent nodes is constructed. Each node only interacts with its neighboring nodes, which not only completely avoids the risk of system paralysis caused by single-point failures in centralized control architectures and significantly improves system reliability, but also greatly reduces the bandwidth pressure and latency of information transmission, enhancing the system's dynamic response capability. By conducting real-time assessments of each energy storage unit, covering its state of charge, health, and safety, and using the state assessment results as the core decision-making basis in subsequent collaborative calculations, the control strategy can strictly follow the state constraints of each unit, thereby effectively preventing damaging behaviors such as overcharging and over-discharging, and achieving proactive protection of the energy storage unit's lifespan. At the same time, a distributed optimization algorithm is used to integrate the system's total power demand, neighboring node information, and local state, and solves the problem with the overall system operation indicators as the optimization objective, thereby achieving a balance between global economic optimization and lifespan optimization while meeting power demand. In addition, a dynamic parameter adjustment mechanism based on actual operating deviations is introduced, enabling the system to adaptively respond to new energy fluctuations and load changes, continuously maintain optimal operating state, and achieve globally optimal control.

[0016] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a distributed collaborative control and optimization method for an energy storage system according to an exemplary embodiment.

[0018] Figure 2 This is an architecture diagram illustrating distributed collaborative control and optimization of an energy storage system according to an exemplary embodiment.

[0019] Figure 3 This is a schematic diagram illustrating information interaction in a hierarchical distributed collaborative control system according to an exemplary embodiment.

[0020] Figure 4 This is a flowchart illustrating the sub-steps of each node interacting with its neighboring nodes through a distributed communication network, according to an exemplary embodiment.

[0021] Figure 5 This is a flowchart illustrating a sub-step of step S3 according to an exemplary embodiment.

[0022] Figure 6 This is a flowchart illustrating a sub-step of step S4 according to an exemplary embodiment. Detailed Implementation

[0023] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0024] First, let's explain the following technical terms: Multi-Agent: A system composed of multiple intelligent units (agents) with autonomous decision-making capabilities and the ability to interact with each other. In this scheme, each energy storage unit can be regarded as an intelligent agent, capable of independently collecting its own state data, executing basic control commands, and collaborating with other energy storage units (intelligent agents) through a communication network to achieve global control objectives, without relying on a single central controller.

[0025] Distributed communication network: In contrast to centralized communication network, it refers to a communication architecture without a fixed central node, where all nodes (such as energy storage units and control modules) have equal status. Nodes exchange information through direct or indirect links. Even if some nodes or links fail, the remaining nodes can still maintain communication through other paths, improving the system's fault tolerance.

[0026] Communication redundancy mechanism: A backup strategy designed to avoid communication interruption caused by the failure of a single communication link or device. It can be a dual-link configuration of primary communication + backup communication (such as Ethernet). When the primary link experiences delays, packet loss or other problems, the system can automatically switch to the backup link to ensure the continuity of information transmission.

[0027] State of Charge (SOC): This describes the percentage of energy remaining in an energy storage unit (such as a battery or flywheel) relative to its rated capacity. It is a core parameter for energy storage system control and directly determines the charging and discharging strategy.

[0028] State of Health (SOH): Reflects the difference between the current performance state of an energy storage unit and its initial state, and is usually measured by indicators such as capacity decay rate and the degree of decline in charge and discharge efficiency.

[0029] Hierarchical Distributed Cooperative Control Strategy: A composite strategy combining hierarchical and distributed control. It consists of two layers: a local control layer (corresponding to a single energy storage unit), which performs basic charging and discharging control based on its own SOC, SOH, and other states; and a cooperative control layer (associating all energy storage units), which optimizes global power allocation through algorithms to ensure the safety of individual units while achieving optimal overall system efficiency.

[0030] With the large-scale development and grid connection of new energy sources (wind power, solar power, etc.), the intermittency, volatility, and randomness of their output have brought enormous challenges to the safe and stable operation of the power system. Energy storage systems, as a key technology for smoothing out the fluctuations of new energy sources and improving the flexibility of the power grid, have been rapidly developed.

[0031] In related technologies, the control methods of energy storage systems are mainly divided into centralized control and distributed control. Centralized control uses a central controller to collect the operating information of all energy storage units and formulate control strategies to distribute to each energy storage unit. Although it has a simple structure and high control accuracy, it has obvious drawbacks such as low reliability, information transmission delay, and significantly increased bandwidth pressure. Distributed control, on the other hand, has low information interaction efficiency between multiple energy storage units during collaborative operation, making it difficult to achieve globally optimal control results.

[0032] The inventors discovered that centralized control, where a central controller collects operational information from all energy storage units and formulates control strategies to distribute to each unit, has significant drawbacks. While this control method is simple in structure and offers high precision, it suffers from several issues: First, the central controller must process massive amounts of information; a failure in the central controller would paralyze the entire energy storage system, resulting in low system reliability. Second, with the increasing number of energy storage units and their geographical dispersion, information transmission delays and bandwidth pressures increase significantly, causing control commands to fail to execute in a timely manner and impacting the system's dynamic response performance.

[0033] While existing distributed control methods have addressed the reliability and information transmission issues of centralized control to some extent, they still have many shortcomings. Some distributed control methods only consider the power allocation optimization of the energy storage system, neglecting the state constraints of the energy storage units, such as state of charge (SOC), charging and discharging power limits, and temperature limits. Long-term operation can easily lead to overcharging and over-discharging of the energy storage units, shortening their service life. Other methods do not fully consider the operational economics of the energy storage system, failing to effectively control the operating costs of the energy storage system, such as charging and discharging losses, maintenance costs, and capacity decay costs. At the same time, in the process of multiple energy storage units operating collaboratively, the information exchange between the units is not efficient enough, making it difficult to achieve globally optimal control results.

[0034] To address the aforementioned issues, a distributed communication network based on multi-agent nodes is constructed. Each node interacts only with its neighboring nodes, completely eliminating the risk of system paralysis due to single-point failures in centralized control architectures and significantly improving system reliability. Furthermore, it drastically reduces bandwidth pressure and latency in information transmission, enhancing the system's dynamic response capabilities. Real-time assessments covering the state of charge, health, and safety status of each energy storage unit are conducted, and the results are used as the core decision-making basis in subsequent collaborative computations. This ensures that the control strategy strictly adheres to the state constraints of each unit, effectively preventing damaging behaviors such as overcharging and over-discharging, and achieving proactive protection of the energy storage unit's lifespan. Simultaneously, a distributed optimization algorithm integrates the system's total power demand, neighboring node information, and local state, using the overall system operating indicators as the optimization objective. This achieves a balance between global economic optimization and lifespan optimization while meeting power requirements. In addition, a dynamic parameter adjustment mechanism based on actual operating deviations is introduced, enabling the system to adaptively respond to new energy fluctuations and load changes, continuously maintaining optimal operating conditions and achieving globally optimal control.

[0035] Figure 1 This is a flowchart illustrating a distributed collaborative control and optimization method for an energy storage system according to an exemplary embodiment. This method can be applied to distributed energy storage systems comprising multiple energy storage units. Please refer to [link to relevant documentation]. Figure 2 For example, distributed energy storage systems that include various types of energy storage units (such as lithium battery energy storage, vanadium redox flow battery energy storage, flywheel energy storage, etc.) can be widely used in scenarios such as new energy microgrids, distribution network frequency regulation and peak shaving, and user-side energy storage. The aim is to improve the operating efficiency, reliability and economy of energy storage systems and promote the efficient consumption and utilization of new energy.

[0036] Please see Figure 1 The distributed collaborative control and optimization method for the energy storage system may include steps S1 to S4.

[0037] Step S1: Construct a distributed communication network, treating each energy storage unit as an intelligent agent node, with each node interacting with neighboring nodes through the distributed communication network.

[0038] A distributed communication network based on multiple agents is constructed, with each energy storage unit acting as an agent node. These agent nodes interact with neighboring nodes via wireless communication (such as LoRa, 5G, WiFi, etc.) or wired communication (such as Ethernet), forming a mesh communication structure without a central node. Communication content may include, but is not limited to, real-time operating status information of each energy storage unit (such as SOC, charging / discharging power, temperature, voltage, current, etc.), local control objectives, and collaborative requirements.

[0039] Step S2: For each energy storage unit, based on the real-time collected operating data, evaluate at least one of the following: state of charge, health status, and safety status, to obtain the status evaluation result.

[0040] The operating data of the energy storage unit can be collected in real time by sensors installed inside the energy storage unit, such as SOC sensors, temperature sensors, voltage sensors, and current sensors. The collected data can be preprocessed by filtering and noise reduction to remove interference signals and ensure data accuracy.

[0041] In one possible implementation, the state of charge is estimated based on voltage and current data using a Kalman filter algorithm combined with the ampere-hour integral method; the health state is evaluated based on at least one of the capacity decay rate, internal resistance change rate, and historical cycle life data, combined with a neural network algorithm; and the safety state is determined based on whether the temperature data exceeds the limit and whether there are abnormal fluctuations in the voltage and current data.

[0042] In one embodiment, based on preprocessed operating data, the state of the energy storage unit is evaluated using multi-dimensional evaluation indicators. The State of Charge (SOC) can be estimated using a Kalman filter algorithm combined with the ampere-hour integral method to improve the accuracy of SOC estimation. The State of Health (SOH) can be evaluated using a neural network algorithm by analyzing parameters such as the cycle life, capacity decay rate, and internal resistance change rate of the energy storage unit. The safety status can be determined by monitoring whether the temperature of the energy storage unit exceeds the threshold, whether the voltage is abnormal, and whether there are fault signals such as leakage. The state evaluation results serve as an important basis for the control decision of the energy storage unit.

[0043] In step S3, each energy storage unit, based on its own state assessment results, combined with information obtained from adjacent nodes and the total power demand of the system, performs collaborative calculations through a distributed optimization algorithm to solve and execute the charging and discharging power commands of each energy storage unit with the goal of optimizing the overall system operation indicators.

[0044] Please see Figure 3 Based on the distributed communication network and energy storage unit status assessment results, a hierarchical distributed cooperative control strategy is designed, consisting of a local control layer and a cooperative control layer: Local Control Layer: Each energy storage unit's local controller formulates a preliminary local control scheme based on its own state assessment results and local control objectives (such as maintaining the State of Charge (SOC) within a reasonable range, avoiding excessive temperature, and achieving local load power supply). When the SOC of the energy storage unit is below the lower threshold, the local controller prioritizes charging that unit; when the temperature exceeds the upper threshold, cooling measures are immediately triggered (such as starting the cooling fan and adjusting the cooling system), and its charging and discharging power is limited. Simultaneously, the local controller sends its own state information and preliminary control scheme to adjacent intelligent agent nodes.

[0045] Cooperative Control Layer: After receiving information from neighboring nodes, each intelligent agent node, combined with the overall power demand of the system (such as power commands obtained from the power grid dispatch center or microgrid control center), performs cooperative calculations using distributed optimization algorithms such as the Alternating Direction Method of Multipliers (ADMM), consensus algorithms, and particle swarm optimization algorithms to optimize global power allocation. During the cooperative calculation process, the optimization objectives are to minimize system operating costs, minimize energy storage unit lifetime losses, and meet system power requirements. An optimization objective function is established as follows: System operating costs include the charging and discharging loss costs of energy storage units (related to charging and discharging power and efficiency), maintenance costs (related to operating time and status), and capacity decay costs (related to the number of charge and discharge cycles and depth), etc. The lifespan loss of energy storage units is quantified by controlling parameters such as the SOC fluctuation range and charge / discharge rate of each unit. The system power demand constraints include the requirement that the sum of the charging and discharging power of each energy storage unit must meet the total power command (charging or discharging) of the system, while the charging and discharging power of each energy storage unit must be within its rated power range.

[0046] The objective function is solved by a distributed optimization algorithm to obtain the optimal charging and discharging power command for each energy storage unit, and then fed back to the local control layer. The local control layer adjusts the local control scheme according to the optimal power command issued by the collaborative control layer to achieve the collaborative operation of each energy storage unit.

[0047] In step S4, each energy storage unit dynamically adjusts its control parameters based on the deviation between the actual operating status and the expected target.

[0048] To address changes in the operating environment and conditions of energy storage systems (such as fluctuations in renewable energy output, load changes, and energy storage unit failures), a dynamic adjustment and adaptive optimization mechanism is implemented. The system monitors operational deviations in real time. Each intelligent agent node calculates the deviation between its actual operating state (such as actual charging and discharging power, SOC changes, etc.) and the optimal control target, i.e. the expected target, and shares the deviation information with neighboring nodes through a distributed communication network.

[0049] The system adaptively adjusts control parameters. When the system's operating deviation exceeds a set threshold, an adaptive optimization mechanism is triggered. Through online learning algorithms (such as reinforcement learning algorithms), the system dynamically adjusts the parameters of the distributed optimization algorithm (such as weight coefficients, convergence factors, etc.) and the thresholds of the local control strategy (such as SOC upper and lower limits, temperature alarm thresholds, etc.) based on the system's historical operating data and real-time deviations. This enables the control strategy to adapt to changes in system operating conditions and maintain a good control effect at all times.

[0050] By constructing a distributed communication network based on multiple agent nodes, each node interacts with only its neighboring nodes. This not only completely avoids the risk of system paralysis caused by single-point failures in centralized control architectures, significantly improving system reliability, but also greatly reduces bandwidth pressure and latency in information transmission, enhancing the system's dynamic response capability. Real-time assessments covering the state of charge, health, and safety status of each energy storage unit are conducted, and the results are used as the core decision-making basis in subsequent collaborative computations. This ensures that the control strategy strictly adheres to the state constraints of each unit, effectively preventing damaging behaviors such as overcharging and over-discharging, and achieving proactive protection of the energy storage unit's lifespan. Simultaneously, a distributed optimization algorithm integrates the system's total power demand, neighboring node information, and local state, using the overall system operating indicators as the optimization objective. This achieves a balance between global economic optimization and lifespan optimization while meeting power requirements. Furthermore, a dynamic parameter adjustment mechanism based on actual operating deviations is introduced, enabling the system to adaptively respond to new energy fluctuations and load changes, continuously maintaining optimal operating conditions and achieving globally optimal control.

[0051] In one possible implementation, please refer to Figure 4 In step S1, each node interacts with its neighboring nodes through a distributed communication network, which may include steps S11 and S12.

[0052] In step S11, each agent node sends its own state assessment results and local power reference values ​​to its neighboring nodes.

[0053] In step S12, each agent node receives the state evaluation results and local power reference values ​​sent by its neighboring nodes.

[0054] The condition assessment results are used to characterize the health status of the energy storage unit.

[0055] The local power reference value quantifies the power allocation decisions in the initial control scheme into a transmittable numerical value, such as "desired charging power 50kW" or "maximum discharge power 30kW".

[0056] Each agent node sends its own state assessment results and local power reference values ​​to neighboring nodes and receives similar information from its neighbors, laying a crucial foundation for efficient and reliable distributed collaborative optimization. Through this structured, lightweight local information exchange, the system not only avoids the bandwidth bottlenecks and latency of centralized communication and ensures real-time control, but also provides each node with the necessary local environmental awareness for global optimization decisions: the state assessment results enable nodes to accurately evaluate the real-time capabilities and constraints of neighboring units, while the local power reference values ​​reflect the power allocation intentions of neighbors based on their own conditions. The fusion of these two types of information allows subsequent distributed optimization algorithms to quickly and convergently negotiate a globally coordinated and cost-effective optimal power allocation scheme while fully respecting the individual states and initial intentions of each unit. This ensures the system's comprehensive advantages in reliability, economy, and dynamic performance at the architectural level.

[0057] In one possible implementation, the distributed communication network is a mesh network without a central node and is configured with a communication redundancy mechanism; the communication redundancy mechanism may include: Configure at least two physical or logical links using different communication protocols for each agent node; Real-time monitoring of the communication quality of the primary link; When the communication quality of the primary link falls below a preset threshold, the communication path is automatically switched to the backup link.

[0058] By configuring heterogeneous dual communication links for each intelligent agent node, monitoring link quality in real time, and automatically switching to backup links when the performance of the primary link deteriorates, a robust communication layer with self-diagnosis and rapid recovery capabilities is constructed. This significantly improves the overall reliability of the system from a physical perspective. It ensures that when the performance of a single communication link degrades or is interrupted due to interference, congestion, or hardware failure, the critical status and control information between energy storage units can still be continuously and reliably exchanged through the backup path. This not only avoids the stagnation or failure of the collaborative optimization process caused by communication interruption and ensures the continuous and stable execution of distributed control logic, but also enables the entire energy storage system to cope with harsh field communication environments and local equipment failures. Thus, it provides a solid underlying communication guarantee for achieving highly reliable distributed collaborative operation.

[0059] In another embodiment, the distributed communication network can also be an edge computing node collaborative network. Under the edge computing node collaborative architecture, an access layer is deployed on the energy storage device side to collect and preprocess data in real time; the collaborative layer uses a federated learning framework and a distributed energy storage collaborative control model to achieve multi-node parameter sharing; the decision layer integrates digital twin technology to verify the control strategy through virtual simulation, aiming to ensure efficient and stable information interaction between energy storage units. All of these focus on solving the communication reliability problem, avoiding single link failures through redundant link switching. The edge computing node collaborative architecture adopts an improved scheme based on time-sensitive networking, using forward error correction coding technology to reduce the bit error rate and deploying dynamic routing protocols to optimize the transmission path, ensuring end-to-end latency stability, improving collaborative efficiency, and reducing communication bandwidth consumption. The ultimate goal is to provide a solid communication guarantee for the distributed collaborative control of the energy storage system, allowing each energy storage unit to perform subsequent status monitoring, collaborative control, fault handling, and other operations based on accurate interactive information.

[0060] In one possible implementation, please refer to Figure 5 Step S3, which involves collaborative calculation using a distributed optimization algorithm to optimize the overall system performance indicators and solve for the charging and discharging power commands of each energy storage unit, may include steps S31 to S33.

[0061] Step S31: Establish an objective function with the optimization objectives of minimizing system operating costs, minimizing energy storage unit lifespan loss, and meeting the total system power demand.

[0062] In one possible implementation, the system operating cost includes charging and discharging loss cost, maintenance cost, and capacity decay cost; the life loss of the energy storage unit is quantified by the fluctuation range of its state of charge and the rate of change of charging and discharging power.

[0063] System operating costs may include the cost of charging and discharging losses of energy storage units (related to charging and discharging power and efficiency), maintenance costs (related to operating time and status), and capacity decay costs (related to the number of charge and discharge cycles and depth).

[0064] The system power requirement can include the sum of the charging and discharging power of each energy storage unit, which must meet the total power command (charging or discharging) of the system, while the charging and discharging power of each energy storage unit must be within its rated power range.

[0065] By setting an objective function to simultaneously optimize the three dimensions of cost, lifetime, and power satisfaction, a specific multi-objective optimization problem can be solved.

[0066] Step S32: The charging and discharging power capability and state of charge range of each energy storage unit are used as constraints.

[0067] Charge / discharge power capability can refer to the maximum charge / discharge power that each energy storage unit can withstand instantaneously (e.g., rated power of ±100kW).

[0068] State of charge (SOC) range refers to the range of cumulative charge allowed for each energy storage unit. For example, lithium batteries are typically set to operate between 20% and 90% to ensure that the SOC of each unit fluctuates within a safe and healthy range throughout the entire operation.

[0069] By explicitly elevating the two key individual attributes of each energy storage unit to the core constraints of the global optimization problem, the collaborative algorithm can proactively and one-to-one ensure the safety and health of each participating unit while meeting the total demand of the power grid (system objective), thereby extending the equipment life.

[0070] Step S33: Using the alternating direction multiplier method, consensus algorithm or particle swarm optimization algorithm, the objective function is solved iteratively to obtain the charging and discharging power command of each energy storage unit.

[0071] After receiving information from neighboring nodes, each intelligent agent node combines the overall power demand of the system, such as power commands obtained from the power grid dispatch center or microgrid control center, and performs collaborative calculations through distributed optimization algorithms such as ADMM, consensus algorithms, and particle swarm optimization algorithms to optimize global power allocation.

[0072] The objective function is solved by a distributed optimization algorithm to obtain the optimal charging and discharging power command for each energy storage unit, and then fed back to the local control layer. The local control layer adjusts the local control scheme according to the optimal charging and discharging power command issued by the collaborative control layer to achieve the collaborative operation of each energy storage unit.

[0073] To cope with changes in the operating environment and conditions of the energy storage system (such as fluctuations in new energy output, load changes, and energy storage unit failures), a dynamic adjustment and adaptive optimization mechanism is set up.

[0074] In one possible implementation, please refer to Figure 6 The dynamic adjustment of control parameters in step S4 may include steps S41 to S43.

[0075] Step S41: Monitor the deviation between the actual charging and discharging power and the change in state of charge and the corresponding expected target values.

[0076] The system monitors operational deviations in real time. Each intelligent agent node calculates its own actual operating status in real time, such as the deviation between the actual charging and discharging power and its corresponding expected target value, as well as the deviation between the change in state of charge (SOC) and its corresponding expected target value. The deviation information is then shared with neighboring nodes through a distributed communication network.

[0077] Step S42: When the absolute value or duration of the deviation exceeds the set threshold, parameter adjustment is triggered.

[0078] Step S43: Apply reinforcement learning algorithm and adjust the weight coefficients or convergence factors of the distributed optimization algorithm based on historical running data and current deviation.

[0079] When the deviation exceeds the set threshold, an adaptive optimization mechanism is triggered. Through online learning algorithms, such as reinforcement learning, the weight coefficients and convergence factors of the distributed optimization algorithm are dynamically adjusted based on the system's historical operating data and the current deviation, so that the control strategy can adapt to changes in system operating conditions and always maintain a good control effect.

[0080] In other embodiments, the thresholds of the local control strategy can also be adjusted, such as the upper and lower limits of SOC, temperature alarm thresholds, etc.

[0081] In one possible implementation, the distributed collaborative control and optimization method for the energy storage system may further include fault diagnosis and fault-tolerant control, specifically including: Based on operational data and communication status, diagnose energy storage unit faults or communication link faults. If the fault is due to abnormal sensor data, the abnormal data will be corrected by a multi-source data fusion algorithm, and the power allocation weight of the corresponding energy storage unit will be reduced. If a cell experiences a severe failure or communication interruption, the faulty cell is isolated from the collaborative computing network, and the collaborative computing steps are re-executed to assign new charge / discharge power commands to the remaining normal cells.

[0082] Fault diagnosis involves analyzing real-time operating data of each energy storage unit (such as abnormal voltage fluctuations, sudden current changes, rapid temperature rises, and excessive deviations in SOC estimation) and the transmission status of the communication network (such as excessively high data packet loss rate and communication delays exceeding thresholds). Fault diagnosis algorithms (such as fuzzy logic algorithms and support vector machine algorithms) are then used to diagnose the fault types (such as battery cell faults, sensor faults, and communication faults) and fault locations of the energy storage units in real time, and fault alarm signals are issued.

[0083] Fault-tolerant control means that when a fault is detected in an energy storage unit, corresponding fault-tolerant measures are taken according to the severity of the fault. For minor faults (such as decreased sensor accuracy), the fault data is corrected by a data fusion algorithm (fusing data from multiple related sensors) to ensure the accuracy of control decisions. For serious faults (such as damage to a single battery cell or interruption of a communication link), the faulty energy storage unit is automatically isolated from the system. At the same time, the charging and discharging power of the remaining normal energy storage units is redistributed through a distributed collaborative control strategy to make up for the power gap caused by the removal of the faulty unit, ensuring that the system can continue to operate stably and preventing the fault from spreading and causing system paralysis.

[0084] In one embodiment, taking a distributed energy storage system in a new energy microgrid as an example, the system includes 10 energy storage units, of which 5 are lithium battery energy storage units (rated capacity 200kWh, rated charge and discharge power 100kW), 3 are vanadium redox flow battery energy storage units (rated capacity 300kWh, rated charge and discharge power 80kW), and 2 are flywheel energy storage units (rated capacity 50kWh, rated charge and discharge power 150kW).

[0085] A distributed communication network is built using LoRa wireless communication technology. Each energy storage unit is equipped with a LoRa communication module, with a communication distance set at 1km, forming a mesh communication structure between the units. Additionally, each unit is equipped with an Ethernet communication module as a backup communication link, automatically switching to Ethernet communication when the LoRa communication link fails.

[0086] Each energy storage unit is equipped with a SOC sensor (measurement accuracy ±2%), a temperature sensor (measurement range -20℃ to 80℃, accuracy ±0.5℃), a voltage sensor (measurement range 0 to 1000V, accuracy ±0.2%), and a current sensor (measurement range -500A to 500A, accuracy ±0.2%). The sensor data is acquired through a data acquisition card (sampling frequency 1kHz). The acquired data is filtered (using Kalman filtering algorithm) and then transmitted to the local controller.

[0087] The local controller employs an embedded processor and deploys a state monitoring and evaluation model (SOC estimation uses Kalman filtering combined with ampere-hour integral method, and SOH evaluation uses BP neural network algorithm) and a local control strategy. Each energy storage unit interacts with information through a distributed communication network. The collaborative control layer deploys the alternating direction multiplier method as a distributed optimization algorithm, and the cost weight coefficients in the optimization objective function are set according to the microgrid's operational requirements. The dynamic adjustment and adaptive optimization mechanism uses a Q-learning reinforcement learning algorithm to adjust the convergence factor and SOC upper and lower limit thresholds of the ADMM algorithm online. The fault diagnosis and fault-tolerant control module deploys a fuzzy logic fault diagnosis algorithm and a fault-tolerant control strategy based on a consensus algorithm.

[0088] After the energy storage system is started, the local controller of each energy storage unit initializes, completes self-checks of sensors, communication modules, and control algorithms, and ensures that all hardware devices and software modules are operating normally. At the same time, each intelligent agent node discovers its neighbors through the distributed communication network, establishes communication connections, determines the list of neighboring nodes, and sends its initial state information (such as initial SOC, initial temperature, rated parameters, etc.) to neighboring nodes.

[0089] Each local controller collects real-time operating data (SOC, temperature, voltage, current, etc.) of the energy storage unit through a data acquisition device. After preprocessing, the data is input into the status monitoring and evaluation model to evaluate the SOC, SOH, and safety status of the energy storage unit in real time. The evaluation results are stored in the local database, and the evaluation results and real-time operating data are sent to adjacent nodes at regular intervals.

[0090] Based on the state assessment results and local control objectives, the local controller formulates a preliminary local control scheme. For example, when the SOC of a lithium battery energy storage unit drops to 18%, the local controller initially formulates a control scheme with a charging power of 50kW; when the temperature of a flywheel energy storage unit rises to 52℃, the local controller initially formulates a control scheme to reduce the charging and discharging power to 100kW and start the cooling fan.

[0091] After receiving the status information and preliminary control scheme from neighboring nodes, each intelligent agent node, in conjunction with the total system power command issued by the microgrid control center (e.g., the energy storage system needs to discharge 200kW at a certain moment), initiates the ADMM distributed optimization algorithm for collaborative calculation. During the calculation process, the objectives are to minimize system operating costs and energy storage unit lifetime losses, while simultaneously satisfying constraints such as the charging and discharging power of each energy storage unit being within its rated range and the total discharge power reaching 200kW. The optimal charging and discharging power command for each energy storage unit is then obtained. For example, calculations show that 5 lithium battery energy storage units discharge 30kW, 35kW, 25kW, 40kW, and 30kW respectively; 3 vanadium redox flow battery energy storage units discharge 15kW, 20kW, and 15kW respectively; and 2 flywheel energy storage units discharge 0kW (due to the current low flywheel SOC, priority is given to retaining power) and 0kW respectively, with a total discharge power of 200kW, and the operating status of each unit meets the constraints.

[0092] Each local controller receives the optimal charging and discharging power command from the collaborative control layer, compares the difference between its initial control scheme and the optimal command, adjusts its local control strategy, and controls the energy storage unit to operate at the optimal charging and discharging power. Simultaneously, it monitors the deviation between the actual operating status and the optimal target in real time. If the deviation is within the allowable range (e.g., power deviation not exceeding ±5kW, SOC deviation not exceeding ±3%), the current control scheme is maintained; if the deviation exceeds the allowable range, a dynamic adjustment mechanism is triggered.

[0093] At a certain moment, due to a sudden increase in wind power output and a decrease in microgrid load demand, the microgrid control center adjusts the power command of the energy storage system to 150kW charging. This causes the actual SOC change rate of some energy storage units to deviate from the optimal target by more than ±3%. At this point, the Q-learning reinforcement learning algorithm is triggered. Based on historical operating data and real-time deviations, the convergence factor of the ADMM algorithm is dynamically adjusted from 0.01 to 0.015. Simultaneously, the upper limit threshold of the SOC for lithium battery energy storage units is increased from 80% to 85% to adapt to the increased charging power. After the adjustment, each energy storage unit re-performs collaborative optimization calculations to obtain a new optimal charging and discharging power command, ensuring that the system can still operate stably and economically under the new conditions.

[0094] During system operation, a voltage sensor in one of the lithium-ion battery energy storage units malfunctioned, causing significant fluctuations in the collected voltage data and a State of Charge (SOC) estimation deviation exceeding ±5%. The fault diagnosis module analyzed the voltage data fluctuation characteristics and SOC estimation deviation, employing a fuzzy logic algorithm to diagnose the voltage sensor malfunction and issue a fault alarm signal. It then initiated a data fusion algorithm, fusing current and temperature data from the faulty lithium-ion battery energy storage unit with voltage data from adjacent units to correct the faulty voltage data and ensure accurate SOC estimation. Simultaneously, the fault information was shared with other intelligent agent nodes, and in subsequent collaborative optimization calculations, the charging and discharging power allocation ratio of the faulty unit was appropriately reduced (from the normal 30kW to 20kW) to prevent the fault from escalating. Once the faulty sensor was replaced, the system automatically restored the unit's normal power allocation ratio.

[0095] The key aspects of this solution focus on five core levels: First, constructing a distributed communication network based on multiple agents, using a mesh structure without a central node combined with a communication redundancy mechanism to achieve efficient and reliable information interaction among energy storage units, laying the foundation for collaborative control; Second, establishing a precise energy storage unit status monitoring and evaluation model, through multi-sensor data acquisition, preprocessing, and multi-algorithm fusion (such as Kalman filtering and neural networks), to monitor SOC, SOH, and safety status in real time, providing data support for control decisions; Third, designing a hierarchical distributed collaborative control strategy, relying on the local control layer to quickly respond to unit state constraints, and the collaborative control layer to achieve optimal global power allocation through algorithms such as ADMM, balancing local safety and overall benefits; Fourth, incorporating dynamic adjustment and adaptive optimization mechanisms, using reinforcement learning algorithms to optimize control parameters in real time according to changes in operating conditions, ensuring the system's adaptability in complex scenarios; Fifth, building a comprehensive fault diagnosis and fault-tolerant control module, through fault identification, data correction, unit isolation, and power reallocation, ensuring stable operation of the system under fault conditions, forming a closed-loop key system of communication-monitoring-control-optimization-fault tolerance.

[0096] The distributed collaborative control and optimization method for this energy storage system has at least the following advantages: By constructing a distributed communication network without a central node and equipping it with communication redundancy mechanisms, fault diagnosis and fault-tolerant control modules, the risk of system paralysis caused by the failure of the central controller in centralized control is completely avoided, and the system reliability is greatly improved. By leveraging real-time status monitoring and evaluation models, the SOC, SOH, and safety status of energy storage units can be accurately grasped. The control strategy strictly follows the status constraints of each unit, effectively avoiding problems such as overcharging, over-discharging, and abnormal temperature, significantly reducing the lifespan loss of energy storage units (such as reducing the capacity decay rate of lithium batteries to 1.5%), and extending their service life. With minimizing system operating costs as one of the optimization objectives, the system comprehensively considers various costs such as charging and discharging losses, maintenance, and capacity decay. It achieves optimal power allocation through a distributed optimization algorithm, thereby effectively reducing overall operating costs. Each energy storage unit interacts with its adjacent nodes only, reducing communication latency. Combined with a hierarchical control strategy, the system can quickly respond to operating conditions such as fluctuations in new energy output and load changes, thereby enhancing dynamic response performance. Its distributed architecture and adaptive optimization mechanism make it highly adaptable. When the system is expanded or adjusted, there is no need for large-scale reconstruction. It can be flexibly adapted to multiple scenarios such as new energy microgrids, distribution network frequency regulation and peak shaving, and user-side energy storage, providing comprehensive support for the efficient and stable operation of energy storage systems.

[0097] This disclosure also provides a distributed energy storage system, including: Multiple energy storage units, A distributed communication network connects multiple energy storage units that act as intelligent agent nodes, enabling information exchange between the nodes. Each intelligent agent node is configured to: assess at least one of the state of charge, health, and safety states of the corresponding energy storage unit based on its real-time operating data, and obtain a state assessment result; based on its own state assessment result, combined with information obtained from neighboring nodes and the total power demand of the system, perform collaborative calculations through a distributed optimization algorithm to solve and execute the charging and discharging power commands of each energy storage unit with the goal of optimizing the overall system operating indicators; and dynamically adjust control parameters according to the deviation between the actual operating state of the corresponding energy storage unit and the expected target.

[0098] Optionally, each agent node is also configured to: send its own state assessment results and local power reference values ​​to neighboring nodes; and receive state assessment results and local power reference values ​​sent by neighboring nodes.

[0099] Optionally, the state of charge is estimated based on voltage and current data using a Kalman filter algorithm combined with the ampere-hour integral method; The health status is evaluated based on at least one of the following: capacity decay rate, internal resistance change rate, and historical cycle life data, combined with a neural network algorithm. The safety status is determined based on whether the temperature data exceeds the limit and whether there are abnormal fluctuations in the voltage and current data.

[0100] Optionally, the agent node is specifically configured as follows: An objective function is established with the optimization objectives of minimizing system operating costs, minimizing energy storage unit lifespan loss, and meeting the total system power demand. The charging and discharging power capacity and state of charge range of each energy storage unit are used as constraints. The objective function is solved iteratively using the alternating direction multiplier method, consensus algorithm, or particle swarm optimization algorithm to obtain the charging and discharging power commands of each energy storage unit.

[0101] Optionally, the system operating cost includes charging and discharging loss cost, maintenance cost, and capacity decay cost; the life loss of the energy storage unit is quantified by the fluctuation range of its state of charge and the charging and discharging rate.

[0102] Optionally, the agent node is specifically configured as follows: Monitor the deviation between the actual charging and discharging power and changes in state of charge and the corresponding expected target values; When the absolute value or duration of the deviation exceeds the set threshold, parameter adjustment is triggered. By applying reinforcement learning algorithms, the weight coefficients or convergence factors of the distributed optimization algorithm are adjusted based on historical operating data and current deviations. This disclosure also provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned distributed cooperative control and optimization method for energy storage systems.

[0103] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0104] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0105] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for distributed collaborative control and optimization of energy storage systems, characterized in that, The method, applied to a distributed energy storage system comprising multiple energy storage units, includes: A distributed communication network is constructed, with each energy storage unit acting as an intelligent node, and each node interacting with its neighboring nodes through the distributed communication network. For each energy storage unit, based on real-time collected operating data, at least one of the following is assessed: state of charge, health status, and safety status, to obtain a status assessment result; Based on its own state assessment results, combined with information obtained from adjacent nodes and the total power demand of the system, each energy storage unit performs collaborative calculations through a distributed optimization algorithm to solve and execute the charging and discharging power commands of each energy storage unit with the goal of optimizing the overall system operation indicators. Each energy storage unit dynamically adjusts its control parameters based on the deviation between its actual operating status and the expected target.

2. The method according to claim 1, characterized in that, Each node interacts with its neighboring nodes through the distributed communication network, including: Each agent node sends its own state assessment results and local power reference values ​​to its neighboring nodes; Each agent node receives the state assessment results and local power reference values ​​sent by its neighboring nodes.

3. The method according to claim 2, characterized in that, The distributed communication network is a mesh network without a central node and is configured with a communication redundancy mechanism; the communication redundancy mechanism includes: Configure at least two physical or logical links using different communication protocols for each agent node; Real-time monitoring of the communication quality of the primary link; When the communication quality of the primary link is lower than a preset threshold, the communication path is automatically switched to the backup link.

4. The method according to claim 1, characterized in that, The state of charge is estimated based on voltage and current data using a Kalman filter algorithm combined with the ampere-hour integral method. The health status is evaluated based on at least one of the following: capacity decay rate, internal resistance change rate, and historical cycle life data, combined with a neural network algorithm. The safety status is determined based on whether the temperature data exceeds the limit and whether there are abnormal fluctuations in the voltage and current data.

5. The method according to claim 1, characterized in that, Collaborative computation using distributed optimization algorithms aims to optimize the overall system performance, solving for the charging and discharging power commands of each energy storage unit, including: An objective function is established with the optimization objectives of minimizing system operating costs, minimizing energy storage unit lifespan loss, and meeting the total system power demand. The charging and discharging power capacity and state of charge range of each energy storage unit are used as constraints. The objective function is iteratively solved using the alternating direction multiplier method, consensus algorithm, or particle swarm optimization algorithm to obtain the charging and discharging power commands for each energy storage unit.

6. The method according to claim 5, characterized in that, The system operating cost includes charging and discharging loss cost, maintenance cost, and capacity decay cost; the lifespan loss of the energy storage unit is quantified by the fluctuation range of its state of charge and the charging and discharging rate.

7. The method according to claim 1, characterized in that, The dynamically adjusted control parameters include: Monitor the deviation between the actual charging and discharging power and changes in state of charge and the corresponding expected target values; When the absolute value or duration of the deviation exceeds a set threshold, parameter adjustment is triggered. By applying reinforcement learning algorithms, the weight coefficients or convergence factors of the distributed optimization algorithm are adjusted based on historical running data and the current deviation.

8. The method according to any one of claims 1-7, characterized in that, The method also includes fault diagnosis and fault-tolerant control, specifically including: Based on the aforementioned operating data and communication status, diagnose energy storage unit faults or communication link faults. If the fault is due to abnormal sensor data, the abnormal data will be corrected by a multi-source data fusion algorithm, and the power allocation weight of the corresponding energy storage unit will be reduced. If a unit experiences a severe failure or communication interruption, the faulty unit is isolated from the collaborative computing network, and the collaborative computing steps are re-executed to assign new charge / discharge power commands to the remaining normal units.

9. A distributed energy storage system, characterized in that, include: Multiple energy storage units, A distributed communication network connects multiple energy storage units that act as intelligent agent nodes, enabling information exchange between the nodes. Each of the aforementioned intelligent agent nodes is configured as follows: Based on the real-time operating data of the corresponding energy storage unit, at least one of its state of charge, health status and safety status is evaluated to obtain the status assessment result; Based on its own state assessment results, combined with information obtained from neighboring nodes and the total power demand of the system, a distributed optimization algorithm is used for collaborative calculation to optimize the overall system operation indicators, and to solve and execute the charging and discharging power commands of each energy storage unit. The control parameters are dynamically adjusted based on the deviation between the actual operating status of the corresponding energy storage unit and the expected target.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the distributed collaborative control and optimization method for energy storage systems as described in any one of claims 1-8.

Images