Energy storage system state of charge estimation platform and method based on improved unscented Kalman filter

Abstract

The present invention provides a state of charge estimation platform and method for an energy storage system based on an improved unscented Kalman filter, which relates to the technical field of state of charge estimation, including: performing benchmark calibration by identifying static working conditions, matching the static terminal voltage with the OCV-SOC curve to generate a three-dimensional calibration vector; temporally and spatially correlating the real-time current, voltage and battery temperature with the benchmark reference point, predicting the SOC value using an unscented Kalman filter, calculating the time decay weight and spatial distance weight of each candidate benchmark point, and constructing a composite weight coefficient to form a valid reference set; dynamically adjusting the process noise and observation noise covariance matrix of the UKF based on the composite weight coefficient of the valid reference set, and updating the SOC estimate. A dynamic safety assessment domain is established in the SOC-temperature plane, and an early warning is triggered when the estimated value exceeds the confidence interval three times in a row. The present invention improves the accuracy of state of charge estimation and system safety.

Description

Technical Field

[0001] The present invention relates to the technical field of state of charge estimation, and in particular to a platform and method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter. Background Art

[0002] Energy storage technology is increasingly being used in modern energy systems, particularly in the integration of renewable energy and the promotion of electric vehicles. As a core component, batteries require accurate state-of-charge (SOC) estimation, making it crucial to accurately estimate their state of charge (SOC). Traditional SOC estimation methods, such as open-circuit voltage and ampere-hour measurement, often rely on static models and simple algorithms. These methods are unable to dynamically adapt to changes in the battery's operating state and environment, compromising the accuracy of SOC estimation. Furthermore, the battery's internal resistance and polarization voltage vary under varying charge and discharge conditions. Traditional algorithms fail to effectively account for these dynamic factors, resulting in inaccurate battery health assessments and, in turn, impacting battery safety and service life.

[0003] In addition, many existing technologies also have shortcomings in noise processing and anomaly detection. During the battery state estimation process, sensor noise and external interference often lead to data instability, which makes it difficult for the battery management system to make accurate decisions. Existing methods generally pay less attention to how to adjust the covariance of process noise and observation noise in real time to adapt to environmental changes. In addition, the detection mechanism of abnormal conditions is often relatively simple and cannot promptly identify situations where the SOC exceeds the safe range, increasing the risk of battery overheating and excessive discharge. In summary, the shortcomings of traditional state of charge estimation technology in processing dynamic characteristics, noise management and anomaly detection urgently need to be improved through more advanced methods to ensure the safe and reliable operation of energy storage systems.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a platform and method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A state of charge estimation platform for energy storage systems based on an improved unscented Kalman filter, specifically comprising:

[0008] The benchmark reference point calibration module is used to identify the static operating condition and perform benchmark calibration during the operation of the energy storage system. The benchmark SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the benchmark SOC value, battery temperature, and polarization voltage is generated as the benchmark reference point for state estimation.

[0009] The spatiotemporal correlation analysis module is used to perform spatiotemporal correlation between the real-time collected current, voltage, and battery temperature and the benchmark reference points, predict the SOC value through the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate benchmark point, construct a composite weight coefficient for SOC correction, and select candidate benchmark points whose composite weight coefficients meet the conditions to form a valid reference set;

[0010] A dynamic noise adjustment module is used to dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the effective reference set, and update the state vector through the observation equation. The updated state vector includes the SOC estimate, battery temperature and polarization voltage;

[0011] The safety assessment and early warning module is used to establish a dynamic safety assessment domain with the SOC estimation value as the core on the SOC-battery temperature plane. When the SOC estimation value exceeds the confidence interval constructed by the benchmark reference point three times in a row, an early warning is triggered and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are applied to the UKF state estimation of the next working cycle.

[0012] Furthermore, the specific logic of executing the reference point calibration module is: when the absolute value of the current is detected to be less than And the voltage fluctuation rate is less than When it is determined to be in a static condition, the average terminal voltage 30 seconds before the end of the static condition is collected. The closest SOC value is matched as the reference SOC value in the OCV-SOC curve database by binary search method. At the same time, the current battery temperature is measured and the polarization voltage at this time is identified by the least square method. and Finally, a three-dimensional calibration vector containing the reference SOC value, polarization voltage and battery temperature is generated as the reference point for state estimation, and the three-dimensional calibration vector is stored in the reference point database together with its calibration timestamp. The expression is as follows:

[0013] ;

[0014] Where, is the benchmark SOC value, and is the polarization voltage, represents the first polarization voltage generated by the R1-C1 parallel circuit, represents the second polarization voltage generated by the R2-C2 parallel circuit, R1 and R2 represent the resistance in the first polarization voltage generating circuit and the resistance in the second polarization voltage generating circuit, respectively, C1 and C2 represent the capacitance in the first polarization voltage generating circuit and the capacitance in the second polarization voltage generating circuit, respectively, is the battery temperature.

[0015] Furthermore, the real-time collected current, voltage, and battery temperature are temporally and spatially correlated with the benchmark reference point. The specific logic is as follows:

[0016] Establish a spatiotemporal correlation model for the current moment Current ,Voltage and battery temperature , retrieve the time window in the datum reference point database The reference points within the range of 0.01 and with a battery temperature difference of no more than 5°C are selected as candidate reference points. Hour, Indicates the current moment;

[0017] Substitute the current current, voltage and battery temperature into the UKF unscented Kalman filter state equation to obtain the current moment Predicted SOC value;

[0018] For each candidate benchmark point, calculate its corresponding time decay weight and spatial distance weight according to the following formula:

[0019] ;

[0020] Where, For the The time decay weight of candidate reference points, For the The calibration timestamps of candidate benchmark points, is the index of the candidate reference point, is the time decay constant, which is used to control the decay rate of the weight over time. is a natural constant; For the The spatial distance weight of candidate benchmark points, Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature when calibrating the candidate reference points.

[0021] Furthermore, a composite weight coefficient for SOC correction is constructed based on the time decay weight and the spatial distance weight, and the formula is as follows:

[0022] ;

[0023] Where, Indicates the The composite weight coefficient of candidate benchmark points, and Respectively represent The time attenuation weight and spatial distance weight of candidate benchmark points, is the total number of candidate benchmark points;

[0024] Calculate the composite weight coefficient for all candidate benchmark points and select the composite weight coefficient The candidate reference points constitute a valid reference set.

[0025] Furthermore, the process noise covariance matrix and observation noise covariance matrix of UKF are dynamically adjusted based on the composite weight coefficient in the effective reference set, and the SOC estimation value is updated through the observation equation. The specific logic is as follows:

[0026] Based on the composite weight coefficients of each candidate reference point in the valid reference set, the adjustment factor of the process noise covariance matrix is ​​calculated according to the formula:

[0027] ;

[0028] Where, is the adjustment factor of the process noise covariance matrix, Indicates the first The composite weight coefficient of candidate benchmark points, is the index of the candidate reference point in the valid reference set, represents the number of candidate reference points in the valid reference set, Refers to the standard deviation of the composite weight coefficients in the effective reference set;

[0029] Adjustment factors based on the process noise covariance matrix Update process noise covariance matrix :

[0030] ;

[0031] Where, is the initial covariance matrix, is the preset lower limit matrix;

[0032] The adjustment coefficient of the observation noise covariance matrix is ​​calculated based on the distance between the current time and the time at which the candidate reference point in the valid reference set is located. The formula is as follows:

[0033] ;

[0034] Where, is the adjustment coefficient of the observation noise covariance matrix, Indicates the average of the Euclidean distances of all candidate benchmark points; Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature when calibrating the candidate reference points;

[0035] Adjustment coefficients based on the observation noise covariance matrix Update the observation noise covariance matrix :

[0036] ;

[0037] Where, is the initial observation noise covariance.

[0038] Furthermore, the UKF observation equation is used to update the SOC estimate, based on the following formula:

[0039] ;

[0040] Where, For the current moment The observed voltage value, Indicates based on The predicted open circuit voltage, and The function uses the piecewise function corresponding to the highest weighted reference point in the valid reference set. Indicates the current time Predicted SOC value; It is the current moment The current, Indicates the ohmic internal resistance value corresponding to the highest weight reference point, and Represents the current time The first polarization voltage and the second polarization voltage, For the current moment The observation noise follows a normal distribution , is the covariance matrix of the observation noise;

[0041] Calculate the Kalman gain by UKF unscented transformation , and update the state vector, the mathematical expression is as follows: ;

[0042] Where, Indicates at time The updated state vector, including the SOC estimate, battery temperature, and polarization voltage, For the moment The predicted state vector, that is, the state prediction of the previous step, For the moment The calculated Kalman gain is, Indicates at time The actual measured voltage value;

[0043] According to the Kalman gain Update the covariance matrix of the state estimate :

[0044] ;

[0045] Where, Indicates at time The updated state estimation covariance matrix, Indicates at time The predicted state covariance matrix, is the prediction residual covariance matrix, represents the Kalman gain The transpose of .

[0046] Furthermore, the specific logic for executing the security assessment and early warning module is as follows:

[0047] A dynamic safety assessment domain with the SOC estimation as the core is established on the SOC-battery temperature plane. When the SOC estimation value is detected to exceed the confidence interval constructed by the benchmark reference point in three consecutive samplings, it is marked as abnormal and triggers an early warning. The abnormal deviation vector is automatically recorded. The abnormal deviation vector includes the SOC deviation, battery temperature deviation and polarization voltage change for subsequent model parameter correction. The corrected model parameters will be applied to the UKF state estimation of the next working cycle; the safety assessment domain is visualized with a heat map, abnormal points are marked with red pulses and deviation data are displayed; the SOC-battery temperature plane is a two-dimensional coordinate system, The axis is the estimated SOC value, Axis is battery temperature;

[0048] Among them, the SOC deviation refers to the difference between the current SOC estimation value and the SOC mean value in the benchmark reference point, the battery temperature deviation refers to the difference between the currently measured battery temperature and the battery temperature mean value in the benchmark reference point, and the polarization voltage change refers to the difference between the currently measured polarization voltage and the polarization voltage mean value in the benchmark reference point.

[0049] The present invention further provides a method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter. The method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter is obtained by using the above-mentioned energy storage system state of charge estimation platform based on an improved unscented Kalman filter, comprising:

[0050] Step 1: During the operation of the energy storage system, the static operating condition is identified and a baseline calibration is performed. The baseline SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the baseline SOC value, polarization voltage, and battery temperature is generated as the baseline reference point for state estimation.

[0051] Step 2: Temporally and spatially correlate the real-time current, voltage, and battery temperature with the reference points, predict the SOC value using the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate reference point, construct a composite weight coefficient for SOC correction, and select candidate reference points whose composite weight coefficients meet the conditions to form a valid reference set;

[0052] Step 3: Dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the valid reference set, and update the state vector through the observation equation. The updated state vector includes the estimated SOC value, battery temperature and internal resistance;

[0053] Step 4: A dynamic safety assessment domain with the SOC estimate as the core is established on the SOC-temperature plane. When the SOC estimate exceeds the confidence interval established by the benchmark reference point three times in a row, an early warning is triggered and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are then applied to the UKF state estimation in the next working cycle. Compared with the prior art, the present invention has the following advantages:

[0054] The present invention uses dynamic calibration of benchmark reference points and spatiotemporal correlation analysis to make SOC estimation no longer dependent on a fixed model, but can adapt to environmental changes and fluctuations in battery status in a timely manner, thereby reducing estimation errors. Secondly, the introduction of a safety assessment and early warning mechanism can issue an alarm in a timely manner when the SOC exceeds the safe range, reducing the risks of battery overheating, over-discharge, etc., and improving the safety of the system. In addition, by recording and analyzing abnormal deviation vectors, a reliable data basis is provided for the precise correction of subsequent models, further enhancing the intelligence level and operating efficiency of the entire energy storage system. These beneficial effects not only improve the safety and economy of battery use, but also provide new ideas and methods for the optimization of battery management systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a schematic diagram of the platform module of the present invention;

[0056] Figure 2 Schematic diagram of the overall method flow of the present invention;

[0057] Figure 3 and Figure 4 These are the changing curves of time attenuation weight and spatial distance weight respectively. DETAILED DESCRIPTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0059] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0060] Example:

[0061] See also Figure 1 , the present invention provides a technical solution:

[0062] A state of charge estimation platform for energy storage systems based on an improved unscented Kalman filter, specifically comprising:

[0063] The benchmark reference point calibration module is used to identify the static operating condition and perform benchmark calibration during the operation of the energy storage system. The benchmark SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the benchmark SOC value, battery temperature, and polarization voltage is generated as the benchmark reference point for state estimation.

[0064] In this embodiment, the specific logic of executing the reference point calibration module is: when the absolute value of the current is detected to be less than 5 minutes continuously, the reference point calibration module is executed. And the voltage fluctuation rate is less than When it is determined to be in a static condition, the average terminal voltage 30 seconds before the end of the static condition is collected. The closest SOC value is matched as the reference SOC value in the OCV-SOC curve database by binary search method. At the same time, the current battery temperature is measured and the polarization voltage at this time is identified by the least square method. and Finally, a three-dimensional calibration vector containing the reference SOC value, polarization voltage and battery temperature is generated as the reference point for state estimation, and the three-dimensional calibration vector is stored in the reference point database together with its calibration timestamp. The expression is as follows:

[0065] ;

[0066] Where, is the benchmark SOC value, and is the polarization voltage, represents the first polarization voltage generated by the R1-C1 parallel circuit, represents the second polarization voltage generated by the R2-C2 parallel circuit, R1 and R2 represent the resistance in the first polarization voltage generating circuit and the resistance in the second polarization voltage generating circuit, respectively, C1 and C2 represent the capacitance in the first polarization voltage generating circuit and the capacitance in the second polarization voltage generating circuit, respectively, is the battery temperature.

[0067] The benchmark reference point calibration module accurately confirms the SOC value based on the OCV-SOC curve by identifying static operating conditions. The generated three-dimensional calibration vector includes the baseline SOC value, battery temperature, and polarization voltage, providing a dynamic and reliable benchmark reference. Compared with the static calibration method of existing technologies, this can better adapt to the changes in the battery under different operating conditions, improving the accuracy and real-time performance of SOC estimation. In addition, through reasonable logical judgment and precise data collection, the module significantly reduces errors caused by environmental factors and equipment aging, ensuring the safety and reliability of the battery management system.

[0068] The implementation of the benchmark reference point calibration module improves the accuracy and stability of the entire energy storage system's state of charge estimation platform. This module provides a solid foundation for subsequent spatiotemporal correlation analysis, dynamic noise adjustment, and safety assessment and early warning. It ensures effective matching of real-time data with the benchmark reference, thereby optimizing the effectiveness of the UKF unscented Kalman filter. With accurate baseline SOC values ​​and temperature data, the system can more effectively identify anomalies and trigger warnings, enhancing the battery's safety management capabilities and providing a reliable basis for modifying model parameters, thereby improving the overall operating efficiency and lifespan of the energy storage system.

[0069] The spatiotemporal correlation analysis module is used to perform spatiotemporal correlation between the real-time collected current, voltage, and battery temperature and the benchmark reference points, predict the SOC value through the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate benchmark point, construct a composite weight coefficient for SOC correction, and select candidate benchmark points whose composite weight coefficients meet the conditions to form a valid reference set;

[0070] In this embodiment, the real-time collected current, voltage, and battery temperature are temporally and spatially correlated with the nearest reference point. The specific logic is as follows:

[0071] Establish a spatiotemporal correlation model to Current at the moment ,Voltage and battery temperature , retrieve the time window in the datum reference point database The reference points within the range of 0.01 and with a battery temperature difference of no more than 5°C are selected as candidate reference points. Hour, Indicates the current moment;

[0072] Substitute the current current, voltage and battery temperature into the UKF unscented Kalman filter state equation to obtain the current moment Predicted SOC value;

[0073] For each candidate benchmark point, calculate its corresponding time decay weight and spatial distance weight according to the following formula:

[0074] ;

[0075] Where, For the The time decay weight of candidate reference points, For the The calibration timestamps of candidate benchmark points, is the index of the candidate reference point, is the time decay constant, seconds, used to control the rate at which weights decay over time, is a natural constant; For the The spatial distance weight of candidate benchmark points, Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature during calibration of the candidate reference points. The temperature deviation coefficient of 1 / 25 is set based on the equivalent relationship between the battery's electrochemical characteristics. It reflects the equivalence of the effects of temperature changes and SOC changes on the spatial distance weight under typical operating conditions, ensuring that the contributions of temperature parameters and SOC parameters in the spatial distance weight calculation are reasonably matched.

[0076] Table 1: Statistics of time decay weight and spatial distance weight

[0077]

[0078] It should be noted that in the above table, the time deviation corresponds to Item, SOC deviation corresponds to Item, temperature deviation corresponds to item.

[0079] See also Figure 3-Figure 4 In this data analysis, the time-decay weight exhibits a typical exponential decay pattern, gradually decreasing from 1 (perfect match) to near 0 as the absolute value of the time offset increases. When the time offset reaches -4.75 hours, the weight has decayed to 0.0149, indicating that the system assigns very low weight to historical benchmarks older than 4 hours. This decaying property ensures that the algorithm prioritizes recent data.

[0080] From the tabular data, we can see that as the SOC deviation and temperature deviation increase, we can observe a changing trend in the spatial distance weight: Initially, when the SOC and temperature deviations are small, the spatial distance weight remains at a high level. For example, when the SOC and temperature are both zero, the spatial distance weight reaches 1, indicating that the system is in an ideal state. However, as the SOC and temperature deviations increase, especially when the SOC reaches 0.5 and above, the spatial distance weight begins to decrease significantly, and even becomes negative, indicating that the robustness of the system is seriously affected. This trend shows that when the SOC and temperature deviations reach a certain threshold, the performance of the battery management system will be significantly worse than expected, which may lead to battery overheating, shortened life, or safety hazards.

[0081] In summary, the increase in SOC and temperature deviation significantly affects the spatial distance weight, demonstrating the importance of accurately monitoring and adjusting SOC and temperature during battery management. This analysis provides theoretical support for further battery management strategies and optimization, helping to improve the overall performance and safety of battery systems.

[0082] Based on the above analysis, the calculation of spatiotemporal weights provides the present invention with an assessment of the relative importance of each candidate reference point. The combination of time-decay weights, SOC deviation, and temperature deviation provides a scientific basis for optimizing battery state estimation. This analysis provides a better understanding of the impact of different candidate reference points on the current state, thus supporting subsequent decision-making. This multi-dimensional assessment method ensures the accuracy and reliability of the battery management system in complex environments.

[0083] The composite weight coefficient for SOC correction is constructed based on the time decay weight and the spatial distance weight, and the formula is as follows:

[0084] ;

[0085] Where, Indicates the The composite weight coefficient of candidate benchmark points, and Respectively represent The time attenuation weight and spatial distance weight of candidate benchmark points, is the total number of candidate benchmark points;

[0086] Calculate the composite weight coefficient for all candidate benchmark points and select the composite weight coefficient The candidate reference points constitute a valid reference set.

[0087] The spatiotemporal correlation analysis module accurately predicts changes in SOC values ​​by effectively matching the real-time current, voltage, and battery temperature with the nearest benchmark reference point in spatiotemporal order. Compared to existing technologies, this module dynamically accounts for time decay and spatial distance factors, ensuring that the selected candidate benchmark points are closer to the current state in time and temperature, significantly improving the accuracy and timeliness of SOC estimation. This UKF-based spatiotemporal correlation method can better adapt to rapid changes in battery state, reduce estimation errors caused by environmental factors, and comprehensively enhance the reliability of the battery management system.

[0088] By adopting the spatiotemporal correlation analysis module, the overall energy storage system state of charge estimation platform is able to achieve a more efficient SOC estimation process. This module provides an accurate data foundation for subsequent dynamic noise adjustment and safety assessment warning, enabling the system to promptly correct and optimize state estimation in real-time changes. The construction of an effective reference set and the generation of composite weight coefficients significantly improve the effectiveness of the UKF unscented Kalman filter, thereby enhancing the intelligence and automation level of battery management. At the same time, the module's capabilities in anomaly detection and warning provide important guarantees for the safe operation of the battery system and promote the overall performance improvement of the energy storage system.

[0089] A dynamic noise adjustment module is used to dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the effective reference set, and update the state vector through the observation equation. The updated state vector includes the SOC estimate, battery temperature and polarization voltage;

[0090] In this embodiment, the process noise covariance matrix and the observation noise covariance matrix of the UKF are dynamically adjusted based on the composite weight coefficient in the effective reference set, and the SOC estimation value is updated through the observation equation. The specific logic is as follows:

[0091] Based on the composite weight coefficients of each candidate reference point in the valid reference set, the adjustment factor of the process noise covariance matrix is ​​calculated according to the formula:

[0092] ;

[0093] Where, is the adjustment factor of the process noise covariance matrix, Indicates the first The composite weight coefficient of candidate benchmark points, represents the maximum value of the composite weight coefficient in the valid reference set, represents the number of candidate reference points in the valid reference set, is the index of the candidate reference point in the valid reference set, Refers to the standard deviation of the composite weight coefficients in the effective reference set.

[0094] Dependent variable in the formula is the adjustment factor of the process noise covariance matrix, which is used to dynamically adjust the process noise covariance matrix in the UKF unscented Kalman filter algorithm to adapt to different working conditions and environmental conditions. The technical effect is reflected in its ability to dynamically optimize the process noise covariance matrix based on the composite weight coefficient of the candidate reference point in the effective reference set, reduce the state estimation error caused by noise, and improve the accuracy and robustness of SOC estimation. This adjustment mechanism enables the system to respond to complex working conditions more flexibly and improves the state estimation stability of the energy storage system. The two independent variables in the formula are and , which are derived from the composite weight coefficient of the candidate reference point, reflecting the matching degree between the reference point and the current state. The weight value of the reference point in the valid reference set that is closest to the current state indicates whether there is a highly matching reference point in the reference set. The higher the value, the lower the noise adjustment requirement and the smaller the impact of process noise. Reflects the degree of dispersion of the reference point weights in the valid reference set. The higher the degree of dispersion, the greater the difference in weights within the reference set. The state estimation may be affected by unstable reference points, and a higher adjustment factor is required to compensate for the noise. When the value is large, it means that there are highly matched reference points in the effective reference set, and the system has less need to adjust to process noise. will decrease, reducing the adjustment range of the process noise covariance matrix. When the value is large, it means that the weight distribution of the reference points in the reference set is more different, and the system needs to adjust the process noise more. will increase to enhance the compensation effect for process noise.

[0095] Adjustment factors based on the process noise covariance matrix Update process noise covariance matrix :

[0096] ;

[0097] Where, is the initial covariance matrix, is the preset lower limit matrix;

[0098] The adjustment coefficient of the observation noise covariance matrix is ​​calculated based on the distance between the current time and the time at which the candidate reference point in the valid reference set is located. The formula is as follows:

[0099] ;

[0100] Where, is the adjustment coefficient of the observation noise covariance matrix, Indicates the average of the Euclidean distances of all candidate benchmark points; Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature when calibrating the candidate reference points;

[0101] The dependent variable in the formula is the adjustment coefficient of the observation noise covariance matrix , used to dynamically adjust the observation noise covariance matrix in the UKF unscented Kalman filter algorithm to adapt to different environmental and state conditions. The technical benefit is its ability to dynamically optimize the observation noise covariance matrix based on the average distance from the current moment to the candidate reference point, thereby improving the accuracy and robustness of the SOC estimation. By adjusting the noise covariance, the system can better reflect the uncertainty of the current state, ensuring the stability and reliability of the SOC estimation under changing operating conditions.

[0102] The independent variables in the formula are mainly , which is the average Euclidean distance of the candidate reference points. This independent variable reflects the closeness between the current state and each reference point in the reference set. When the distance between the current state and the reference point is small, it means that the current state matches the reference set well, and the system has a low need to adjust the observation noise. will decrease accordingly; conversely, when the distance is large, the matching degree is low, and the system will require a higher observation noise covariance adjustment to enhance the compensation for uncertainty. Therefore, the independent variable directly affects the size of the dependent variable by reflecting the matching degree of the current state, thereby achieving dynamic adjustment of observation noise.

[0103] When the average distance of the candidate reference points increases, it indicates that the matching degree between the current state and the reference set decreases, which means that the system needs a higher observation noise covariance adjustment to cope with possible state estimation errors. The positive correlation between the observation noise and the distance between the current state and the reference point will increase, thereby increasing the value of the observation noise covariance matrix. This positive correlation design ensures that the system can dynamically adjust the observation noise according to the distance between the current state and the reference point, thereby maintaining the accuracy and stability of the SOC estimation under different operating conditions.

[0104] Adjustment coefficients based on the observation noise covariance matrix Update the observation noise covariance matrix :

[0105] ;

[0106] Where, is the initial observation noise covariance.

[0107] The UKF observation equation is used to update the SOC estimate, based on the following formula:

[0108] ;

[0109] Where, For the current moment The observed voltage value, Indicates based on The predicted open circuit voltage, and The function uses the piecewise function corresponding to the highest weighted reference point in the valid reference set. Indicates the current time Predicted SOC value; It is the current moment The current, Indicates the ohmic internal resistance value corresponding to the highest weight reference point, and Represents the current time The first polarization voltage and the second polarization voltage, For the current moment The observation noise follows a normal distribution , is the covariance matrix of the observation noise;

[0110] Calculate the Kalman gain by UKF unscented transformation , and update the state vector, the mathematical expression is as follows: ;

[0111] Where, Indicates at time The updated state vector, including the SOC estimate, battery temperature, and polarization voltage, For the moment The predicted state vector, that is, the state prediction of the previous step, For the moment The calculated Kalman gain is, Indicates at time The actual measured voltage value;

[0112] According to the Kalman gain Update the covariance matrix of the state estimate :

[0113] ;

[0114] Where, Indicates at time The updated state estimation covariance matrix, Indicates at time The predicted state covariance matrix, is the prediction residual covariance matrix, represents the Kalman gain The transpose of .

[0115] Reflects the uncertainty of the state estimate. In Kalman filtering, updating the covariance matrix is ​​a key step, ensuring that the system can adjust its confidence in the state based on new observations. By updating the covariance matrix, the credibility of the current state can be more accurately assessed. represents the uncertainty assessment of the state when there is no new observation information. This matrix is ​​based on the state estimate at the previous moment and the dynamic model for prediction, providing a preliminary covariance estimate. Determines the degree to which new observations correct the uncertainty of the state estimate. The calculation of the Kalman gain takes into account the effects of the predicted state estimate covariance and observation noise to ensure that the prior information and measurement information are optimally combined during the update process. represents the uncertainty introduced by the observation noise during the update process. By combining the Kalman gain with the residual covariance matrix, the updated covariance matrix It can effectively reflect the uncertainty of the latest observation results on the state estimation. In the update formula, the product term of Kalman gain Subtracting from the forecast covariance matrix means that the uncertainty in the state estimate generally decreases as new observations are introduced. This process makes the state estimate more accurate and improves the system's ability to predict future states.

[0116] The dynamic noise adjustment module dynamically adjusts the UKF's process noise covariance matrix and observation noise covariance matrix based on actual conditions by analyzing the composite weight coefficients in the effective reference set in real time. Compared with existing technologies, this module improves the system's adaptability to different operating conditions and environments, thereby reducing the impact of noise on SOC estimation. This flexible noise processing mechanism enables the system to better cope with fluctuations caused by battery status and environmental changes, ensuring the accuracy and reliability of SOC estimation, ultimately improving the operating efficiency and safety of the energy storage system.

[0117] The implementation of the dynamic noise adjustment module significantly improves the performance of the overall energy storage system state of charge estimation platform. By adjusting the noise covariance matrix in real time, the system effectively optimizes the UKF state estimation process, making the SOC estimate more accurate and stable. This not only enhances the ability to detect abnormal conditions but also provides a more solid foundation for subsequent safety assessments and early warnings, reducing potential risks and thus promoting the overall safety and efficiency of the energy storage system. Overall, the introduction of this module ensures the system's continued efficient operation under complex operating conditions and enhances its intelligence level.

[0118] The safety assessment and warning module is used to establish a dynamic safety assessment domain centered on the SOC estimate on the SOC-battery temperature plane. When the SOC estimate exceeds the confidence interval constructed by the benchmark reference point three times in a row, an alert is triggered and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are then applied to the UKF state estimation in the next working cycle.

[0119] In this embodiment, the specific logic used by the security assessment and early warning module is as follows:

[0120] A dynamic safety assessment domain with the SOC estimation as the core is established on the SOC-battery temperature plane. When the SOC estimation value is detected to exceed the confidence interval constructed by the benchmark reference point in three consecutive samplings, it is marked as abnormal and triggers an early warning. The abnormal deviation vector is automatically recorded. The abnormal deviation vector includes the SOC deviation, battery temperature deviation and polarization voltage change for subsequent model parameter correction. The corrected model parameters will be applied to the UKF state estimation of the next working cycle; the safety assessment domain is visualized with a heat map, abnormal points are marked with red pulses and deviation data are displayed; the SOC-battery temperature plane is a two-dimensional coordinate system, The axis is the estimated SOC value, Axis is battery temperature;

[0121] Among them, the SOC deviation refers to the difference between the current SOC estimation value and the SOC mean value in the benchmark reference point, the battery temperature deviation refers to the difference between the currently measured battery temperature and the battery temperature mean value in the benchmark reference point, and the polarization voltage change refers to the difference between the currently measured polarization voltage and the polarization voltage mean value in the benchmark reference point.

[0122] The safety assessment and early warning module establishes a dynamic safety assessment domain centered around the estimated SOC value, enabling real-time monitoring of the battery's operating status and automatically triggering early warnings when the SOC value continuously exceeds the confidence interval. This mechanism offers significant advantages over traditional static monitoring methods, as it can promptly identify potential safety hazards and avoid system failures or safety incidents caused by delayed responses. Furthermore, the module automatically records abnormal deviation vectors, providing data support for subsequent model parameter corrections, thereby continuously optimizing the performance of the battery management system and improving overall safety and reliability.

[0123] The introduction of the safety assessment and early warning module significantly enhances the intelligence and safety of the entire energy storage system's state of charge estimation platform. The module's real-time monitoring and early warning capabilities not only improve rapid response to abnormal situations but also provide a basis for dynamic noise adjustment, ensuring the accuracy of SOC estimation. When an anomaly occurs, the module automatically adjusts model parameters to optimize state estimation, thereby improving system stability and reliability. This closed-loop feedback mechanism makes the battery management system more adaptable in dynamic environments, ensuring the safe and efficient operation of the energy storage system and promoting overall technological advancement.

[0124] See also Figure 2 The energy storage system state of charge estimation method based on the improved unscented Kalman filter includes the following steps:

[0125] Step 1: During the operation of the energy storage system, the static operating condition is identified and a baseline calibration is performed. The baseline SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the baseline SOC value, polarization voltage, and battery temperature is generated as the baseline reference point for state estimation.

[0126] Step 2: Temporally and spatially correlate the real-time current, voltage, and battery temperature with the reference points, predict the SOC value using the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate reference point, construct a composite weight coefficient for SOC correction, and select candidate reference points whose composite weight coefficients meet the conditions to form a valid reference set;

[0127] Step 3: Dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the valid reference set, and update the state vector through the observation equation. The updated state vector includes the estimated SOC value, battery temperature and internal resistance;

[0128] Step 4: A dynamic safety assessment domain with the SOC estimate as the core is established on the SOC-temperature plane. When the SOC estimate exceeds the confidence interval constructed by the benchmark reference point three times in a row, an early warning is triggered, and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are applied to the UKF state estimation of the next working cycle.

[0129] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0130] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0131] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0132] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A state of charge estimation platform for energy storage systems based on an improved unscented Kalman filter, characterized in that: Specifically include: The benchmark reference point calibration module is used to identify the static operating condition and perform benchmark calibration during the operation of the energy storage system. The benchmark SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the benchmark SOC value, battery temperature, and polarization voltage is generated as the benchmark reference point for state estimation. The spatiotemporal correlation analysis module is used to perform spatiotemporal correlation between the real-time collected current, voltage, and battery temperature and the benchmark reference points, predict the SOC value through the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate benchmark point, construct a composite weight coefficient for SOC correction, and select candidate benchmark points whose composite weight coefficients meet the conditions to form a valid reference set; A dynamic noise adjustment module is used to dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the effective reference set, and update the state vector through the observation equation. The updated state vector includes the SOC estimate, battery temperature and polarization voltage; The safety assessment and early warning module is used to establish a dynamic safety assessment domain with the SOC estimation value as the core on the SOC-battery temperature plane. When the SOC estimation value exceeds the confidence interval constructed by the benchmark reference point three times in a row, an early warning is triggered and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are applied to the UKF state estimation of the next working cycle.

2. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 1, characterized in that: The specific logic for executing the reference point calibration module is: when the absolute value of the current is detected to be less than And the voltage fluctuation rate is less than When it is determined to be in a static condition, the average terminal voltage 30 seconds before the end of the static condition is collected. The closest SOC value is matched as the reference SOC value in the OCV-SOC curve database by binary search method. At the same time, the current battery temperature is measured and the polarization voltage at this time is identified by the least square method. and Finally, a three-dimensional calibration vector containing the reference SOC value, polarization voltage and battery temperature is generated as the reference point for state estimation, and the three-dimensional calibration vector is stored in the reference point database together with its calibration timestamp. The expression is as follows: ； Where, is the benchmark SOC value, and is the polarization voltage, represents the first polarization voltage generated by the R1-C1 parallel circuit, represents the second polarization voltage generated by the R2-C2 parallel circuit, R1 and R2 represent the resistance in the first polarization voltage generating circuit and the resistance in the second polarization voltage generating circuit, respectively, C1 and C2 represent the capacitance in the first polarization voltage generating circuit and the capacitance in the second polarization voltage generating circuit, respectively, is the battery temperature.

3. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 1, characterized in that: The real-time collected current, voltage, and battery temperature are temporally and spatially correlated with the benchmark reference point. The specific logic is as follows: Establish a spatiotemporal correlation model for the current moment Current ,Voltage and battery temperature , retrieve the time window in the datum reference point database The reference points within the range of 0.01 and with a battery temperature difference of no more than 5°C are selected as candidate reference points. Hour, Indicates the current moment; Substitute the current current, voltage and battery temperature into the UKF unscented Kalman filter state equation to obtain the current moment Predicted SOC value; For each candidate benchmark point, calculate its corresponding time decay weight and spatial distance weight according to the following formula: ； Where, For the The time decay weight of candidate reference points, For the The calibration timestamps of candidate benchmark points, is the index of the candidate reference point, is the time decay constant, which is used to control the decay rate of the weight over time. is a natural constant; For the The spatial distance weight of candidate benchmark points, Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature when calibrating the candidate reference points.

4. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 3, characterized in that: The composite weight coefficient for SOC correction is constructed based on the time decay weight and the spatial distance weight, and the formula is as follows: ； Where, Indicates the The composite weight coefficient of candidate benchmark points, and Respectively represent The time attenuation weight and spatial distance weight of candidate benchmark points, is the total number of candidate benchmark points; Calculate the composite weight coefficient for all candidate benchmark points and select the composite weight coefficient The candidate reference points constitute a valid reference set.

5. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 1, characterized in that: The process noise covariance matrix and observation noise covariance matrix of UKF are dynamically adjusted based on the composite weight coefficient in the effective reference set, and the SOC estimation value is updated through the observation equation. The specific logic is as follows: Based on the composite weight coefficients of each candidate reference point in the valid reference set, the adjustment factor of the process noise covariance matrix is ​​calculated according to the formula: ； Where, is the adjustment factor of the process noise covariance matrix, Indicates the first The composite weight coefficient of candidate benchmark points, is the index of the candidate reference point in the valid reference set, represents the number of candidate reference points in the valid reference set, Refers to the standard deviation of the composite weight coefficients in the effective reference set; Adjustment factors based on the process noise covariance matrix Update process noise covariance matrix : ； Where, is the initial covariance matrix, is the preset lower limit matrix; The adjustment coefficient of the observation noise covariance matrix is ​​calculated based on the distance between the current time and the time at which the candidate reference point in the valid reference set is located. The formula is as follows: ； Where, is the adjustment coefficient of the observation noise covariance matrix, Indicates the average of the Euclidean distances of all candidate benchmark points; Indicates the current time Predicted SOC value, Indicates the The benchmark SOC value calibrated by the candidate benchmark points, Indicates the current time Measured battery temperature, Indicates the The battery temperature when calibrating the candidate reference points; Adjustment coefficients based on the observation noise covariance matrix Update the observation noise covariance matrix : ； Where, is the initial observation noise covariance.

6. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 1, characterized in that: The UKF observation equation is used to update the SOC estimate, based on the following formula: ； Where, For the current moment The observed voltage value, Indicates based on The predicted open circuit voltage, and The function uses the piecewise function corresponding to the highest weighted reference point in the valid reference set. Indicates the current time Predicted SOC value; It is the current moment The current, Indicates the ohmic internal resistance value corresponding to the highest weight reference point, and Represents the current time The first polarization voltage and the second polarization voltage, For the current moment The observation noise follows a normal distribution , is the covariance matrix of the observation noise; Calculate the Kalman gain by UKF unscented transformation , and update the state vector, the mathematical expression is as follows: ; Where, Indicates at time The updated state vector, including the SOC estimate, battery temperature, and polarization voltage, For the moment The predicted state vector, that is, the state prediction of the previous step, For the moment The calculated Kalman gain is, Indicates at time The actual measured voltage value; According to the Kalman gain Update the covariance matrix of the state estimate : ； Where, Indicates at time The updated state estimation covariance matrix, Indicates at time The predicted state covariance matrix, is the prediction residual covariance matrix, represents the Kalman gain The transpose of .

7. The energy storage system state of charge estimation platform based on improved unscented Kalman filtering according to claim 1, characterized in that: The specific logic for executing the security assessment and early warning module is as follows: A dynamic safety assessment domain centered around the SOC estimate is established on the SOC-battery temperature plane. When the SOC estimate is detected to exceed the confidence interval established by the benchmark reference point in three consecutive samplings, it is marked as abnormal and triggers an alert. The abnormal deviation vector, including the SOC deviation, battery temperature deviation, and polarization voltage change, is automatically recorded for subsequent model parameter correction. The corrected model parameters will be applied to the UKF state estimation in the next working cycle. The safety assessment domain is visualized with a heat map, abnormal points are marked with red pulses and deviation data are displayed; the SOC-battery temperature plane is a two-dimensional coordinate system. The axis is the estimated SOC value, Axis is battery temperature; Among them, the SOC deviation refers to the difference between the current SOC estimation value and the SOC mean value in the benchmark reference point, the battery temperature deviation refers to the difference between the currently measured battery temperature and the battery temperature mean value in the benchmark reference point, and the polarization voltage change refers to the difference between the currently measured polarization voltage and the polarization voltage mean value in the benchmark reference point.

8. A method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter is characterized by: The method for estimating the state of charge of an energy storage system based on an improved unscented Kalman filter is obtained by using the energy storage system state of charge estimation platform based on an improved unscented Kalman filter according to any one of claims 1 to 7, comprising: Step 1: During the operation of the energy storage system, the static operating condition is identified and a baseline calibration is performed. The baseline SOC value is determined by matching the static terminal voltage with the OCV-SOC curve. A three-dimensional calibration vector containing the baseline SOC value, polarization voltage, and battery temperature is generated as the baseline reference point for state estimation. Step 2: Temporally and spatially correlate the real-time current, voltage, and battery temperature with the reference points, predict the SOC value using the UKF unscented Kalman filter state equation, calculate the time decay weight and spatial distance weight of each candidate reference point, construct a composite weight coefficient for SOC correction, and select candidate reference points whose composite weight coefficients meet the conditions to form a valid reference set; Step 3: Dynamically adjust the process noise covariance matrix and observation noise covariance matrix of the UKF based on the composite weight coefficient in the valid reference set, and update the state vector through the observation equation. The updated state vector includes the estimated SOC value, battery temperature and internal resistance; Step 4: A dynamic safety assessment domain with the SOC estimate as the core is established on the SOC-temperature plane. When the SOC estimate exceeds the confidence interval constructed by the benchmark reference point three times in a row, an early warning is triggered, and the abnormal deviation vector is automatically recorded for model parameter correction. The corrected model parameters are applied to the UKF state estimation of the next working cycle.

Images