Battery cluster SOC estimation and parameter self-learning method and system based on double SOCs

Abstract

The invention discloses a battery cluster SOC (State of Charge) estimation and parameter self-learning method and system based on double SOCs (State of Charge). According to the method, high-frequency real-time estimation and low-frequency parameter learning are decoupled, high-precision RSOC is obtained through CCV calibration of multi-condition matching in a high-frequency process, the VSOC is updated by adopting a dual-mode asymmetric strategy based on a VSOC / RSOC difference value and a relative size, and accurate, smooth and quick-response SOC output is achieved; the low-frequency process is operated independently, and the maximum capacity of the battery is updated asynchronously after complete charge-discharge circulation is detected. According to the SOC estimation method, self-learning and updating of key parameters are realized while the real-time performance and the stability of SOC estimation are ensured through a separated architecture of'real-time state tracking 'and'low-speed model correction', and the overall precision, the robustness and the adaptability of SOC estimation of an energy storage system in long-term operation are fundamentally improved.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, specifically to a method and system for estimating the SOC and self-learning parameters of a battery cluster based on dual SOC. Background Technology

[0002] State of Charge (SOC) is a core parameter of battery management systems, and its estimation accuracy directly affects the operating efficiency, safety, and lifespan of energy storage systems. Traditional SOC estimation methods, such as the ampere-hour integration method, are susceptible to drift due to initial errors and the accumulation of current measurement errors; while calibration methods based on open-circuit voltage (OCV) require batteries to be stored for extended periods, making them difficult to apply in dynamically operating energy storage scenarios.

[0003] In existing technologies, filtering or smoothing algorithms are often used to process the real-time estimated SOC (which can be called RSOC) to obtain a smoothed SOC value (which can be called VSOC) for display and control. However, conventional smoothing algorithms (such as low-pass filtering) have a contradiction between response lag and poor dynamic characteristics: when battery current or state changes abruptly, VSOC may not be able to track the actual change in charge in time, or a "false high" phenomenon may appear at the end of charging, affecting user trust and system control. In addition, how to efficiently and accurately use the closed-circuit voltage (CCV) to calibrate RSOC in real time during dynamic charging and discharging processes, especially at different rates, is also a technical challenge.

[0004] Therefore, there is an urgent need for a SOC estimation method that can balance estimation accuracy, response speed, and smooth stability, and is applicable to dynamic multi-rate energy storage scenarios. Summary of the Invention

[0005] In view of this, the present invention provides a battery cluster SOC estimation method and system based on dual SOC, aiming to solve the problem that it is difficult to balance accuracy, responsiveness and smoothness in SOC estimation under dynamic scenarios in the prior art.

[0006] In a first aspect, the present invention provides a battery cluster SOC estimation and parameter self-learning method based on dual SOC, including a high-frequency SOC estimation process and a low-frequency parameter self-learning process executed in parallel: The high-frequency SOC estimation process includes: Step S1, Data Collection and Processing: Collect the operating status data of the battery cluster and preprocess the data to remove outliers; Step S2, Working Status Determination and Calibration Data Collection: Determine the current working status of the battery cluster, and store the voltage data into the corresponding sliding data window according to the working status and charge / discharge rate; Step S3, CCV calibration: Based on the data in the sliding data window, query the pre-stored CCV-SOC mapping table to obtain the reference SOC value RSOC; the CCV-SOC mapping table stores the mapping relationship between SOC and voltage calibration point under different charge and discharge rates, and the voltage range of the voltage calibration point is compensated in the current direction according to the charge and discharge rate. Step S4, Dual-mode VSOC Update: Compare the absolute value of the difference between the current VSOC and the RSOC with a first threshold: Step S41: If the absolute value of the difference is less than or equal to the first threshold, then perform VSOC calibration in the fully vented state. Step S42: If the absolute value of the difference is greater than the first threshold, then execute the VSOC dynamic update strategy. The strategy uses an asymmetric rule to smoothly update the VSOC based on the relative size relationship between the current VSOC and RSOC; then perform VSOC calibration in the full-discharge state. The low-frequency parameter self-learning process is executed asynchronously, independently of the high-frequency SOC estimation process, and includes: Step S5, Capacity Calibration: After detecting that the battery cluster has completed a full charge-discharge cycle that meets the preset conditions, update the maximum available capacity of the battery cluster and provide the updated capacity value to the high-frequency SOC estimation process.

[0007] Preferably, the specific process of querying the CCV-SOC mapping table in step S3 includes: Traverse the calibration points in the CCV-SOC mapping table until a calibration point that satisfies all of the following conditions is found: The system status recorded at the calibration point is consistent with the current operating status of the battery cluster; The current current of the battery cluster is within the current range corresponding to the calibration point, and the current range is determined according to ±5% of the calibrated charge and discharge rate; If the calibration point is a full charge or empty charge calibration point, the extreme voltage of the battery cluster is within the voltage range corresponding to that calibration point; If the calibration point is a normal calibration point, then more than half of the voltage data in the sliding data window are within the voltage range corresponding to the calibration point.

[0008] Preferably, the VSOC dynamic update strategy in step S42 is as follows: When the difference between VSOC and RSOC satisfies the first directional condition, the first update sub-policy is executed; When the difference between VSOC and RSOC satisfies the second directional condition, the second update sub-policy is executed; Wherein, the update magnitude of the first update sub-strategy is greater than or equal to that of the second update sub-strategy.

[0009] More preferably, the first directional condition is that VSOC lags behind RSOC; the first update sub-strategy is: if the absolute value of a single change in RSOC is within a preset second threshold range, then VSOC is updated to completely follow the single change, and within a certain period after the update, an additional fixed granularity is added to VSOC at fixed time intervals.

[0010] More preferably, the second directional condition is that VSOC leads or equals RSOC; the second update sub-strategy is: query the preset shaving ratio according to the current current of the battery cluster; if the absolute value of the single change of RSOC is within the preset second threshold range, then let VSOC be updated in proportion to the shaving ratio, partially following the single change.

[0011] Preferably, the deceleration ratio is related to the absolute value of the current; the smaller the absolute value of the current, the smaller the deceleration ratio.

[0012] Preferably, in step S41, the VSOC calibration under the fully charged and discharged state means that when the battery cluster voltage reaches the voltage range corresponding to the fully charged or discharged calibration point, the VSOC is directly calibrated to the SOC value corresponding to the calibration point.

[0013] Preferably, in step S2, the working state includes at least a charging state, a discharging state, and a resting state, and the sliding data window is established according to different charge and discharge rates.

[0014] Preferably, step S5 in the low-frequency parameter self-learning process is executed during system idle periods or low-priority tasks, and a smoothing filtering algorithm is used when updating the capacity value to avoid abrupt interference to the high-frequency SOC estimation process.

[0015] In a second aspect, the present invention provides a battery management system, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any of the first aspects.

[0016] Beneficial effects The technical solution provided by this invention, by introducing a dual SOC mechanism (RSOC and VSOC) and designing a collaborative update strategy, achieves the following beneficial effects: High-precision real-time calibration: Through the CCV calibration method with multi-condition matching (state, current range, voltage window), the closed-circuit voltage can be used to perform high-precision and adaptive real-time correction of RSOC during battery dynamic operation, effectively suppressing the cumulative error of ampere-hour integration.

[0017] Balancing intelligent smoothing and rapid response: The designed dual-mode VSOC update mechanism activates an asymmetric dynamic update strategy when there is a large deviation between VSOC and RSOC. This strategy allows for rapid catching up or even auxiliary increases when VSOC is "lagging behind," while slowing down and following according to the current ratio when VSOC is "leading." This perfectly resolves the inherent contradiction between smoothness, response speed, and prevention of false highs at the end, making the displayed SOC both stable and accurate.

[0018] Strong robustness: Multiple protection mechanisms are set up throughout the process, such as limiting the single change of RSOC and judging the statistical data of sliding window, which effectively filter out the interference of sensor noise and occasional jumps on SOC estimation and improve the reliability of the system.

[0019] Clear and reliable architecture: By decoupling high-frequency state estimation from low-frequency parameter learning, the interference of critical parameter mutations on the real-time system is avoided, making the overall architecture clearer and the system operation more stable and reliable, which conforms to the design principles of highly reliable embedded systems.

[0020] Highly practical for engineering applications: The method fully considers the characteristics of multi-rate operation of energy storage batteries. Through the multi-rate sliding window and voltage and current compensation mechanism, the algorithm can maintain good performance under different load conditions and is suitable for complex practical energy storage application scenarios. Attached Figure Description

[0021] Figure 1 This is a basic design flowchart of a battery cluster SOC estimation method based on dual SOC provided in an embodiment of the present invention.

[0022] Figure 2 This is a detailed flowchart of CCV calibration (step S3) in one embodiment of the present invention.

[0023] Figure 3 This is the main flowchart of dual-mode VSOC update (step S4) in one embodiment of the present invention.

[0024] Figure 4 This is a detailed flowchart of the VSOC dynamic update strategy (step S42) in a charging state according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0026] Example 1 See Figure 1 This embodiment provides a battery cluster SOC estimation method based on dual SOC, the basic design process of which includes: Step 101: The system periodically collects operating status data such as individual cell voltage, total bus voltage, total bus current, and temperature of the battery cluster. Obvious outliers are removed through data validity verification and filtering algorithms (such as median filtering). The extreme values ​​of individual cell voltages within the cluster, the average extreme values ​​of the module voltages, and the current charge / discharge rate C-rate are calculated.

[0027] Step 102: Determine the current operating state of the battery cluster based on the current direction (positive for charging, negative for discharging, and quiescent when close to zero). Based on the current state and the calculated voltage rating, store the average extreme value of the module voltage in a pre-established sliding data window for that voltage rating. Different voltage ratings have independent sliding windows used to store historical voltage data at that rating; the window length can be set as needed (e.g., 30 seconds).

[0028] Step 103: Execute the CCV calibration procedure. Use the data in the sliding window to query the pre-stored CCV-SOC mapping table to obtain a high-precision reference SOC value, i.e., RSOC. This step is used to periodically correct the accumulated errors of methods such as ampere-hour integration.

[0029] Step 104: Execute the dual-mode VSOC update process. Compare the current smoothed SOC value (VSOC) used for display / control with the RSOC obtained in step 103. Select different update paths based on whether the absolute value of the difference between the two exceeds a threshold (e.g., 1%).

[0030] Simultaneously, the system runs an independent low-frequency parameter self-learning process: step 105. This process continuously monitors the battery's historical operating data. When it detects that the battery cluster has completed a full charge-discharge cycle that meets quality requirements (e.g., from low SOC to high SOC, then discharge to low SOC, with stable and valid current and voltage data throughout), it triggers a capacity calibration calculation. This calculation is performed during system idle or low-priority tasks, and by analyzing the changes in net charge-discharge ampere-hour capacity and start-end SOC of the cycle, it calculates the updated maximum usable capacity value. To maintain system stability, the new capacity value is smoothly updated to the parameters used in the high-frequency SOC estimation process through methods such as first-order low-pass filtering, thereby progressively correcting the benchmark of the ampere-hour integration method and maintaining long-term accuracy.

[0031] See Figure 2 The procedure for step 103 (CCV calibration) is described in detail.

[0032] Step 201: Obtain the current system status (charging / discharging / resting), charge / discharge rate C, and sliding window data.

[0033] Step 202: Determine whether the historical data within the sliding window needs to be cleared (e.g., when the status or multiplier changes drastically). If so, clear and then collect again.

[0034] Step 203: Query the total number N of calibration points at the current magnification C from the CCV-SOC mapping table.

[0035] Step 204: Initialize the loop counter i = 0.

[0036] Step 205: Judge the loop condition. When i < N, perform the following matching judgment.

[0037] Step 206: Judge whether the current system state is consistent with the state recorded in the ith calibration point. If not, jump to Step 213, increment i by 1, and continue the next loop.

[0038] Step 207: Judge whether the current magnification mode is a single current mode (that is, this calibration point only corresponds to a specific current value). If so, skip the current interval judgment and directly execute Step 209.

[0039] Step 208: If not, then judge whether the current actual current I is within the current interval [C0.95, C1.05] defined by this calibration point. If not, jump to Step 213.

[0040] Step 209: Judge whether the ith calibration point is an end calibration point for full charge or full discharge. If so, execute Step 210; if not, execute Step 211.

[0041] Step 210: Judge whether the extreme voltage (the highest or lowest single-cell voltage) of the current battery cluster is within the voltage interval [V_min_i, V_max_i] defined by this calibration point. The center value of the voltage interval has been compensated in the current direction according to the magnification during calibration. If not, jump to Step 213.

[0042] Step 211: For ordinary calibration points, judge whether more than half of the voltage data stored in the sliding data window fall within the voltage interval [V_min_i, V_max_i] of the calibration point. If not, jump to Step 213.

[0043] Step 212: The matching is successful. Return the SOC value corresponding to this calibration point as RSOC, and end the process after abnormal calibration processing.

[0044] Step 213: Increment the loop counter i by 1, and return to Step 205 for matching the next calibration point.

[0045] See Figure 3 , which details the main process of Step 104 (dual-mode VSOC update).

[0046] Step 301: Save the current VSOC granularity information (such as the last update time, additional cumulative value, etc.).

[0047] Step 302: Determine whether the battery cluster is in the charge / discharge state and calculate |VSOC - RSOC|.

[0048] Step 303: Determine whether |VSOC - RSOC| is greater than 1%. Here, 1% is the first threshold. If the difference is greater than 1%, execute Step 304; if the difference is less than or equal to 1%, directly jump to Step 305; Step 304: According to whether it is the charging state or the discharging state currently, respectively jump to execute the corresponding dynamic update sub-process (charging state strategy or discharging state strategy), and then jump to Step 305. The charging and discharging states adopt an asymmetric update strategy with the same logic, only symmetrically reversing the definitions of "leading" and "lagging".

[0049] Step 305: Perform VSOC calibration in the full / empty state. For example, when the system determines that the battery is fully charged (the voltage reaches the full charge calibration point), directly set VSOC to 100% (or the calibration point SOC); the same applies when it is empty.

[0050] Step 306: The update process ends, and the final VSOC value is output for system display and control.

[0051] See Figure 4 , taking the charging state as an example, details the VSOC dynamic update strategy (charging state strategy) in Step 304.

[0052] Step 401: Determine whether VSOC is less than RSOC (i.e., VSOC lags behind the actual battery level) in the charging state.

[0053] Step 402: If VSOC < RSOC (lagging state), enter the "catch-up mode". First, determine whether the single change amount |ΔRSOC| of RSOC is within the limit range (the second threshold, such as 0.5%). If so, execute Step 403; if not, jump to Step 404.

[0054] Step 403: VSOC is updated completely following the current change amount of RSOC: VSOC_new = VSOC_old + ΔRSOC.

[0055] Step 404: Start the time-assisted catch-up mechanism. Query the cumulative time of the additional change of VSOC since the last update. Then, VSOC is additionally increased at a preset tiny granularity (such as 0.1%) and a fixed time interval (such as 1 second) to ensure that in the constant current charging stage, even if RSOC remains unchanged temporarily due to model errors, the progress bar can advance smoothly.

[0056] Step 405: If it is determined in step 401 that VSOC ≥ RSOC (leading or equal state), enter the "anti-false high mode". First, based on the current charging current I, look up the preset speed reduction ratio mapping table to obtain the speed reduction ratio r (0 <r ≤ 1)。

[0057] Step 406: Determine if |ΔRSOC| is within the limit. If yes, proceed to step 407; otherwise, proceed to step 408.

[0058] Step 407: VSOC partially follows RSOC changes with a reduction ratio r: VSOC_new = VSOC_old + r ΔRSOC. The reduction ratio r is negatively correlated with the current magnitude, that is, the smaller the current (such as at the end of a constant voltage circuit), the smaller the value of r, and the slower the VSOC increases, effectively preventing the displayed power from being falsely high.

[0059] Step 408: The process ends, and the updated VSOC value is returned. The strategy for discharge state is completely symmetrical; only the "less than" and "greater than or equal to" conditions need to be swapped.

[0060] Example 2 This embodiment provides a battery management system (BMS) including a processor and a memory. The memory stores a computer program containing instructions for executing the battery cluster SOC estimation method based on dual SOC described in Embodiment 1. By executing these instructions, the processor achieves accurate estimation, display, and control of the battery cluster SOC.

[0061] Example 3 This embodiment provides an energy storage system, which includes one or more battery clusters and a battery management system as described in Embodiment 2, which is electrically and communicatively connected to each battery cluster. This energy storage system can be applied to various scenarios such as grid-side energy storage, user-side energy storage, and integrated wind-solar-energy storage.

[0062] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for estimating the SOC and self-learning parameters of a battery cluster based on dual SOC, characterized in that, This includes a parallel high-frequency SOC estimation process and a low-frequency parameter self-learning process: The high-frequency SOC estimation process includes: Step S1, Data Collection and Processing: Collect the operating status data of the battery cluster and preprocess the data to remove outliers; Step S2, Working Status Determination and Calibration Data Collection: Determine the current working status of the battery cluster, and store the voltage data into the corresponding sliding data window according to the working status and charge / discharge rate; Step S3, CCV calibration: Based on the data in the sliding data window, query the pre-stored CCV-SOC mapping table to obtain the reference SOC value RSOC; the CCV-SOC mapping table stores the mapping relationship between SOC and voltage calibration point under different charge and discharge rates, and the voltage range of the voltage calibration point is compensated in the current direction according to the charge and discharge rate. Step S4, Dual-mode VSOC Update: Compare the absolute value of the difference between the current VSOC and the RSOC with a first threshold: Step S41: If the absolute value of the difference is less than or equal to the first threshold, then perform VSOC calibration in the fully vented state. Step S42: If the absolute value of the difference is greater than the first threshold, then execute the VSOC dynamic update strategy. The strategy uses an asymmetric rule to smoothly update the VSOC based on the relative size relationship between the current VSOC and RSOC; then perform VSOC calibration in the full-discharge state. The low-frequency parameter self-learning process is executed asynchronously, independently of the high-frequency SOC estimation process, and includes: Step S5, Capacity Calibration: After detecting that the battery cluster has completed a full charge-discharge cycle that meets the preset conditions, update the maximum available capacity of the battery cluster and provide the updated capacity value to the high-frequency SOC estimation process.

2. The method according to claim 1, characterized in that, The specific process of querying the CCV-SOC mapping table in step S3 includes: Traverse the calibration points in the CCV-SOC mapping table until a calibration point that satisfies all of the following conditions is found: The system status recorded at the calibration point is consistent with the current operating status of the battery cluster; The current current of the battery cluster is within the current range corresponding to the calibration point, and the current range is determined according to ±5% of the calibrated charge and discharge rate; If the calibration point is a full charge or empty charge calibration point, the extreme voltage of the battery cluster is within the voltage range corresponding to the calibration point; If the calibration point is a normal calibration point, then more than half of the voltage data in the sliding data window are within the voltage range corresponding to the calibration point.

3. The method according to claim 1, characterized in that, The VSOC dynamic update strategy in step S42 is as follows: When the difference between VSOC and RSOC satisfies the first directional condition, the first update sub-policy is executed; When the difference between VSOC and RSOC satisfies the second directional condition, the second update sub-policy is executed; Wherein, the update magnitude of the first update sub-strategy is greater than or equal to that of the second update sub-strategy.

4. The method according to claim 3, characterized in that, The first directional condition is that VSOC lags behind RSOC; The first update sub-strategy is as follows: if the absolute value of a single change in RSOC is within a preset second threshold range, then VSOC is updated to completely follow the single change, and after the update, an additional fixed granularity is added to VSOC at fixed time intervals for a period of time.

5. The method according to claim 3, characterized in that, The second directional condition is that VSOC leads or equals RSOC; The second update sub-strategy is as follows: query the preset shaving ratio based on the current current of the battery cluster; if the absolute value of the single change of RSOC is within the preset second threshold range, then let VSOC be updated in proportion to the shaving ratio, partially following the single change.

6. The method according to claim 5, characterized in that, The deceleration ratio is related to the absolute value of the current; the smaller the absolute value of the current, the smaller the deceleration ratio.

7. The method according to claim 1, characterized in that, In step S41, the VSOC calibration under the fully charged and discharged state means that when the battery cluster voltage reaches the voltage range corresponding to the fully charged or discharged calibration point, the VSOC is directly calibrated to the SOC value corresponding to the calibration point.

8. The method according to claim 1, characterized in that, In step S2, the working state includes at least charging state, discharging state and resting state, and the sliding data window is established according to different charge and discharge rates.

9. The method according to claim 1, characterized in that, Step S5 in the low-frequency parameter self-learning process is executed during system idle periods or low-priority tasks, and a smoothing filtering algorithm is used when updating the capacity value to avoid sudden interference to the high-frequency SOC estimation process.

10. A battery management system, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 9.

Images