A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries

Abstract

The present invention proposes a reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries. First, a multi-level reconfigurable battery array architecture is constructed, in which each battery module integrates an independent three-port power switch unit and a wireless communication node, which can realize self-management and rapid response of the battery module. The bottom battery cluster adopts a MOSFET matrix topology, and the FPGA controls the PWM signal to dynamically adjust the battery series-parallel structure. The central controller calculates the optimal connection pair in real time based on power demand and controls the contactor to establish a new path. When multiple modules fail, the system automatically activates three-layer collaborative reconstruction, and the faulty module is switched to bypass through the three-port switch. The central controller refreshes the charge and discharge plan according to the system status cycle, thereby improving the system cycle efficiency and reducing the battery array life attenuation rate. The present invention realizes the online dynamic replacement of faulty batteries, improves the system operation stability and efficiency, and effectively reduces maintenance costs and downtime.

Description

Technical Field

[0001] The present invention belongs to the field of battery systems, and more particularly relates to a reconfigurable energy storage system for online dynamic replacement of faulty batteries and a control method thereof. Background Art

[0002] In today's society, the volatility and uncertainty of renewable energy necessitate the use of energy storage devices to provide stable power. Battery energy storage devices are widely used due to their excellent performance and high energy efficiency. However, battery energy storage systems typically consist of a large number of battery cells. If a single cell fails or its performance degrades, the performance of the entire energy storage system will be affected.

[0003] Traditional battery energy storage systems often require system shutdown and replacement of the entire battery cell when a battery cell failure occurs. This not only causes a power outage but also increases maintenance costs. Furthermore, due to differences in the operating conditions of individual battery cells, if one cell fails, other cells may not function properly under a sudden increase in load, resulting in decreased performance of the entire system.

[0004] To address these issues, battery management systems have been proposed to perform online diagnosis and prediction of batteries by monitoring various performance parameters. However, these systems are usually unable to achieve online dynamic replacement of faulty batteries, and after a battery cell failure, manual intervention is required in the tedious process of replacing the faulty battery.

[0005] Therefore, developing a reconfigurable energy storage system and control method that can dynamically replace faulty batteries online can improve the system's self-healing ability and operating efficiency, reduce maintenance costs and downtime, and has important practical application value. Summary of the Invention

[0006] The main technical problem to be solved by the present invention is how to effectively realize the online dynamic replacement of faulty batteries, so that when a battery cell fails, there is no need to shut down the system to replace the entire battery cell. At the same time, it can improve the operating efficiency and self-healing ability of the battery energy storage system, reduce the cost and time of fault switching and system maintenance, and ensure the stable and reliable operation of the battery energy storage system.

[0007] In order to achieve the above object, the present invention is implemented by adopting the following technical solutions: the method comprises:

[0008] Step 1: Build a multi-level reconfigurable battery array architecture. Design an energy storage array consisting of N standardized battery modules. Each module integrates an independent three-port power switch unit and a wireless communication node. The array adopts a hierarchical cascade topology. The bottom layer is battery cluster-level reconfiguration, the middle layer is an inter-cluster ring interconnection architecture, and the top layer is connected to the DC bus through a multi-winding bidirectional DC-DC converter. This allows any single-point or multi-point fault module to be electrically isolated, and triggers adjacent modules to automatically switch connection modes, forming a new power transmission path.

[0009] Step 2: Each battery module integrates a three-port power switching unit, designed using gallium nitride MOSFET devices. These units include an input main power channel, an output load channel, and a bypass backup channel. Nanosecond switching is achieved through an optocoupler isolation drive circuit. The module's built-in wireless communication node is networked based on the Zigbee-3.0 protocol, transmitting voltage and current data to a central controller in real time. The array's underlying battery cluster uses a MOSFET matrix topology, with an FPGA controlling the PWM signal to dynamically adjust the battery series-parallel configuration. A single module failure triggers a cluster-wide reconfiguration mechanism, automatically reorganizing healthy modules into series compensation mode to maintain cluster voltage stability.

[0010] Step 3: Connect each battery cluster via inter-cluster ring copper busbars. Solid-state contactors and bidirectional current detection circuits are deployed between every two adjacent clusters to form a dynamic parallel path. A central controller calculates the optimal connection pair in real time based on power demand and controls the contactors to establish new paths. The top layer uses a multi-winding, high-frequency, isolated, bidirectional DC-DC converter. The primary winding of its magnetically integrated transformer is independently connected to each battery cluster, and the secondary output is connected in parallel to the DC bus. Phase-shift modulation is used to achieve directional power distribution.

[0011] Step 4: When multiple modules fail, the system automatically activates three-layer coordinated reconstruction. The failed module switches to bypass mode via a three-port switch, and the central controller updates the charge and discharge schedule based on the system status cycle, improving system cycle efficiency and reducing battery array life attenuation.

[0012] Step 5: Implement online dynamic replacement of faulty batteries. The system supports electrical isolation, bypass switching, cluster-level reconstruction, dynamic parallel connection between clusters, top-level power directional allocation, as well as capacity migration, parameter alignment, and capacity reorganization of faulty batteries.

[0013] In one embodiment, the three-port power switch unit uses a gallium nitride MOSFET device and implements nanosecond switching through an optocoupler isolation drive circuit.

[0014] In one embodiment, the wireless communication nodes are networked based on the Zigbee-3.0 protocol and can transmit voltage and current data to the central controller in real time.

[0015] In one solution, the bottom battery cluster adopts a MOSFET matrix topology, and the FPGA controls the PWM signal to dynamically adjust the battery series-parallel structure.

[0016] In one solution, the middle layer connects the battery clusters through a ring copper busbar, and a solid-state contactor and a current bidirectional detection circuit are deployed between every two adjacent clusters to form a dynamic parallel path.

[0017] In one solution, the top layer uses a multi-winding high-frequency isolated bidirectional DC-DC converter, the primary winding of its magnetic integrated transformer is independently connected to each battery cluster, and the secondary side is output in parallel to the DC bus.

[0018] In one solution, when multiple modules fail, the system can automatically activate three-layer coordinated reconstruction. The failed module is switched to bypass through a three-port switch, and the charging and discharging plan is periodically refreshed through the central controller.

[0019] In one solution, the system uses a central controller to calculate the optimal connection pair in real time based on power demand and control the contactors to establish new paths, thereby achieving directional power distribution and improving system cycle efficiency.

[0020] Beneficial effects of the present invention:

[0021] First, the present invention achieves online dynamic replacement of faulty batteries by constructing a multi-level reconfigurable battery array architecture. Traditional energy storage systems typically require downtime and replacement of the entire battery cell upon a battery cell failure, a process that is both time-consuming and increases maintenance costs. However, the present invention enables online dynamic replacement of faulty batteries, eliminating downtime for replacement and significantly improving system efficiency.

[0022] Secondly, the present invention integrates an independent three-port power switch unit and wireless communication node into each battery module, enabling each battery module to self-manage and quickly respond to failures. This design enhances the system's self-healing capabilities, ensuring that the entire system remains stable even when single or multiple battery modules fail. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, a reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries include the following steps:

[0026] Step 1: Build a multi-level reconfigurable battery array architecture

[0027] The design involves an energy storage array consisting of N standardized battery modules, each integrating an independent three-port power switch unit (connecting the input, output, and bypass path) and a wireless communication node. The array employs a hierarchical cascade topology: a bottom layer for cluster-level reconfiguration (using a MOSFET matrix to enable intra-cluster series and parallel switching of batteries), a middle layer for inter-cluster ring interconnection (supporting dynamic parallel connection of any two clusters), and a top layer for connection to the DC bus via a multi-winding bidirectional DC-DC converter. This architecture allows any module with single or multiple faults to be electrically isolated, while triggering adjacent modules to automatically switch their connection modes, forming new power transmission paths.

[0028] To enable dynamic online replacement of faulty batteries, an energy storage array consisting of standardized battery modules was first constructed. Each module features a three-port power switching unit using gallium nitride (GaN) MOSFETs. These switches comprise a main input power channel, an output load channel, and a bypass backup channel. Optocoupler isolation and drive circuitry enable nanosecond switching. The module's built-in wireless communication nodes, based on the Zigbee 3.0 protocol, transmit real-time voltage and current data to a central controller. The array's bottom-layer battery clusters utilize an 8×4 MOSFET matrix topology, with 48 PWM signals controlled by an FPGA to dynamically adjust the battery series-parallel structure. When a single module failure is detected, a cluster-wide reconfiguration mechanism is triggered, automatically reorganizing healthy modules into a series compensation mode to maintain cluster voltage stability. The middle layer connects the battery clusters via circular copper busbars, with solid-state contactors and bidirectional current detection circuits deployed between every two adjacent clusters to form dynamic parallel paths. A central controller calculates the optimal connection pair (e.g., connecting Cluster A and Cluster D) in real time based on power demand and controls the contactors to establish a new path within 10ms. The top layer utilizes a multi-winding, high-frequency, isolated, bidirectional DC-DC converter. The primary winding of its specially designed magnetically integrated transformer is independently connected to each battery cluster, with the secondary output connected in parallel to the 800V DC bus. Phase-shift modulation is used to achieve directional power distribution. When multiple modules fail, the system automatically activates three-layer collaborative reconstruction: the faulty module switches to bypass mode through a three-port switch to disconnect the electrical connection. Adjacent modules form a power bypass path based on a graph theory algorithm (for example, when module 7 fails, module 6 → module 8 are directly connected). Simultaneously, the inter-cluster ring architecture initiates cross-cluster power support (for example, when cluster 1 fails, clusters 2 and 3 are dynamically connected in parallel to bear its load). Finally, the multi-winding converter maintains the bus voltage fluctuation within ±1.5% by adjusting the duty cycle of each winding in real time.

[0029] Step 2: Develop online fault diagnosis and pre-replacement decision algorithms

[0030] An online diagnostic system based on deep reinforcement learning is deployed. This system collects the voltage ripple spectrum, internal resistance change rate, and temperature field gradient of each module in real time, and uses this information as input into a time-series neural network to predict the remaining health (RUL). When the module failure probability exceeds a threshold, a three-level response mechanism is triggered. The first level initiates pre-synchronization control of modules adjacent to the fault point (ensuring that the backup module voltage tracks the busbar); the second level generates a dynamic replacement path planning diagram, using graph theory algorithms to calculate the optimal reconstruction path to minimize system losses; the third level sends the fault coordinates and pre-reconstruction plan to the operation and maintenance terminal, awaiting manual confirmation or automatic execution.

[0031] The online fault diagnosis system uses a deep reinforcement learning (DRL) framework as its core, building a dual-channel heterogeneous data fusion network. The fault prediction model uses a temporal convolution-long short-term memory (TCN-LSTM) hybrid neural network as its backbone, with its input layer simultaneously collecting real-time signals from three dimensions:

[0032] 1. Voltage ripple spectral density (0.1-10kHz frequency band eigenvalue after Fourier transform)

[0033] 2. Internal resistance change rate (Online identification value based on Kalman filter)

[0034] 3. Temperature field gradient (Spatial derivative of the 4×4 thermocouple array inside the module)

[0035] After three layers of feature extraction, the time domain features are processed by the LSTM unit:

[0036]

[0037]

[0038] Spatial domain features are extracted by dilated causal convolutional layers:

[0039]

[0040] The final fusion layer outputs the health score and remaining life (Unit is hours), when The three-level response mechanism is triggered.

[0041] The pre-replacement decision first starts the pre-synchronization control of the fault point neighborhood module: select the backup module within a radius of three hops, and adjust its bidirectional DC-DC converter duty cycle D through the PI controller to make its output voltage Tracks bus voltage exponentially :

[0042]

[0043] The synchronization error is constrained to .

[0044] The dynamic path planning engine is then called to model the entire battery array as a weighted undirected graph , where node V is the battery module and the weight of edge E is Include:

[0045] Path resistance (including switch conduction loss)

[0046] Reconstruction of switch operation times

[0047] Historical failure probability

[0048] The improved Dijkstra algorithm is used to minimize the objective function:

[0049]

[0050] Calculate and obtain a topology reconstruction solution (for example, a detour path: module 12 → module 15 → module 17).

[0051] Finally, a pre-replacement instruction package in JSON format is generated, containing the coordinates of the faulty module, the reconstruction path sequence, and the capacity migration schedule. This package is encrypted and transmitted to the operation and maintenance terminal via the MQTT protocol. If the instruction is manually confirmed or the system does not receive a rejection within 200ms, the switching control logic in step 3 is triggered.

[0052] Step 3: Implement zero-downtime fault-tolerant switching control

[0053] A distributed collaborative control strategy is employed: When the system maintains continuous power output, the faulty module's three-port switch switches the current to a preset bypass channel within 200μs and activates the energy storage buffer capacitor (pre-installed at the bus interface) to compensate for the power shortfall. Simultaneously, the control center triggers three actions via power line carrier communication: ① Reconfigure the topology of the nearest backup battery cluster (automatically switching to a pre-calculated equivalent impedance connection scheme); ② Start power redistribution of the multi-winding DC-DC converter (based on modular multi-level predictive control); and ③ The soft-start circuit of the newly connected module automatically adjusts the output voltage phase. Ultimately, seamless power transfer is completed within 10ms, ensuring bus voltage fluctuations of less than ±2%.

[0054] The fault switching control adopts a distributed collaborative architecture and completes a four-step linkage operation within the critical 200 microsecond window period after the central controller issues the switching command. First, the three-port power switch unit of the fault module responds to the optocoupler isolation drive signal and uses the ultra-low turn-off delay characteristics of the gallium nitride MOSFET (<50ns) to switch the main power channel current to the Forced to switch to the bypass channel, and at the same time activate the energy storage buffer capacitor group connected in parallel at the output end (the capacity is designed to be ,in is the rated power, is the switching duration, The capacitor bank injects compensation current into the busbar through a bidirectional Buck-Boost circuit pre-installed at the DC-DC converter interface. , filling the power gap in real time and suppressing voltage mutations.

[0055] Synchronously trigger the three-layer coordinated reconstruction: ① At the battery cluster level, based on the topology reconstruction plan generated in step 2, the FPGA controller of the non-faulty cluster closest to the fault point (such as cluster B) immediately executes the pre-stored equivalent impedance connection strategy - switching the original series structure to an adaptive parallel mode through the MOSFET matrix (for example, reorganizing 8 series and 4 parallel to 4 series and 8 parallel), making the cluster output impedance The error with the pre-fault state is controlled within ±3%. ② The multi-winding bidirectional DC-DC converter starts the modular multi-level predictive control (MMC-PC) to redistribute the power weight according to the health status of each cluster. Its real-time optimization goal is:

[0056]

[0057] in is the actual power of the kth winding, Dynamically adjust according to load demand, is the duty cycle change (constraint ), the modulation wave is updated every 100μs through rolling time domain optimization; ③ The newly connected backup module adopts a zero-voltage soft start strategy, and its output stage series IGBT / H-bridge circuit tracks the bus voltage phase in real time based on a phase-locked loop (PLL) , and adjust the drive pulse phase angle satisfy:

[0058]

[0059] The module output voltage is synchronized with the busbar within ±1° to avoid circulating current shock.

[0060] The entire switching process is synchronized by power line carrier communication (frequency band 150kHz-1MHz, baud rate 1Mbps) to achieve nanosecond signal synchronization, and the status of each subsystem is coordinated by a distributed time-triggered electrical mechanism (TTE). When the voltage drops linearly to zero within 8ms, the backup module completes its rated power carrying capacity, and the peak-to-peak value of the bus voltage fluctuation is limited to within ±15V (±2% @800VDC), and the system achieves fault-tolerant switching without perception.

[0061] Step 4: Dynamic access to virtual battery pool and seamless capacity migration

[0062] After the faulty module is electrically isolated, the intelligent capacity migration mechanism is activated: the control center first calculates the real-time remaining capacity (SOC) and output impedance characteristics of the faulty module, and automatically matches the backup battery cluster with the closest electrical parameters from the hot backup cluster using the pre-set virtual battery mapping table. The three-step seamless access process is then executed:

[0063] ① Power path reconstruction: Control the multi-port bidirectional switches of the backup cluster to switch to the equivalent circuit topology, so that the output impedance deviation of the backup cluster from the original faulty module is ≤5%;

[0064] Controls the multi-port solid-state switch matrix (GaN FET array) of the backup cluster to switch to a topology equivalent to the faulty module (e.g., converting a star connection to a delta connection). After the topology is adjusted, the output impedance matching must meet the following requirements:

[0065]

[0066] By adjusting the adjustable inductor in the topology in real time With resistance box Achieve precise compensation.

[0067] ② Soft capacity transfer: A controllable ramp current (dI / dt ≤ 10A / ms) is injected through the bidirectional DC-DC converter group to gradually transfer the load power to the backup cluster, while the bus buffer capacitor absorbs the switching transient impact;

[0068] A cascaded bidirectional DC-DC converter group is deployed at the output of the backup cluster to generate a slope-controlled ramp current:

[0069]

[0070] At the same time, the buffer capacitor group connected in parallel with the busbar (design capacity ,in is the maximum load mutation value, (allowable pressure difference) absorbs transient impact energy and ensures that the load power is The exponential law migration avoids power steps.

[0071] ③ Online parameter alignment: An extended Kalman filter is used to observe the voltage difference between the new and old modules in real time, completing active SOC balancing and adaptive internal resistance compensation within 30ms, ultimately achieving a capacity migration error of less than 1%.

[0072] During the migration process, an extended Kalman filter (EKF) is implanted to observe the voltage deviation between the backup cluster and the original system at a sampling rate of 200kHz. , the state equation is modeled as:

[0073]

[0074] The observation matrix , is the derivative of the open circuit voltage-SOC curve), process noise The internal resistance compensation is completed by the adaptive PID of the DC-DC converter:

[0075]

[0076] Finally achieved within 30ms and , the migration error is stable within the 1% threshold.

[0077] After the faulty module is electrically isolated, the intelligent capacity migration mechanism is activated based on the virtual battery mapping engine. The control center first calculates the real-time status parameters of the faulty module: remaining capacity and dynamic output impedance (100Hz-1kHz sweep frequency measurement value). Virtual battery mapping table preset by hash index (including SOC, Z, temperature coefficient of 512 groups of hot backup clusters) , etc.), and use weighted Euclidean distance to match the closest backup cluster:

[0078]

[0079] (Weight ), select the backup cluster with the smallest deviation to activate the access process.

[0080] Step 5: Dynamic capacity reorganization and system optimization

[0081] After a new module is connected, adaptive capacity reorganization is initiated: First, the new module's internal resistance-capacity curve is measured. Combined with the current SOC distribution of the battery array, a game-theoretic optimal matching model is used to reallocate the power scheduling weights of each cluster. The new module is then incorporated into the virtual battery pool (VBP) using a consistent hashing algorithm, dynamically updating the system's total available capacity (MAP) curve. Finally, a capacity self-calibration is performed: a multi-band AC injection signal is applied during light-load periods, the impedance spectrum of the entire array is measured, and a capacity-health correction factor is generated. This factor is then updated in the rolling optimization scheduling model of the energy management system (EMS).

[0082] After the new module is successfully connected, the system immediately starts the three-layer adaptive reorganization mechanism. First, a 0.1C-1C multi-step charge and discharge test pulse (pulse width 20ms) is applied through the bidirectional DC-DC converter group, and the voltage response is synchronously collected. , solve the internal resistance-capacity characteristic curve of the new module and maximum available capacity (Based on the joint calibration of ampere-hour integration and Peukert equation). Combined with the real-time status distribution diagram of the battery array (including SOC, SOH, and temperature field data), and a non-cooperative game theory optimization model is used to allocate power scheduling weights: the objective function is defined as minimizing the total system loss:

[0083]

[0084] (in is the scheduling weight of cluster k, The optimal power allocation scheme is calculated using the Nash equilibrium solver (Gurobi optimization engine), with the following constraints:

[0085] Inter-cluster SOC differences

[0086] Single cluster power mutation rate

[0087] Then, the new module is compiled into the virtual battery pool based on the consistent hashing algorithm: The triplet generates a 128-bit hash value as a virtual node to join the hash ring and dynamically update the system's global capacity MAP curve (Describes the maximum sustainable power supply time of the system under different load requirements.) During the reconstruction process, virtual slot migration technology is used to redistribute traffic to only two adjacent physical nodes, ensuring the disturbance time of topology changes. .

[0088] Finally, perform full array online capacity calibration: During the night light load period (system load rate ), the multi-band AC excitation signal is injected by the multi-winding converter (sweep range 100Hz-10kHz, amplitude The impedance spectrum of each cluster is measured synchronously by distributed acquisition units. , combined with electrochemical impedance spectroscopy (EIS) analysis to generate a capacity-health correction factor:

[0089]

[0090] This coefficient is combined with the cyclic aging model to update the actual available capacity of the cluster. , data is synchronized to the EMS rolling optimization model in real time. EMS uses model predictive control (MPC) to refresh the charge and discharge plan in a 5-minute cycle. The optimization goals include: improving system cycle efficiency . Battery array life attenuation rate is reduced .

[0091] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0092] It should be understood that the detailed description of the technical solutions of the present invention using the preferred embodiments above is illustrative and not restrictive. A person skilled in the art, after reading the present specification, may modify the technical solutions described in the embodiments or replace some of the technical features therein with equivalents; such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries, characterized in that: The method includes: Step 1: Build a multi-level reconfigurable battery array architecture. Design an energy storage array consisting of N standardized battery modules. Each module integrates an independent three-port power switch unit and a wireless communication node. The array adopts a hierarchical cascade topology. The bottom layer is battery cluster-level reconfiguration, the middle layer is an inter-cluster ring interconnection architecture, and the top layer is connected to the DC bus through a multi-winding bidirectional DC-DC converter. This allows any single-point or multi-point fault module to be electrically isolated, and triggers adjacent modules to automatically switch connection modes, forming a new power transmission path. Step 2: Each battery module integrates a three-port power switching unit, designed using gallium nitride MOSFET devices. These units include an input main power channel, an output load channel, and a bypass backup channel. Nanosecond switching is achieved through an optocoupler isolation drive circuit. The module's built-in wireless communication node is networked based on the Zigbee-3.0 protocol, transmitting voltage and current data to a central controller in real time. The array's underlying battery cluster uses a MOSFET matrix topology, with an FPGA controlling the PWM signal to dynamically adjust the battery series-parallel configuration. A single module failure triggers a cluster-wide reconfiguration mechanism, automatically reorganizing healthy modules into series compensation mode to maintain cluster voltage stability. Step 3: Connect each battery cluster via inter-cluster ring copper busbars. Solid-state contactors and bidirectional current detection circuits are deployed between every two adjacent clusters to form a dynamic parallel path. A central controller calculates the optimal connection pair in real time based on power demand and controls the contactors to establish new paths. The top layer uses a multi-winding, high-frequency, isolated, bidirectional DC-DC converter. The primary winding of its magnetically integrated transformer is independently connected to each battery cluster, and the secondary output is connected in parallel to the DC bus. Phase-shift modulation is used to achieve directional power distribution. Step 4: When multiple modules fail, the system automatically activates three-layer coordinated reconstruction. The failed module switches to bypass mode via a three-port switch, and the central controller updates the charge and discharge schedule based on the system status cycle, improving system cycle efficiency and reducing battery array life attenuation. Step 5: Implement online dynamic replacement of faulty batteries. The system supports electrical isolation, bypass switching, cluster-level reconstruction, dynamic parallel connection between clusters, top-level power directional allocation, as well as capacity migration, parameter alignment, and capacity reorganization of faulty batteries.

2. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1, characterized in that: The three-port power switch unit adopts gallium nitride MOSFET devices and realizes nanosecond switching through an optical coupler isolation drive circuit.

3. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1, characterized in that: The wireless communication nodes are networked based on the Zigbee-3.0 protocol and can transmit voltage and current data to the central controller in real time.

4. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1, characterized in that: The bottom battery cluster adopts MOSFET matrix topology, and the FPGA controls the PWM signal to dynamically adjust the battery series-parallel structure.

5. The reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1 is characterized in that: The middle layer connects the battery clusters through an annular copper busbar, and a solid-state contactor and a current bidirectional detection circuit are deployed between every two adjacent clusters to form a dynamic parallel path.

6. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1, characterized in that: The top layer adopts a multi-winding high-frequency isolated bidirectional DC-DC converter, the primary winding of the magnetic integrated transformer of which is independently connected to each battery cluster, and the secondary winding is output in parallel to the DC bus.

7. A reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1, characterized in that: When multiple modules fail, the system can automatically activate three-layer coordinated reconstruction. The failed module is switched to bypass through a three-port switch, and the charging and discharging plan is periodically refreshed through the central controller.

8. The reconfigurable energy storage system and control method for online dynamic replacement of faulty batteries according to claim 1 is characterized in that: The system calculates the optimal connection pair in real time based on power demand through a central controller, and controls the contactor to establish a new path, thereby achieving directional power distribution and improving system cycle efficiency.

Images