Method and device for estimating state of charge of energy storage system against fluctuating current

Abstract

The application discloses a kind of anti-fluctuation current energy storage system state-of-charge estimation method and device.For the problem that energy storage system power fluctuation easily causes estimated value shock, the method obtains real-time charge-discharge current and terminal voltage of battery pack, calculates basic state-of-charge and real-time current rate;Adopt double threshold hysteresis determination mechanism to determine the current operating condition in combination with the state of the previous moment, and trigger the operating condition latching mechanism to lock the determination result under the power critical fluctuation condition;According to the condition, a segmented nonlinear weight distribution strategy is used to distribute dynamic weight coefficients for the first and second state-of-charge estimation algorithms;Finally, based on the weight coefficient, the modified parameters output in parallel by the two algorithms are fused to obtain the final state-of-charge estimation value.The application can alleviate the determination interference caused by high-frequency fluctuation current, realize smooth transition of multi-algorithm correction quantity and stable output of state-of-charge.

Description

Technical Field

[0001] This invention relates to the field of battery state estimation technology, and in particular to a method and apparatus for estimating the state of charge of an energy storage system resistant to fluctuating current. Background Technology

[0002] Electrochemical energy storage systems have wide applications in energy dispatching scenarios such as grid frequency regulation and peak-valley arbitrage. In the operation and daily management of energy storage systems, the state of charge (SOC) of the battery pack is a key parameter characterizing the remaining battery capacity and guiding charge and discharge control strategies. Obtaining relatively accurate SOC estimates provides an important data foundation for the grid-connected operation of energy storage power stations, capacity utilization assessment, and battery pack lifespan management.

[0003] Current energy storage SOC estimation schemes mostly rely on calculation methods such as ampere-hour integration, combined with different state estimation algorithms for segmented correction. Conventional processing logic typically determines operating conditions based on a single set current critical threshold. That is, under low-rate steady-state conditions, one type of filtering algorithm is mainly used to suppress errors and noise, while under high-rate conditions, a different control algorithm is switched to handle the situation. These existing schemes often employ fixed numerical weights or simple linear transition methods, and the switching logic of the state estimation algorithm depends solely on the comparison between the current real-time current and this single threshold. As energy storage systems become more deeply involved in complex grid dispatching, the existing single-judgment logic and algorithm weight allocation mechanisms still need further improvement in adapting to dynamic operating conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for estimating the state of charge (SOC) of an energy storage system that is resistant to fluctuating currents, in order to solve the problems mentioned in the background art, such as the oscillating output of the SOC estimate caused by frequent power fluctuations, and the algorithm correction lag and jump problems that may occur during long-term operation.

[0005] To achieve the above objectives, according to a first aspect of the present invention, a method for estimating the state of charge (SOC) of an energy storage system resistant to fluctuating current is provided, applied to an energy storage battery management system, the method comprising:

[0006] Acquire real-time operating data of the energy storage battery pack, the real-time operating data including at least real-time charge and discharge current and measured terminal voltage; and determine the basic state of charge based on the real-time charge and discharge current, and calculate the real-time current ratio;

[0007] A dual-threshold hysteresis determination mechanism is used to compare the real-time current ratio with a preset first current ratio threshold and a preset second current ratio threshold, and combined with the operating status of the energy storage battery pack at the previous moment to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold.

[0008] If the current operating condition is determined to be a critical power fluctuation condition, the operating condition latching mechanism is triggered and the determination result of the current operating condition is kept locked for a preset latching time.

[0009] Based on the determined current operating conditions, a piecewise nonlinear weight allocation strategy is used to assign dynamic weight coefficients to the preset first state estimation algorithm and the preset second state estimation algorithm;

[0010] Based on the correction parameters obtained by fusing the first state estimation algorithm and the second state estimation algorithm in parallel using the dynamic weighting coefficients, the basic state of charge is corrected to obtain the final state of charge estimate of the energy storage battery pack.

[0011] In one possible implementation, the current operating condition includes a steady-state low-current condition, a power critical fluctuation condition, and a steady-state high-current condition. The step of using a dual-threshold hysteresis determination mechanism to compare the real-time current ratio with a pre-set first preset current ratio threshold and a second preset current ratio threshold, and combining this with the previous operating state of the energy storage battery pack, to determine the current operating condition of the energy storage battery pack, includes:

[0012] If the real-time current multiplier of the current cycle is less than the first preset current multiplier threshold and is not within the preset latching time, the current operating condition is determined to be the steady-state low-current operating condition.

[0013] If the real-time current multiplier of the current cycle is greater than or equal to the first preset current multiplier threshold and less than the second preset current multiplier threshold, the current operating condition is determined to be the power critical fluctuation condition.

[0014] If the real-time current multiplier of the current cycle is greater than or equal to the second preset current multiplier threshold, the current operating condition is determined to be the steady-state high-current operating condition.

[0015] Specifically, after determining that the system has entered the critical power fluctuation condition, the operating condition latching mechanism is triggered to start timing. Within the preset latching time, the system ignores any leap in the real-time current multiplier to a value less than the first preset current multiplier threshold, and maintains the determination that the current operating condition is the critical power fluctuation condition.

[0016] In one possible implementation, the step of assigning dynamic weight coefficients to the first state estimation algorithm and the second state estimation algorithm using a piecewise nonlinear weight allocation strategy based on the determined current operating condition includes:

[0017] Throughout the entire state of charge estimation period, the dynamic weight coefficients of the first state estimation algorithm are kept constant and greater than zero;

[0018] When the current operating condition is the steady-state low-current condition, the dynamic weight coefficient of the first state estimation algorithm is configured to the highest proportion extreme value, and the dynamic weight coefficient of the second state estimation algorithm is configured to the lowest proportion extreme value.

[0019] When the current operating condition is the critical power fluctuation condition, the dynamic weight coefficient of the first state estimation algorithm is made to decrease nonlinearly according to a preset quadratic proportional function as the difference between the real-time current ratio and the first preset current ratio threshold increases.

[0020] When the current operating condition is the steady-state high-current condition, the dynamic weight coefficient of the second state estimation algorithm is made to increase nonlinearly with the increase of the real-time current ratio according to a preset natural exponential function, and the dynamic weight coefficient of the first state estimation algorithm is reduced accordingly, so as to satisfy that the sum of the dynamic weight coefficients of the two is constant at a preset fixed value.

[0021] In one possible implementation, the first state estimation algorithm is a Kalman filter algorithm; the second state estimation algorithm is a proportional-integral-differential algorithm based on an integral separation mechanism; the process of obtaining the correction parameters of the parallel output of the second state estimation algorithm specifically includes:

[0022] Obtain a pre-configured battery equivalent circuit model, and calculate the estimated terminal voltage at the current moment based on the battery equivalent circuit model, thereby obtaining the preliminary voltage residual between the estimated terminal voltage and the measured terminal voltage;

[0023] A preset threshold limiting operation is performed on the initial voltage residual to obtain the voltage residual after limiting;

[0024] The voltage residual after the limit is input into the proportional-integral-differential algorithm to calculate the single-cycle state of charge correction amount, wherein the integral separation mechanism is configured to block the accumulation operation of the error integral term when the absolute value of the voltage residual after the limit is greater than the preset integral separation threshold.

[0025] The calculated single-cycle state of charge correction is limited to a preset upper limit and output as a correction parameter for the second state estimation algorithm.

[0026] In one possible implementation, calculating the estimated terminal voltage at the current moment based on the battery equivalent circuit model includes:

[0027] A first-order RC polarization model is adopted as the equivalent circuit model of the battery.

[0028] A recursive relationship for polarization voltage is established based on model parameters calibrated in advance through pulse testing;

[0029] By combining the open-circuit voltage of the energy storage battery pack with the recursive relationship between the polarization voltage, the estimated terminal voltage adapted to the long-term operating characteristics of the energy storage system is calculated.

[0030] In one possible implementation, the step of determining the base state of charge based on the real-time charge-discharge current includes:

[0031] The ampere-hour integration method is used to perform integral calculations of the basic state based on the real-time charge and discharge current;

[0032] Based on the current charge / discharge depth data and ambient temperature data of the energy storage battery pack, the integral charge / discharge efficiency of the base state is compensated and corrected.

[0033] In one possible implementation, the method further includes a long-term operating error calibration mechanism, which includes:

[0034] If the continuous static time of the energy storage battery pack reaches the preset static time threshold and the absolute value of its real-time charging and discharging current is less than the preset static current threshold, a secondary calibration operation is performed according to the pre-calibrated and stored mapping table of open circuit voltage and state of charge to eliminate the cumulative estimation error generated by the energy storage battery pack during long-term operation.

[0035] According to a second aspect of the present invention, a state-of-charge estimation device for an energy storage system resistant to fluctuating current is provided, comprising:

[0036] The data acquisition module is configured to acquire real-time operating data of the energy storage battery pack, wherein the real-time operating data includes at least real-time charging and discharging current and measured terminal voltage;

[0037] The basic processing module is configured to determine the basic state of charge based on the real-time charge and discharge current and to calculate the real-time current ratio.

[0038] The operating condition determination module is configured to use a dual-threshold hysteresis determination mechanism to compare the real-time current ratio with a preset first current ratio threshold and a preset second current ratio threshold, and combine the operating state of the energy storage battery pack at the previous moment to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold.

[0039] The anti-oscillation latch module is configured to trigger a condition latching mechanism and maintain the determination result of the current operating condition in a locked state for a preset latching time when the current operating condition is determined to be a critical power fluctuation condition.

[0040] The dynamic weight allocation module is configured to allocate dynamic weight coefficients to a preset first state estimation algorithm and a preset second state estimation algorithm based on the determined current operating conditions using a piecewise nonlinear weight allocation strategy.

[0041] The nonlinear fusion module is configured to perform correction processing on the basic state of charge based on the correction parameters output in parallel from the first state estimation algorithm and the second state estimation algorithm, which are fused with the dynamic weight coefficients, so as to obtain the final state of charge estimate of the energy storage battery pack.

[0042] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0043] At least one processor; and

[0044] A memory communicatively connected to the at least one processor, the memory being used to store executable instructions;

[0045] The processor executes the executable instructions to implement the energy storage system state-of-charge estimation method for resisting fluctuating current as described in any of the possible implementations of the first aspect above.

[0046] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the method for estimating the state of charge of an energy storage system resistant to fluctuating current as described in any possible implementation of the first aspect above.

[0047] The above-described one or more technical solutions in the embodiments of this application have at least one or more of the following technical effects:

[0048] This invention provides a method for estimating the state of charge (SOC) of an energy storage system to withstand fluctuating current. Addressing the instability caused by frequent power fluctuations in energy storage systems, this invention introduces a dual-threshold hysteresis determination mechanism combined with operating condition latching logic. This effectively filters out state determination disturbances caused by high-frequency repetitive current jumps near critical thresholds. This mechanism avoids repeated system jumps under critical operating conditions, improving the system's anti-interference capability and overall stability in the face of complex dynamic operating conditions. Simultaneously, this invention employs a piecewise nonlinear weight allocation strategy, dynamically configuring weight coefficients based on the first and second state estimation algorithms operating in parallel under the determined operating conditions. Compared to conventional numerical hard switching, this nonlinear weighted fusion mechanism can follow real-time current changes, achieving a smooth transition of multi-algorithm correction parameters and effectively mitigating the step jump phenomenon in SOC estimation values ​​caused by sudden changes in operating conditions or algorithm switching. In summary, this solution achieves smooth and stable SOC output under all operating conditions, providing robust data support for the long-term reliable operation and high-precision energy dispatch of energy storage systems.

[0049] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0050] Figure 1 This is a schematic flowchart of a method for estimating the state of charge of an energy storage system resistant to fluctuating current, according to an exemplary embodiment.

[0051] Figure 2 This is a schematic diagram of the module composition structure of a state of charge estimation device for an energy storage system resistant to fluctuating current, according to an exemplary embodiment.

[0052] Figure 3 This is a schematic diagram illustrating the changes in a dual hysteresis threshold and weights for energy storage according to an exemplary embodiment.

[0053] Figure 4 This is a comparison curve of SOC output under fluctuating operating conditions provided according to an exemplary embodiment.

[0054] Figure labeling: 100, Data acquisition module; 200, Basic processing module; 300, Operating condition determination module; 400, Anti-vibration latching module; 500, Dynamic weight allocation module; 600, Nonlinear fusion module. Detailed Implementation

[0055] 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.

[0056] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of systems and methods consistent with some aspects of the invention as detailed in the appended claims.

[0057] Figure 1 Here is a flowchart of a method for estimating the state of charge (SOC) of an energy storage system resistant to fluctuating current, provided according to an exemplary embodiment. Figure 1 As shown, the method is applied to an energy storage battery management system, and the method includes:

[0058] In step S100, real-time operating data of the energy storage battery pack is acquired. This real-time operating data includes at least real-time charging / discharging current and measured terminal voltage. Based on the real-time charging / discharging current, the baseline state of charge is determined, and the real-time current ratio is calculated. Specifically, the energy storage battery management system continuously acquires the physical state information of the energy storage battery pack through a built-in high-precision sampling circuit, current sensor, and voltage acquisition module, at a preset sampling frequency, such as once every 100 milliseconds or a shorter sampling period. This real-time operating data serves as the fundamental input for all subsequent algorithm corrections and operating condition determinations. The real-time charging / discharging current characterizes the current energy flow intensity of the battery pack, while the measured terminal voltage reflects the polarization response and external characteristics of the battery pack under the current current.

[0059] Furthermore, after acquiring the real-time charge and discharge current, the system first uses basic calculation logic such as the ampere-hour integration method to determine the basic state of charge (SOC). Specifically, the system combines the initial SOC value saved at the previous moment with the real-time charge and discharge current within the current sampling period to obtain a preliminary change in charge, and updates the current SOC accordingly. This SOC can reflect the battery's charge trend under ideal current fluctuations, but in scenarios of long-term operation or drastic current fluctuations, dynamic correction is still required in subsequent stages.

[0060] Based on this, the system quantifies the severity of the current operating condition by calculating the real-time current ratio. The real-time current ratio is defined as the ratio of the absolute value of the real-time charge / discharge current to the rated capacity of the energy storage battery pack. The purpose of calculating the real-time current ratio is to normalize the operating intensity of energy storage battery packs of different specifications and capacities, thereby providing a standardized basis for the subsequent dual-threshold hysteresis judgment mechanism. By converting the real-time current into a ratio form, the judgment error caused by the difference in individual battery cells can be effectively avoided, improving the versatility and adaptability of the algorithm in energy storage systems of different scales.

[0061] Preferably, during the acquisition of real-time operating data and the determination of the basic state of charge, the system also preprocesses the raw data. For example, digital filtering technology is used to filter out high-frequency glitches and noise in the current sampling, or temperature compensation correction is performed on the measured terminal voltage to ensure that the data input to the subsequent state of charge estimation stage has a high degree of confidence. The accurate acquisition of these underlying physical data and the preliminary calculation of the basic electrical quantity constitute the solid data foundation of the anti-fluctuation current estimation scheme of the present invention.

[0062] In step S200, a dual-threshold hysteresis determination mechanism is used to compare the real-time current ratio with a pre-set first preset current ratio threshold and a second preset current ratio threshold, and combined with the previous operating state of the energy storage battery pack, to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold. Specifically, to address the problem of misjudgment of operating conditions and repeated state reversals that are easily caused by conventional single-threshold determination logic in complex power grid dispatch scenarios, this embodiment introduces a dual-threshold hysteresis determination mechanism with physical hysteresis characteristics. The core criterion of this mechanism no longer relies solely on the real-time current ratio at the current moment, but deeply integrates the previous operating state of the energy storage battery pack, thereby constructing a state transition buffer at the control logic level. Since the second preset current ratio threshold is numerically strictly greater than the first preset current ratio threshold, a dead zone that can accommodate current fluctuations is naturally formed between the two.

[0063] Furthermore, the underlying logic of this hysteresis determination mechanism is as follows: When the energy storage battery pack was operating in a low-rate steady-state condition at the previous moment and the system power demand suddenly increases, even if the real-time current rate exceeds the first preset current rate threshold, the system will not immediately switch the operating condition determination result. At this time, the system will maintain the determination of the low-rate steady-state condition until the real-time current rate continues to climb and strictly exceeds the higher second preset current rate threshold, at which point the system will confirm that the current operating condition has transitioned to a high-rate dynamic condition. Conversely, when the energy storage battery pack was in a high-rate dynamic condition at the previous moment and the system power demand begins to decline, the real-time current rate must drop below the first preset current rate threshold before the system will revert the current operating condition back to the low-rate steady-state condition. Through this nonlinear hysteresis logic similar to a Schmitt trigger, the algorithm determination instability caused by the high-frequency shuttling of the real-time current near a certain critical point can be avoided.

[0064] Preferably, the design of the aforementioned dual-threshold hysteresis determination mechanism fully aligns with the physical hysteresis effect of internal chemical reactions and ion migration in electrochemical batteries. In practical energy storage applications, the establishment and dissipation of the internal polarization voltage of the battery are not instantaneous but exhibit significant inertia and time delay. For example, the system can set the first preset current rate threshold to 0.5C and the second preset current rate threshold to 0.8C. The buffer zone established using these two specific values ​​can effectively filter out pseudo-state switching caused by measurement noise from brief power spikes or high-frequency sampling. This comprehensive determination method, which combines historical operating states, effectively improves the anti-interference capability and robustness of operating condition identification, and is expected to achieve a smoother determination output that closely matches the actual physical boundaries of the battery, thus laying a solid foundation for the stable allocation of subsequent model algorithm weights.

[0065] In step S300, if the current operating condition is determined to be a critical power fluctuation condition, a condition latching mechanism is triggered, and the determination result of the current operating condition is kept locked for a preset latching time. Specifically, to address the high-frequency current oscillation phenomenon that may occur when energy storage systems participate in the grid's rapid frequency response or cope with impact loads, this embodiment further enhances the robustness of the determination logic by introducing a time-dimensional condition latching mechanism. The critical power fluctuation condition refers to a specific state where the real-time current ratio fluctuates between a first preset current ratio threshold and a second preset current ratio threshold, or within the hysteresis interval formed by the two, in a short-term, high-frequency back-and-forth motion. Under such extreme conditions, even with the aforementioned dual-threshold hysteresis mechanism, if the fluctuation frequency is extremely high, the algorithm may still frequently switch unnecessarily between steady-state and dynamic logic.

[0066] Furthermore, the operating condition latching mechanism essentially constructs a dynamic time masking window at the control algorithm level. Once the system detects that the current has entered a critical power fluctuation state, it will immediately start a timing task and forcibly freeze the current operating condition judgment output within a preset latching duration. For example, the preset latching duration can be calibrated based on the typical scheduling cycle of the energy storage power station or the polarization response constant of the battery cell, and is usually set to a fixed value between 500 milliseconds and 2000 milliseconds, such as 1000 milliseconds. Within this latching period, no matter how drastically the real-time current ratio changes, the system maintains the current operating condition judgment result unchanged until the timer returns to zero and the current fluctuation tends to smooth out, only then is the next round of state update process allowed.

[0067] Preferably, the application of the operating condition latching mechanism can effectively filter out false judgment signals caused by electrical measurement errors, sampling asynchrony, or transient electromagnetic interference. By introducing this logical self-holding function at the software level, nonlinear low-pass filtering of the complex underlying physical signals is achieved, eliminating the "ping-pong effect" at the algorithm level. This design not only effectively improves the system's judgment stability under extremely complex operating conditions but also provides a secondary guarantee for the stability of subsequent weight allocation coefficients, and is expected to achieve a more continuous and stepless operating condition identification effect, thereby ensuring that the entire state of charge estimation architecture still has extremely high determinism in dynamic environments.

[0068] In step S400, based on the determined current operating condition, a piecewise nonlinear weight allocation strategy is used to assign dynamic weight coefficients to the preset first state estimation algorithm and the preset second state estimation algorithm. Specifically, during the operation of the energy storage system, different state estimation algorithms exhibit significant differences in adaptability to various operating conditions. For example, the first state estimation algorithm may focus more on the convergence of global errors and noise suppression in the underlying equivalent circuit model, and is generally suitable for low-rate smooth operation; while the second state estimation algorithm may have stronger nonlinear gain adjustment capabilities and perform better in dealing with high-frequency polarization and high-rate dynamic response. To effectively improve the comprehensive estimation accuracy under all operating conditions, this embodiment abandons the traditional fixed-value hard switching logic and instead introduces a piecewise nonlinear weight allocation strategy. This strategy can dynamically assign corresponding proportional weights to the parallel first state estimation algorithm and the second state estimation algorithm in real time according to the actual evolution trend of the current operating condition.

[0069] Furthermore, to ensure the reliable operation of the underlying logic of complex systems, this piecewise nonlinear weight allocation strategy typically relies on a pre-defined mathematical mapping function or multi-dimensional lookup table interpolation logic. For example, the system can pre-set a function such as an S-shaped smoothing curve or an exponential decay function as the mapping benchmark for weight allocation in its control kernel. When it is determined that the system's current operating condition is smoothly transitioning from a low-rate steady state to a high-rate dynamic state, the dynamic weight coefficient of the first state estimation algorithm does not drop abruptly, but gradually decreases along a pre-defined nonlinear curve trajectory; simultaneously, the dynamic weight coefficient of the second state estimation algorithm rises gently along a corresponding complementary nonlinear growth trajectory. Within any given computational cycle, the sum of the dynamic weight coefficients allocated by the two algorithms is always strictly equal to 100%.

[0070] Preferably, the aforementioned piecewise nonlinear weight allocation strategy effectively decouples the performance limitations of a single algorithm under extreme conditions, resulting in excellent continuity and smoothness in the switching transition of the underlying algorithm. By transforming discrete operating condition determination results into continuously changing weight coefficients, the data oscillations and error accumulation induced by sudden changes in operating conditions or direct switching of control models are resolved. Furthermore, considering the physical hysteresis characteristics of energy storage batteries with different chemical systems, key node parameters of the nonlinear mapping curve, such as inflection point positions and transition slopes, can be optimized through prior offline calibration or adaptive fine-tuning during operation, which is expected to achieve a more accurate and seamless model synergy effect.

[0071] In step S500, the basic state of charge (SOC) is corrected based on the correction parameters obtained by fusing the parallel outputs of the first and second state estimation algorithms using the dynamic weighting coefficients, to obtain the final SOC estimate of the energy storage battery pack. Specifically, in the underlying computational kernel of the energy storage battery management system, the first and second state estimation algorithms maintain parallel and independent data processing processes. Due to the difference in the settings of the underlying observation model and the core gain matrix, these two algorithms calculate the first and second correction parameters, representing the convergence trends of their respective states, based on real-time operating data acquired at the same time. After obtaining the dynamic weighting coefficients allocated in the preceding steps, the system executes the underlying mathematical weighted summation logic. That is, the system directly multiplies the first correction parameter with its corresponding dynamic weighting coefficient, and simultaneously multiplies the second correction parameter with its corresponding dynamic weighting coefficient, then linearly accumulates the results of the two products to extract a comprehensive correction parameter that fully integrates the advantages of low-rate steady-state noise immunity and high-rate dynamic following.

[0072] Furthermore, after completing the underlying fusion calculation of the multi-algorithm correction parameters, the system accurately superimposes this comprehensive correction parameter onto the previously acquired basic state of charge through feedback compensation. At the physical level, this process is equivalent to utilizing the closed-loop correction capability of advanced state estimation algorithms to dynamically smooth out the long-term cumulative drift error generated by the open-loop ampere-hour integral model. Since this correction parameter is derived from a smooth transition based on nonlinear dynamic weights, its superposition will not cause abrupt impacts on the original basic state of charge.

[0073] Preferably, through the aforementioned flexible fusion processing based on dynamic weights, this embodiment not only achieves smooth decoupling and seamless handover of parameters from multiple observer algorithms at the underlying mathematical logic level, but also mitigates the risk of step jumps in the state of charge (SOC) estimation caused by frequent changes in operating conditions at the practical engineering level. This closed-loop architecture combining qualitative judgment and quantitative fusion effectively improves the fault tolerance of energy storage systems when facing long-term extreme scheduling. The final SOC estimation value output after this joint correction processing is expected to achieve a smooth, continuous estimation accuracy that closely approximates the actual physical limits of the battery across the entire operating range, thereby providing an extremely stable and solid data foundation for the safe grid connection, accurate capacity calculation, and high-level energy management strategies of energy storage systems.

[0074] This invention effectively filters out state determination disturbances caused by high-frequency reciprocating current crossings near critical thresholds by introducing a dual-threshold hysteresis determination mechanism combined with operating condition latching logic. This mechanism avoids repeated system jumps under critical operating conditions, improving the system's anti-interference capability and overall stability in the face of complex dynamic operating conditions. Simultaneously, this invention employs a piecewise nonlinear weight allocation strategy, dynamically configuring weight coefficients based on the first and second state estimation algorithms operating in parallel according to the determined operating conditions. Compared to conventional numerical hard switching modes, this nonlinear weighted fusion mechanism can follow real-time current changes, achieving a smooth transition of multi-algorithm correction parameters and effectively mitigating the step jump phenomenon in state of charge estimation caused by sudden changes in operating conditions or algorithm switching. In summary, this scheme achieves smooth and stable output of state of charge under all operating conditions, providing solid data support for the long-term reliable operation and high-precision energy dispatch of energy storage systems.

[0075] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0076] In an exemplary embodiment, the current operating condition includes a steady-state low-current operating condition, a power critical fluctuation operating condition, and a steady-state high-current operating condition. The step of using a dual-threshold hysteresis determination mechanism to compare the real-time current ratio with a pre-set first preset current ratio threshold and a second preset current ratio threshold, and combining this with the operating state of the energy storage battery pack at the previous moment, to determine the current operating condition of the energy storage battery pack, includes:

[0077] If the real-time current multiplier of the current cycle is less than the first preset current multiplier threshold and is not within the preset latching time, the current operating condition is determined to be the steady-state low-current operating condition.

[0078] If the real-time current multiplier of the current cycle is greater than or equal to the first preset current multiplier threshold and less than the second preset current multiplier threshold, the current operating condition is determined to be the power critical fluctuation condition.

[0079] If the real-time current multiplier of the current cycle is greater than or equal to the second preset current multiplier threshold, the current operating condition is determined to be the steady-state high-current operating condition.

[0080] Specifically, after determining that the system has entered the critical power fluctuation condition, the operating condition latching mechanism is triggered to start timing. Within the preset latching time, the system ignores any leap in the real-time current multiplier to a value less than the first preset current multiplier threshold, and maintains the determination that the current operating condition is the critical power fluctuation condition.

[0081] Specifically, to accurately depict the actual operating state of energy storage systems in complex grid interactions, this invention scientifically divides the entire lifecycle operating conditions into steady-state low-current conditions, power critical fluctuation conditions, and steady-state high-current conditions. These three conditions correspond to typical physical scenarios of energy storage power stations, such as standby and gradual power replenishment, high-frequency regulation transition, and high-power peak-valley arbitrage. By comparing the real-time current ratio with two thresholds exhibiting physical hysteresis characteristics, the invention effectively reduces the misjudgment of the state and system oscillations that are easily caused by a single threshold at the boundary.

[0082] Furthermore, if the real-time current multiplier of the current cycle is less than the first preset current multiplier threshold, for example, 1.8C, and the system is not currently within the preset latching duration, then the system determines that the current operating condition is the steady-state low-current condition. This underlying logic design ensures that the state machine will only switch to steady-state mode after the system has truly escaped the high-frequency fluctuation state and the internal polarization has become relatively flat. This mechanism avoids erroneous algorithm switching caused by brief current dips, providing a stable premise for the system to fully leverage its high-precision noise immunity advantages at low current multipliers.

[0083] Furthermore, if the real-time current multiplier of the current cycle is greater than or equal to the first preset current multiplier threshold and less than the second preset current multiplier threshold, for example, 2.2C, the system determines that the current operating condition is the critical power fluctuation condition. This condition essentially establishes a physical buffer zone between high and low current multipliers. When the current is in this range, it usually means that the system is responding to dynamic scheduling commands, at which point the underlying parameters change drastically and are highly susceptible to electromagnetic interference. Separating this into a critical fluctuation zone helps the subsequent system configure dedicated smoothing weights for it, thereby effectively improving the continuity and smoothness during multi-algorithm handover.

[0084] Preferably, if the real-time current rate of the current cycle is greater than or equal to the second preset current rate threshold, the system determines that the current operating condition is the steady-state high-current condition. Under this condition, ion migration inside the energy storage battery is intense, and physical polarization is significantly amplified. By clearly defining this high-rate range, the system can provide a precise trigger criterion for the subsequent rapid introduction of an anti-saturation correction module, which is expected to achieve a more timely and effective countermeasure against the cumulative error and polarization drift caused by high rates.

[0085] Specifically, the core underlying logic of the operating condition latching mechanism triggered after determining that the system has entered the critical power fluctuation condition lies in the introduction of forced shielding in the time dimension. Once the system state machine transitions to the critical power fluctuation condition, the underlying hardware timer or software counter immediately starts counting. Within the preset latching duration, even if the real-time current ratio briefly jumps to below the first preset current ratio threshold due to grid harmonics, high-frequency noise, or sudden changes in scheduling commands, the system will actively cut off and ignore these downslope signals, forcibly maintaining the current operating condition as the critical power fluctuation condition. This underlying processing method, which introduces physical impedance decoupling, is equivalent to performing nonlinear low-pass filtering on the state machine's transition commands, eliminating repeated jumps in current under critical conditions, fundamentally preventing high-frequency start-stop of the algorithm, and providing a solid state guarantee for the long-term high-reliability operation of the energy storage system.

[0086] In an exemplary embodiment, the step of assigning dynamic weight coefficients to the first state estimation algorithm and the second state estimation algorithm using a piecewise nonlinear weight allocation strategy based on the determined current operating condition includes:

[0087] Throughout the entire state of charge estimation period, the dynamic weight coefficients of the first state estimation algorithm are kept constant and greater than zero;

[0088] When the current operating condition is the steady-state low-current condition, the dynamic weight coefficient of the first state estimation algorithm is configured to the highest proportion extreme value, and the dynamic weight coefficient of the second state estimation algorithm is configured to the lowest proportion extreme value.

[0089] When the current operating condition is the critical power fluctuation condition, the dynamic weight coefficient of the first state estimation algorithm is made to decrease nonlinearly according to a preset quadratic proportional function as the difference between the real-time current ratio and the first preset current ratio threshold increases.

[0090] When the current operating condition is the steady-state high-current condition, the dynamic weight coefficient of the second state estimation algorithm is made to increase nonlinearly with the increase of the real-time current ratio according to a preset natural exponential function, and the dynamic weight coefficient of the first state estimation algorithm is reduced accordingly, so as to satisfy that the sum of the dynamic weight coefficients of the two is constant at a preset fixed value.

[0091] Specifically, to visually demonstrate the evolution of the underlying computing power allocation across all operating conditions, please refer to [link / reference needed]. Figure 3 , Figure 3 This diagram illustrates the changes in the dual hysteresis threshold and weights for energy storage provided in this embodiment of the invention. Addressing the issue of abrupt changes in the estimated state of charge caused by hard start-stop of algorithm modules in conventional hierarchical correction architectures, this embodiment establishes a core benchmark for globally smooth transitions in the underlying weight allocation logic. For example... Figure 3 As shown in the diagram, the relationship between the current multiplier on the horizontal axis and the weight ratio on the vertical axis indicates that throughout the entire state-of-charge estimation cycle, the system forcibly maintains the dynamic weight coefficients of the first state estimation algorithm, namely the Kalman filter algorithm, at a constant greater than zero. This design, which keeps the basic filtering module running throughout the entire cycle, effectively provides a persistent anti-interference foundation for the underlying data flow, ensuring output smoothness during long-term continuous operation.

[0092] Furthermore, combined Figure 3 In the left-hand interval shown, when the current operating condition is the steady-state low-current condition (i.e., the real-time current rate is less than 1.8C), the internal physical polarization of the battery is weak. The system configures the dynamic weight coefficient of the first state estimation algorithm to its highest extreme value of 1, and the dynamic weight coefficient of the second state estimation algorithm (i.e., the proportional-integral-derivative algorithm) to its lowest extreme value of 0. This extreme value configuration ensures that the system fully relies on a filtering algorithm with excellent noise reduction capabilities during the low-power charging phase, and is expected to achieve more stable steady-state static accuracy.

[0093] Furthermore, such as Figure 3 As shown in the transition interval of the middle section, when the current operating condition is the critical power fluctuation condition, i.e., the real-time current ratio is between 1.8C and 2.2C, the underlying current begins to transition between high and low ratio thresholds. To achieve the smooth weight transfer and crossover shown in the figure, the system causes the dynamic weight coefficient of the first state estimation algorithm to nonlinearly decay as the difference between the real-time current ratio and the first preset current ratio threshold increases. The specific weight allocation formula is as follows:

[0094]

[0095] Meanwhile, the formula for calculating the dynamic weight coefficients of the second state estimation algorithm is as follows:

[0096]

[0097] In the formula, The dynamic weights characterize the first state estimation algorithm at time k. The dynamic weights characterize the second-state estimation algorithm at time k. It characterizes the real-time current rate. By setting the denominator of the transition range to 0.4C and the buffer coefficient to 0.2, a specific nonlinear gradual change effect is formed, which closely matches the slow establishment process of the battery polarization voltage in this range and effectively filters out measurement glitches at the critical point.

[0098] Preferably, such as Figure 3 As shown in the extended interval on the right, when the current operating condition transitions to the steady-state high-current condition (i.e., the real-time current rate is greater than or equal to 2.2C), ion migration inside the energy storage battery becomes intense, and the system must respond quickly to the model drift caused by high-rate polarization. At this time, the system causes the dynamic weight coefficients of the second state estimation algorithm to increase non-linearly with the increase of the real-time current rate according to a preset natural exponential function. The specific formula for the natural exponential function is:

[0099]

[0100] The dynamic weight coefficient of the first state estimation algorithm is reduced synchronously, and the specific formula is as follows:

[0101]

[0102] In the formula, the nonlinear mapping function containing the natural exponential term has an extremely fast initial response slope, which can quickly increase the proportion of the algorithm with anti-integral saturation capability in the early stage of high current operation. In the above dynamic allocation process of all operating conditions, the sum of the dynamic weight coefficients of the two is strictly satisfied to be a preset fixed value of 1.

[0103] Finally, after obtaining continuous and oscillation-free dynamic weighting coefficients for the corresponding operating conditions, the system fuses the final state-of-charge parameters output in parallel by the two algorithms, and the fusion formula strictly follows:

[0104]

[0105] In the formula, This represents the final estimated state of charge output. This represents the numerical value output by the first-state estimation algorithm. This represents the numerical value output by the second-state estimation algorithm. Through the aforementioned multi-dimensional collaborative weight allocation and fusion mechanism, the impact of sudden changes in power grid dispatch commands on the algorithm model is mitigated, providing a solid data foundation for the reliable grid connection of energy storage systems.

[0106] In an exemplary embodiment, the first state estimation algorithm is a Kalman filter algorithm; the second state estimation algorithm is a proportional-integral-differential algorithm based on an integral separation mechanism; the process of obtaining the correction parameters of the parallel output of the second state estimation algorithm specifically includes:

[0107] Obtain a pre-configured battery equivalent circuit model, and calculate the estimated terminal voltage at the current moment based on the battery equivalent circuit model, thereby obtaining the preliminary voltage residual between the estimated terminal voltage and the measured terminal voltage;

[0108] A preset threshold limiting operation is performed on the initial voltage residual to obtain the voltage residual after limiting;

[0109] The voltage residual after the limit is input into the proportional-integral-differential algorithm to calculate the single-cycle state of charge correction amount, wherein the integral separation mechanism is configured to block the accumulation operation of the error integral term when the absolute value of the voltage residual after the limit is greater than the preset integral separation threshold.

[0110] The calculated single-cycle state of charge correction is limited to a preset upper limit and output as a correction parameter for the second state estimation algorithm.

[0111] Specifically, to balance the embedded computing power and long-term operational accuracy of the energy storage management system, the pre-configured battery equivalent circuit model in this embodiment is preferably a first-order polarization model. Compared to higher-order models, this model has lower computational complexity and stronger long-term stability. The system first uses this model to calculate the estimated terminal voltage at the current moment, and its underlying calculation logic follows the following formula:

[0112]

[0113] In the formula, The model-predicted terminal voltage at the current moment. The open-circuit voltage is characterized by a lookup table based on the fundamental state of charge derived from the ampere-hour integral. Characterizing real-time charge and discharge current, Characterizing the ohmic internal resistance, The polarization voltage is characterized. The recursive process of the polarization voltage combines the state at the previous moment with the physical polarization time constant; the specific calculation formula is as follows:

[0114]

[0115] In the formula, Characterizing the sampling period, Characterizing the polarization time constant, Characterizing the polarization internal resistance. After obtaining the model-predicted terminal voltage, the system directly subtracts it from the measured terminal voltage to obtain the preliminary voltage residual: In the formula, Characterizing the initial voltage residual, This characterizes the measured terminal voltage. This preliminary residual can objectively reflect the degree of deviation between the current physical model and the actual battery operating state.

[0116] Furthermore, after obtaining the initial voltage residual, to prevent abnormal amplification of the residual value due to high-rate current surges or instantaneous sampling interference, the system will forcibly execute a preset threshold limiting operation. Specifically, the system compares the initial voltage residual with the set residual limiting threshold, strictly confining it within a reasonable physical trust range. The calculation logic for the voltage residual after limiting is as follows:

[0117]

[0118] In the formula, Characterizing the voltage residual after limiting, This represents the voltage residual limiting threshold. This mathematical constraint cuts off the path for abnormal noise to propagate to the subsequent closed-loop algorithm.

[0119] Preferably, after the voltage limiting process is completed, the system inputs the voltage residual after limiting into the proportional-integral-differential algorithm (PIDA), which serves as the second state estimation algorithm. For the high-rate, long-duration charge-discharge conditions unique to energy storage systems, conventional error integral terms are prone to severe integral saturation, leading to algorithm response lag or even divergence. Therefore, this scheme innovatively introduces an integral separation mechanism. When the system determines that the absolute value of the voltage residual after limiting is greater than a preset integral separation threshold, this threshold reuses the limiting threshold, immediately blocking further accumulation of the error integral term at the underlying level; conversely, when the residual is within a controllable range, the integral term is normally introduced to eliminate steady-state error. The complete formula for calculating the single-cycle state-of-charge correction is as follows:

[0120]

[0121] In the formula, Characterizes the single-cycle state of charge correction amount. , , The proportional, integral, and derivative coefficients are represented respectively. This anti-saturation design effectively improves the system's rapid correction capability and robustness under high-power dynamic response.

[0122] Specifically, to completely eliminate abrupt changes in the estimated state of charge (SOC) value that might be caused by a single computational iteration, the system implements a safety limiting interception logic before outputting the correction amount. This means that the absolute value of the calculated single-cycle SOC correction must be limited to a preset upper limit.

[0123]

[0124] In the formula, This represents the upper limit of the single-cycle state of charge correction. If the calculated value exceeds this limit, the upper limit boundary value will be forcibly used.

[0125] Finally, the system directly superimposes this correction, which is subject to both strict physical and mathematical constraints, onto the current fundamental state of charge, thereby outputting the correction parameters for the second state estimation algorithm:

[0126]

[0127] In the formula, The correction parameters characterize the output of the second-state estimation algorithm. It characterizes the base state of charge determined by ampere-hour integration. This integrated processing flow, which combines integral separation and multi-level limiting, is fully adapted to the high-reliability, long-term operation requirements of energy storage systems under fluctuating current environments.

[0128] Furthermore, the calculation of the estimated terminal voltage at the current moment based on the battery equivalent circuit model includes:

[0129] A first-order RC polarization model is adopted as the equivalent circuit model of the battery.

[0130] A recursive relationship for polarization voltage is established based on model parameters calibrated in advance through pulse testing;

[0131] By combining the open-circuit voltage of the energy storage battery pack with the recursive relationship between the polarization voltage, the estimated terminal voltage adapted to the long-term operating characteristics of the energy storage system is calculated.

[0132] Specifically, considering the limited computing power of the embedded platform at the bottom of the energy storage battery management system and the objective physical scenario of long-term operation of electrochemical energy storage power stations, this embodiment abandons high-order equivalent circuit models, such as second-order or third-order RC polarization models, which have extremely high computational complexity and are prone to state overfitting. Instead, it preferably adopts a first-order RC polarization model as the underlying battery equivalent circuit model. This equivalent circuit model physically consists of an ideal voltage source, an ohmic internal resistance, and a series-connected RC network composed of a polarization resistor and a polarization capacitor connected in parallel. This architecture not only reduces the floating-point computation of the control chip but also accurately and stably characterizes the core physical polarization characteristics of the energy storage battery during smooth charging and discharging and long-term cycling, exhibiting high engineering adaptability to the operating characteristics of the energy storage system itself.

[0133] Furthermore, to ensure that the model parameters accurately reflect the underlying electrochemical state, the system establishes a recursive relationship for polarization voltage in the discretized time domain based on the core model parameters obtained through offline calibration using hybrid power pulse capability characteristic tests. Specifically, this recursive logic incorporates the physical hysteresis effect of battery polarization establishment and dissipation, and deduces the current polarization voltage by combining the polarization state of the previous moment with the real-time current excitation of the current cycle. Its recursive formula is expressed as:

[0134]

[0135] In the formula, The polarization voltage at time k, which is the current time, is represented. Characterizing the polarization voltage at the previous moment, Characterizes the sampling period set by the underlying hardware. Characterizes the real-time charge and discharge current after filtering. The polarization time constant, which is determined by both polarization resistance and polarization capacitance, is used to characterize the polarization time constant. The polarization resistance is characterized. Through the above recursive calculations including the natural exponential decay term, the underlying model can effectively capture and quantify the nonlinear accumulation and release evolution of polarization voltage under complex fluctuating currents.

[0136] Preferably, after obtaining the polarization voltage reflecting the dynamic response characteristics, the system integrates it with the static thermodynamic reference of the energy storage battery pack to calculate the final estimated terminal voltage. Specifically, the system first uses the baseline state of charge obtained by the ampere-hour integration method and then accurately retrieves the open-circuit voltage corresponding to the current charge state through multi-dimensional lookup table interpolation. Subsequently, the system combines the transient voltage drop generated by the real-time current across the ohmic internal resistance with the polarization voltage output by the aforementioned recursive network to complete the closed-loop solution of the entire first-order RC polarization model. The specific calculation formula for the estimated terminal voltage is as follows:

[0137]

[0138] In the formula, This represents the model-predicted terminal voltage obtained from the calculation at the current moment. Characterized by the open-circuit voltage mapped from the fundamental state of charge. The ohmic internal resistance obtained from the calibration is characterized. Through this inference mechanism that deeply integrates the physical characteristics of energy storage batteries, the output predicted terminal voltage is expected to achieve extremely high long-term stability, effectively avoiding the cumulative drift phenomenon that is prone to occur in high-order parameter models under long-term continuous operation. This provides solid benchmark data support for obtaining high-confidence voltage residuals and performing multi-algorithm joint correction.

[0139] In an exemplary embodiment, the step of determining the base state of charge based on the real-time charge-discharge current includes:

[0140] The ampere-hour integration method is used to perform integral calculations of the basic state based on the real-time charge and discharge current;

[0141] Based on the current charge / discharge depth data and ambient temperature data of the energy storage battery pack, the integral charge / discharge efficiency of the base state is compensated and corrected.

[0142] Specifically, in the underlying logic of an energy storage battery management system, the ampere-hour integral method is the core foundation for building a basic energy ledger. The system calculates the net change in energy entering and leaving the battery cluster within each sampling period by continuously integrating the high-precision real-time charge and discharge current over time. This net change is then superimposed on the initial state of the battery pack or the previous state of charge, thus achieving real-time updates and projections of the basic state. This fundamental calculation method, based on the principle of charge conservation, can objectively reflect the physical increase and decrease trends of energy under ideal steady-state current. However, for energy storage systems that frequently participate in complex energy dispatching processes such as grid peak shaving and frequency regulation, the actual charge and discharge efficiency of the battery is not an ideal, fixed constant, but rather dynamically drifts with changes in the internal electrochemical state and the external environment.

[0143] Furthermore, to address the long-term cumulative error problem caused by relying solely on current-time integration, this embodiment introduces multi-dimensional compensation and correction calculations in the basic calculation stage. The system acquires and combines the current charge / discharge depth data and ambient temperature data of the energy storage battery pack in real time through the underlying sensor network. Those skilled in the art will understand that the chemical activity of lithium-ion insertion / extraction within the battery varies significantly under different ambient temperatures, such as -20°C or 45°C; simultaneously, the different degrees of battery polarization within different charge / discharge depth ranges also lead to significant nonlinear changes in its coulombic efficiency. Therefore, the system dynamically adjusts the coefficients of the integral charge / discharge efficiency during the basic state calculation process using a pre-set temperature and depth compensation mapping matrix within the chip, thereby suppressing the estimation divergence caused by open-loop integration at its physical source.

[0144] Preferably, considering the specific requirements of long-term continuous operation of the energy storage system, the system is configured with adaptive correction logic under boundary conditions in addition to the aforementioned continuous compensation. When the energy storage battery pack operates to a fully charged and idle stage or a discharged and cut-off state, which have clear physical boundaries, the system's underlying layer will actively trigger a full-cluster calibration mechanism. This full-cluster calibration mechanism effectively clears the long-term accumulated errors caused by the small zero drift of the current sensor or the small deviation of efficiency compensation by re-anchoring the exact energy reference at this time. Through this underlying maintenance strategy composed of dynamic efficiency correction in the operating state and forced zeroing in the boundary state, the accuracy of the basic data under long-term continuous operation is guaranteed, thus providing an extremely robust and highly reliable reference platform for subsequent multi-algorithm weighted fusion to deal with power fluctuation conditions.

[0145] In an exemplary embodiment, the method further includes a long-term operating error calibration mechanism, the long-term operating error calibration mechanism comprising:

[0146] If the continuous static time of the energy storage battery pack reaches the preset static time threshold and the absolute value of its real-time charging and discharging current is less than the preset static current threshold, a secondary calibration operation is performed according to the pre-calibrated and stored mapping table of open circuit voltage and state of charge to eliminate the cumulative estimation error generated by the energy storage battery pack during long-term operation.

[0147] Specifically, given the complex and lengthy operation cycle of energy storage power stations, even the most precise dynamic closed-loop algorithms struggle to completely avoid the long-term cumulative drift caused by minute zero drift of underlying hardware sensors or the slow aging of polarization model parameters. Therefore, this embodiment, in addition to the dynamic weighted fusion architecture, introduces a long-term operational error calibration mechanism based on the battery's underlying thermodynamic equilibrium state. During the short period when the battery is charging / discharging or has just finished operating, there is a significant ohmic voltage drop and lingering polarization voltage within it. The measured terminal voltage at this time cannot accurately reflect the battery's true intrinsic state. Therefore, the system continuously monitors the battery's resting state through underlying logic. Only when the continuous resting time reaches a preset resting time threshold, such as 2 hours or longer, and the absolute value of the real-time charging / discharging current is strictly limited to less than the preset resting current threshold (i.e., physically approaching 0 amperes), does the system determine that the ion concentration gradient and physical polarization effect within the energy storage battery pack have been sufficiently dissipated, and the battery has returned to a stable thermodynamic equilibrium state.

[0148] Furthermore, after confirming that the energy storage battery pack fully meets the aforementioned stringent physical resting boundary conditions, the measured terminal voltage obtained by the high-precision voltage sampling circuit can be confidently equated to the current true open-circuit voltage. Subsequently, the system actively triggers a secondary calibration operation, calling upon the open-circuit voltage-state-of-charge mapping table (i.e., the conventional open-circuit voltage curve) that has been pre-calibrated and stored in the battery management system's storage unit through offline pulse testing. Through precise table lookup and interpolation calculations, the system can directly map this true open-circuit voltage to a reference state of charge with absolute physical reference value. The system will then use this reference state of charge to forcibly overwrite or deeply integrate and compensate the estimated values ​​currently circulating in the system, thereby instantly clearing away the minute errors accumulated over long-term operation.

[0149] Preferably, this mechanism, which fully utilizes the energy storage system for static secondary calibration during grid dispatch intervals, provides a robust data correction baseline for the entire anti-fluctuation current estimation method. This mechanism cleverly combines the inherent operating and dormant rhythms of the energy storage power station, eliminating the need for additional manual intervention or equipment shutdown. Through this dynamic-static combined global correction strategy, the risk of long-term cumulative error divergence is mitigated, and it is expected to achieve high-precision data output with no cumulative drift or abrupt changes throughout the entire lifecycle. This provides long-term and stable state support for accurate capacity metering of the energy storage system and reliable grid dispatch.

[0150] To visually demonstrate the actual estimation performance of the embodiments of the present invention in dealing with complex dynamic operating conditions of energy storage systems, please refer to [link / reference]. Figure 4 , Figure 4 The SOC output comparison curve under fluctuating operating conditions is provided in the embodiments of the present invention.

[0151] Specifically, when energy storage systems participate in energy dispatching scenarios such as grid frequency regulation and peak-valley arbitrage, the charging and discharging power frequently alternates between high and low rates, which can easily cause the underlying current to repeatedly cross near a preset critical threshold. For example Figure 4 As shown in the figure, the vertical axis represents the percentage value of the state of charge (SOC), and the horizontal axis represents the continuous running time or sampling sequence. The figure intuitively presents two sets of comparative data: the curve with large fluctuations, frequent sawtooth-like jumps, and even obvious step jumps represents the SOC output effect of the existing conventional single threshold judgment and linear / hard switching schemes; while the curve with a smooth overall trend and a stable and continuous gradual change represents the final SOC output effect after adopting the scheme of the embodiment of the present invention.

[0152] Furthermore, combined Figure 4The comparative evolution trend clearly shows that in the critical power fluctuation range, existing technologies inevitably lead to abrupt changes and integral saturation in the underlying voltage residual due to the repeated start-stop and mechanical switching of the filtering algorithm and proportional-integral-derivative correction module, which in turn directly causes high-frequency oscillations in the SOC estimation value. This highly unstable output data not only seriously interferes with the high-level decision-making of the battery management system, but also fails to support precise energy dispatch on the grid side.

[0153] Preferably, such as Figure 4 As shown in the smooth curve, this embodiment of the invention innovatively introduces a dual-threshold hysteresis judgment latch mechanism at the underlying level, filtering out the repeated switching of algorithms triggered by fluctuating current from the physical source. Simultaneously, in conjunction with a piecewise nonlinear weighting function, it achieves seamless handover and smooth transition of multi-algorithm computing power without disabling the basic Kalman filter. Therefore, even under extreme charging and discharging fluctuation conditions, the final state-of-charge estimate output by this invention can still strictly maintain a smooth trajectory without oscillations or jumps. This highly stable nonlinear fusion output performance completely eliminates the filter accumulation drift under long-term continuous high-power operation, strictly controlling the estimation error under steady-state low-current conditions to an extremely low level (e.g., ≤1.7%), while controlling the error under critical power fluctuations and high-rate conditions within an excellent tolerance range (e.g., ≤2.1%). This set of measured comparison curves fully demonstrates the excellent anti-interference capability and high-precision estimation level of this invention's solution in long-term energy storage operation scenarios, providing an extremely solid data verification foundation for the large-scale application of various electrochemical energy storage power stations.

[0154] In an exemplary embodiment, please refer to Figure 2 The present invention also provides a state of charge estimation device for an energy storage system resistant to fluctuating current, comprising:

[0155] The data acquisition module 100 is configured to acquire real-time operating data of the energy storage battery pack, wherein the real-time operating data includes at least real-time charging and discharging current and measured terminal voltage.

[0156] The basic processing module 200 is configured to determine the basic state of charge based on the real-time charge and discharge current and to calculate the real-time current ratio.

[0157] The operating condition determination module 300 is configured to use a dual-threshold hysteresis determination mechanism to compare the real-time current ratio with a preset first current ratio threshold and a preset second current ratio threshold, and combine the operating state of the energy storage battery pack at the previous moment to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold.

[0158] The anti-oscillation latch module 400 is configured to trigger a condition latching mechanism and maintain the determination result of the current operating condition in a locked state for a preset latching time when the current operating condition is determined to be a power critical fluctuation condition.

[0159] The dynamic weight allocation module 500 is configured to allocate dynamic weight coefficients to a preset first state estimation algorithm and a preset second state estimation algorithm according to the determined current operating conditions using a piecewise nonlinear weight allocation strategy.

[0160] The nonlinear fusion module 600 is configured to perform correction processing on the basic state of charge based on the correction parameters output in parallel by fusing the first state estimation algorithm and the second state estimation algorithm with the dynamic weight coefficient, so as to obtain the final state of charge estimate of the energy storage battery pack.

[0161] Specifically, the energy storage system state-of-charge estimation device for resisting fluctuating current provided in this embodiment is a concrete instance of the aforementioned estimation method within an energy storage battery management system. The data acquisition module 100, serving as the underlying hardware interface and data sensing source of the entire device, uses a high-frequency sampling current sensor and voltage acquisition circuit to retrieve the basic physical parameters of the energy storage battery pack in real time. This continuous, uninterrupted real-time operating data provides the most original and highly accurate data foundation for the logical deduction of all subsequent control modules.

[0162] Furthermore, the basic processing module 200 receives the data stream from the data acquisition module 100. On one hand, it performs ampere-hour integration through its internal integration logic to establish an initial energy change ledger, i.e., the basic state of charge. On the other hand, the module uses its internal arithmetic unit to divide the absolute current value by the rated capacity of the energy storage battery pack, thereby calculating the standardized real-time current ratio. This design, which performs parameter normalization processing directly at the underlying level, improves the device's universal adaptability in energy storage power stations of different capacities and scales.

[0163] Furthermore, the operating condition determination module 300 constitutes the core state machine of the device. It internally incorporates a dual-threshold hysteresis determination mechanism, setting a first preset current multiplier threshold, for example, 1.8C, and a second preset current multiplier threshold, for example, 2.2C. This module no longer employs mechanical single-point numerical comparison but instead establishes a state buffer register to record the operating state at the previous moment. By combining the current real-time current multiplier's crossing direction with historical states, this module can construct a physical buffer at the underlying logic level, effectively filtering current spikes caused by high-frequency grid scheduling and outputting extremely stable current operating condition determination commands.

[0164] Preferably, to cope with extremely severe electromagnetic interference and high-frequency power fluctuations, the anti-oscillation latch module 400 is configured as an independent, cooperative safety valve for the operating condition determination module 300. When the system detects that it has entered a critical power fluctuation condition, the latch module immediately activates its internal hardware timer or software countdown timer. Within the preset latch duration, the module forcibly cuts off and shields the response to the operating condition jump interruption at the software architecture level, maintaining the determination result in a locked state. This modular design, which introduces time-dimensional impedance decoupling, eliminates repeated jumps and frequent starts and stops of the subsequent main correction module from the underlying hardware architecture.

[0165] Specifically, the dynamic weight allocation module 500 is responsible for the scheduling of computing power and weights across the entire operating range. This module has pre-set piecewise nonlinear mapping curve functions for different operating conditions. Based on the received stable operating condition instructions, the module continuously and smoothly allocates dynamic weight coefficients to the parallel-running first state estimation algorithm (Kalman filtering) and the second state estimation algorithm (anti-saturation proportional-integral-differential algorithm). For example, under high-rate operating conditions, it can quickly increase the proportion of the second state estimation algorithm according to a preset natural exponential function. This nonlinear weight allocation strategy effectively alleviates the divergence of control system parameters and data step jumps caused by hard switching of algorithms in conventional devices.

[0166] Furthermore, the nonlinear fusion module 600, as the final output terminal of the device, performs low-level mathematical multiplication and addition operations on the correction parameters independently solved by the aforementioned parallel algorithms and their corresponding dynamic weight coefficients. Its internal core operation unit executes the following weighted fusion formula:

[0167]

[0168] This module incorporates the fused, comprehensively corrected parameters into the baseline state of charge, thereby outputting the final estimated state of charge value of the energy storage battery pack. Through the highly collaborative and hierarchically decoupled architecture of the aforementioned multiple modules, this device not only achieves smooth handover of parameters from multiple observer algorithms at the underlying computational logic level, but also mitigates the risk of estimation oscillations caused by long-term charging and discharging and sudden power changes. It is expected to achieve smooth, continuous estimation accuracy that closely approximates the true physical limits of the battery across the entire operating range.

[0169] In an exemplary embodiment, the present invention also provides an electronic device, comprising:

[0170] At least one processor; and

[0171] A memory communicatively connected to the at least one processor, the memory being used to store executable instructions;

[0172] The processor executes the executable instructions to implement the energy storage system state of charge estimation method for resisting fluctuating current as described in the above embodiments.

[0173] Specifically, the electronic device provided in this embodiment can be concretely embodied as the main control unit of the battery management system in an energy storage system or the core edge computing node of the energy management system. The at least one processor serves as the computation and control center of the entire electronic device, and can be, for example, a digital signal processor with powerful floating-point computing capabilities, an advanced microcontroller, or a field-programmable gate array (FPGA) or other underlying hardware chip. When facing the complex charging and discharging conditions of the energy storage system, the processor can efficiently schedule underlying computing power to meet the high-frequency real-time computation requirements of matrix iterative calculations in the Kalman filter algorithm and the integral-separated proportional-integral-differential algorithm, ensuring the timeliness of the underlying control logic execution and data throughput capability.

[0174] Furthermore, the memory and the processor achieve a deep communication connection at the underlying level through an internal high-speed communication interface, such as a serial peripheral interface bus or an integrated circuit built-in bus. For applications involving long-term continuous operation of energy storage systems, the memory includes not only random access memory for high-frequency reading and writing of underlying real-time operating data and intermediate fusion variables, but also flash memory or electrically erasable programmable read-only memory for non-volatile persistent storage. Specifically, the first-order RC polarization model parameters, the mapping table of open-circuit voltage and state of charge obtained from the aforementioned offline pulse test calibration, and the set first and second preset current multiplier thresholds, as well as other core parameters and configuration files, are all securely stored in the non-volatile memory. Simultaneously, a dedicated state register area is provided within this memory to continuously cache the operating state of the previous moment during the operating condition latching period, providing consistent and reliable historical data support for the dual-threshold hysteresis determination mechanism.

[0175] Preferably, after the electronic device is connected to the energy storage battery cluster and put into actual operation, the processor can completely instantiate the aforementioned data processing and logic judgment architecture at the system bottom layer by retrieving and executing the executable instructions line by line from the memory. For example, the processor can obtain voltage and current message data uploaded by the distributed slave control unit in real time through the underlying CAN bus, and then strictly execute the parallel calculation and weighted fusion of multiple algorithms according to the configured piecewise nonlinear mapping function. Through the deep decoupling and collaborative fusion of the underlying hardware architecture and the upper-level nonlinear software algorithm, the electronic device effectively improves the overall anti-interference performance of the system when facing high-frequency power fluctuations in the power grid. It is expected to achieve high-precision, non-jumping stable output of state of charge throughout its entire life cycle, thus providing an extremely reliable physical computing power carrier for the safe grid connection and long-term high-precision scheduling of large-scale electrochemical energy storage power stations.

[0176] In an exemplary embodiment, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for estimating the state of charge of an energy storage system resistant to fluctuating current as described in the above embodiments.

[0177] Specifically, the computer-readable storage medium can be embodied in a non-volatile memory chip within the energy storage battery management system or an external underlying firmware carrier. Those skilled in the art will understand that this computer-readable storage medium includes, but is not limited to, electrically erasable programmable read-only memory, flash memory chips, magnetic random access memory, or any other physical hardware capable of stably embedding the underlying control logic code. These physical carriers provide an extremely reliable static residency space for the aforementioned complex multi-algorithm parallel computing architecture, integral-integral-differential model, and dual-threshold hysteresis determination mechanism, ensuring the data integrity and security of core code instructions under extreme operating conditions such as prolonged power outages or severe electromagnetic interference in the energy storage power station.

[0178] Furthermore, when the computer program embedded on the computer-readable storage medium is loaded and executed by a processor with low-level data processing capabilities, the aforementioned core task flows, such as data acquisition, state evolution, logical judgment, and nonlinear weighted fusion, will be instantiated sequentially within the system's main control software architecture. For example, during the high-frequency polling execution of the low-level program, the system strictly follows the preset instruction set, continuously and without omission fetching real-time charging and discharging current and measured terminal voltage. Once the conditional judgment statement identifies that the real-time current ratio has entered a critical power fluctuation condition, the program will immediately trigger the operating condition latching mechanism code segment in the corresponding memory block, forcibly cutting off the software-level operating condition switching interrupt response. This software-level instruction flow and logical interlocking seamlessly weaves the abstract piecewise nonlinear weight allocation strategy with the underlying first-order polarization model calculation logic into a closed-loop control program with extremely high real-time response capabilities.

[0179] Preferably, by fully encapsulating and burning the method for estimating the state of charge of the energy storage system resistant to fluctuating current into the aforementioned computer-readable storage medium, the engineering portability and system-level deployment efficiency of the core differentiated technical solution of this invention are improved. This medium can be directly embedded into and backward compatible with the motherboards of mainstream energy storage battery management systems with various topologies, systematically resolving the algorithm divergence and step jumps in estimated values ​​that were originally easily caused by high-power peak-valley arbitrage switching or frequent grid frequency regulation commands at the software firmware level. It is expected to achieve smooth, continuous, and interference-resistant high-precision data output throughout the entire cycle, thus laying an extremely solid foundation for the underlying software support of building a highly digitalized and intelligent modern new energy storage power station dispatch network.

[0180] Any aspects of this invention not described in detail are well-known to those skilled in the art.

[0181] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for estimating the state of charge (SOC) of an energy storage system resistant to fluctuating current, applied to an energy storage battery management system, characterized in that, The method includes: Acquire real-time operating data of the energy storage battery pack, the real-time operating data including at least real-time charge and discharge current and measured terminal voltage; and determine the basic state of charge based on the real-time charge and discharge current, and calculate the real-time current ratio; A dual-threshold hysteresis determination mechanism is used to compare the real-time current ratio with a preset first current ratio threshold and a preset second current ratio threshold, and combined with the operating status of the energy storage battery pack at the previous moment to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold. If the current operating condition is determined to be a critical power fluctuation condition, the operating condition latching mechanism is triggered and the determination result of the current operating condition is kept locked for a preset latching time. Based on the determined current operating conditions, a piecewise nonlinear weight allocation strategy is used to assign dynamic weight coefficients to the preset first state estimation algorithm and the preset second state estimation algorithm; Based on the correction parameters obtained by fusing the first state estimation algorithm and the second state estimation algorithm in parallel using the dynamic weighting coefficients, the basic state of charge is corrected to obtain the final state of charge estimate of the energy storage battery pack.

2. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 1, characterized in that, The current operating conditions include steady-state low-current operating conditions, power critical fluctuation operating conditions, and steady-state high-current operating conditions. The dual-threshold hysteresis determination mechanism compares the real-time current ratio with a pre-set first preset current ratio threshold and a second preset current ratio threshold, and combines this with the previous operating state of the energy storage battery pack to determine the current operating condition of the energy storage battery pack, including: If the real-time current multiplier of the current cycle is less than the first preset current multiplier threshold and is not within the preset latching time, the current operating condition is determined to be the steady-state low-current operating condition. If the real-time current multiplier of the current cycle is greater than or equal to the first preset current multiplier threshold and less than the second preset current multiplier threshold, the current operating condition is determined to be the power critical fluctuation condition. If the real-time current multiplier of the current cycle is greater than or equal to the second preset current multiplier threshold, the current operating condition is determined to be the steady-state high-current operating condition. Specifically, after determining that the system has entered the critical power fluctuation condition, the operating condition latching mechanism is triggered to start timing. Within the preset latching time, the system ignores any leap in the real-time current multiplier to a value less than the first preset current multiplier threshold, and maintains the determination that the current operating condition is the critical power fluctuation condition.

3. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 2, characterized in that, The step of assigning dynamic weight coefficients to the first state estimation algorithm and the second state estimation algorithm based on the determined current operating condition using a piecewise nonlinear weight allocation strategy includes: Throughout the entire state of charge estimation period, the dynamic weight coefficients of the first state estimation algorithm are kept constant and greater than zero; When the current operating condition is the steady-state low-current condition, the dynamic weight coefficient of the first state estimation algorithm is configured to the highest proportion extreme value, and the dynamic weight coefficient of the second state estimation algorithm is configured to the lowest proportion extreme value. When the current operating condition is the critical power fluctuation condition, the dynamic weight coefficient of the first state estimation algorithm is made to decrease nonlinearly according to a preset quadratic proportional function as the difference between the real-time current ratio and the first preset current ratio threshold increases. When the current operating condition is the steady-state high-current condition, the dynamic weight coefficient of the second state estimation algorithm is made to increase nonlinearly with the increase of the real-time current ratio according to a preset natural exponential function, and the dynamic weight coefficient of the first state estimation algorithm is reduced accordingly, so as to satisfy that the sum of the dynamic weight coefficients of the two is constant at a preset fixed value.

4. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 1, characterized in that, The first state estimation algorithm is the Kalman filter algorithm; the second state estimation algorithm is the proportional-integral-differential algorithm based on the integral separation mechanism. The process of obtaining the correction parameters of the parallel output of the second state estimation algorithm specifically includes: Obtain a pre-configured battery equivalent circuit model, and calculate the estimated terminal voltage at the current moment based on the battery equivalent circuit model, thereby obtaining the preliminary voltage residual between the estimated terminal voltage and the measured terminal voltage; A preset threshold limiting operation is performed on the initial voltage residual to obtain the voltage residual after limiting; The voltage residual after the limit is input into the proportional-integral-differential algorithm to calculate the single-cycle state of charge correction amount, wherein the integral separation mechanism is configured to block the accumulation operation of the error integral term when the absolute value of the voltage residual after the limit is greater than the preset integral separation threshold. The calculated single-cycle state of charge correction is limited to a preset upper limit and output as a correction parameter for the second state estimation algorithm.

5. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 4, characterized in that, The calculation of the estimated terminal voltage at the current moment based on the battery equivalent circuit model includes: A first-order RC polarization model is adopted as the equivalent circuit model of the battery. A recursive relationship for polarization voltage is established based on model parameters calibrated in advance through pulse testing; By combining the open-circuit voltage of the energy storage battery pack with the recursive relationship between the polarization voltage, the estimated terminal voltage adapted to the long-term operating characteristics of the energy storage system is calculated.

6. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 1, characterized in that, The step of determining the basic state of charge based on the real-time charge and discharge current includes: The ampere-hour integration method is used to perform integral calculations of the basic state based on the real-time charge and discharge current; Based on the current charge / discharge depth data and ambient temperature data of the energy storage battery pack, the integral charge / discharge efficiency of the base state is compensated and corrected.

7. The method for estimating the state of charge of an energy storage system resistant to fluctuating current according to claim 1, characterized in that, The method further includes a long-term operating error calibration mechanism, which includes: If the continuous static time of the energy storage battery pack reaches the preset static time threshold and the absolute value of its real-time charging and discharging current is less than the preset static current threshold, a secondary calibration operation is performed according to the pre-calibrated and stored mapping table of open circuit voltage and state of charge to eliminate the cumulative estimation error generated by the energy storage battery pack during long-term operation.

8. A state-of-charge estimation device for an energy storage system resistant to fluctuating current, characterized in that, include: The data acquisition module is configured to acquire real-time operating data of the energy storage battery pack, wherein the real-time operating data includes at least real-time charging and discharging current and measured terminal voltage; The basic processing module is configured to determine the basic state of charge based on the real-time charge and discharge current and to calculate the real-time current ratio. The operating condition determination module is configured to use a dual-threshold hysteresis determination mechanism to compare the real-time current ratio with a preset first current ratio threshold and a preset second current ratio threshold, and combine the operating state of the energy storage battery pack at the previous moment to determine the current operating condition of the energy storage battery pack, wherein the second preset current ratio threshold is greater than the first preset current ratio threshold. The anti-oscillation latch module is configured to trigger a condition latching mechanism and maintain the determination result of the current operating condition in a locked state for a preset latching time when the current operating condition is determined to be a critical power fluctuation condition. The dynamic weight allocation module is configured to allocate dynamic weight coefficients to a preset first state estimation algorithm and a preset second state estimation algorithm based on the determined current operating conditions using a piecewise nonlinear weight allocation strategy. The nonlinear fusion module is configured to perform correction processing on the basic state of charge based on the correction parameters output in parallel from the first state estimation algorithm and the second state estimation algorithm, which are fused with the dynamic weight coefficients, so as to obtain the final state of charge estimate of the energy storage battery pack.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor, the memory being used to store executable instructions; The processor executes the executable instructions to implement the energy storage system state of charge estimation method for resisting fluctuating current as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for estimating the state of charge of an energy storage system resistant to fluctuating current as described in any one of claims 1 to 7.

Images