A method, system, device, and medium for state estimation of an energy storage battery

Abstract

The application discloses a kind of energy storage battery state estimation method, system, equipment and medium, and it is related to power system technical field.Capacity data of energy storage battery are used to carry out capacity attenuation experiment, and generate battery health state change data set.Using battery health state change data set carries out trajectory fitting, determines initial life experience model.Based on parameter identification method and the time data of discharge to cut-off voltage corresponding to battery health state change data set, the initial life experience model is updated to generate the target life experience model.Using the linear constraint corresponding to the target life experience model, the initial economic dispatching model is updated to generate the target economic dispatching model.The battery health state change data set is input into the target life experience model and the target economic dispatching model respectively, and state estimation and economic dispatching data are generated.The aging characteristics of charge and discharge cycle process are fully explored, and higher precision energy storage battery health state estimation is realized.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to methods, systems, devices and media for estimating the state of energy storage batteries. Background Technology

[0002] With the large-scale integration of new energy sources into the grid, the issue of grid safety is becoming increasingly prominent. Energy storage systems, as a key emerging device in the power system, can play an important role in achieving peak shaving and valley filling, and ensuring the economical and stable operation of the power grid. The health status of energy storage, as a crucial operational performance parameter, is vital. If this health status is not accurately assessed after being put into actual power system operation, it can lead to distortions in the analysis of energy storage power balance or safety and stability. Therefore, employing appropriate models to accurately characterize the health status of energy storage systems is of paramount importance.

[0003] However, in practical applications of energy storage batteries in power systems, the performance of individual cells and their operating environments inevitably vary, making direct measurement impossible. Therefore, the state estimation problem for energy storage batteries is essentially an implicit state estimation problem for time-varying nonlinear systems. Currently, there are three main solution methods: empirical estimation, filtering estimation, and time-series estimation. Although filtering and time-series estimation methods offer high accuracy, their high computational complexity makes them unsuitable for practical engineering needs. Therefore, empirical estimation methods are currently the primary approach for energy storage battery state estimation.

[0004] Empirical estimation methods assume that the capacity degradation of power batteries follows a certain inherent mathematical relationship. By using historical data to determine undetermined coefficients, a mathematical model describing the aging law of power batteries can be established, thus solving the problem of battery health state estimation. However, existing battery health state estimation methods suffer from low accuracy in scenarios such as sudden changes in maximum usable battery capacity caused by battery relaxation, due to insufficient consideration of the impact of battery polarization effects. Summary of the Invention

[0005] This invention provides a method, system, device, and medium for estimating the state of energy storage batteries. It solves the technical problem that existing battery state estimation methods have low accuracy in estimating the health state of energy storage batteries in scenarios such as sudden changes in maximum usable battery capacity caused by battery relaxation, due to insufficient consideration of the influence of battery polarization effects.

[0006] This invention provides a method for estimating the state of a storage battery, comprising:

[0007] Acquire the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a dataset of battery health status changes.

[0008] The battery health state change dataset was used to perform trajectory fitting to determine the initial lifetime empirical model;

[0009] The initial lifetime empirical model is updated based on the parameter identification method and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate the target lifetime empirical model.

[0010] The initial economic scheduling model is updated using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model;

[0011] The battery health status change dataset is input into the target lifetime empirical model and the target economic scheduling model, respectively, to generate the state estimation and economic scheduling data corresponding to the energy storage battery.

[0012] Optionally, the step of using the battery health state change dataset to perform trajectory fitting and determine the initial lifetime empirical model includes:

[0013] The battery health status change dataset is used to perform trajectory fitting using a single exponential model to generate the first fitted data.

[0014] The battery health status change dataset is used to perform trajectory fitting using a dual exponential model to generate a second fitted data;

[0015] A third fitted data is generated by using the battery health state change dataset to perform trajectory fitting through a linear model;

[0016] The fourth fitting data is generated by using the battery health state change dataset to perform trajectory fitting through a multinomial model.

[0017] The fifth fitted data is generated by using the battery health state change dataset through the Fairhast algorithm model to perform trajectory fitting.

[0018] Select data from the first, second, third, fourth, and fifth fitted data that are greater than a preset effect threshold to generate an effect dataset;

[0019] The model corresponding to the maximum value in the effect dataset is used as the initial lifetime empirical model.

[0020] Optionally, the step of updating the initial lifetime empirical model based on the parameter identification method and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate the target lifetime empirical model includes:

[0021] The parameter identification method is used to identify the parameters of the initial lifetime empirical model and generate undetermined coefficients.

[0022] The initial lifetime empirical model is updated using the battery health state change dataset and the undetermined coefficients to generate an intermediate lifetime empirical model.

[0023] The discharge-to-cutoff voltage time data corresponding to the battery health state change dataset are preprocessed to generate intermediate feature data.

[0024] The intermediate feature data is denoised using a neural network method to generate the target feature data.

[0025] A smooth approximation fitting function is used to characterize the relationship curve between the target feature data and the number of cycles corresponding to the battery health state change dataset, thereby generating relationship curve data;

[0026] The intermediate lifetime empirical model is updated using the relationship curve data to generate the target lifetime empirical model.

[0027] Optionally, the step of updating the initial lifetime empirical model using the battery health state change dataset and the undetermined coefficients to generate an intermediate lifetime empirical model includes:

[0028] The initial weights of the random factors are determined by using the probabilities of each random factor in the battery health status change dataset.

[0029] The initial model parameter set is constructed by substituting the change data corresponding to the initial weights and the random factors into the corresponding weighted average function;

[0030] The minimum initial model parameter corresponding to each random factor in the initial model parameter set is used as the target model parameter to generate an intermediate model parameter set.

[0031] The intermediate model parameter set is used to verify the estimation effect, and verification data is generated.

[0032] When the verification data shows an error exceeding a threshold, a forgetting factor is used to dynamically adjust the initial weights of the random factors to generate target weights.

[0033] The target weight is used as the initial weight, and the process jumps to the step of substituting the change data corresponding to the initial weight and the random factor into the corresponding weighted average function to construct the initial model parameter set.

[0034] When the verification data shows that all errors are within the allowable range, the intermediate model parameter set at the current moment is used as the target model parameter set.

[0035] The initial lifetime empirical model is updated using the target model parameter set and the undetermined coefficients to generate an intermediate lifetime empirical model.

[0036] Optionally, the step of preprocessing the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate intermediate feature data includes:

[0037] Extract feature values ​​from the discharge-to-cutoff-voltage time data corresponding to the battery health state change dataset to generate initial feature data;

[0038] The initial feature data is subjected to repeated observation processing and default value detection and removal to generate the first preprocessed data;

[0039] The first preprocessed data is subjected to outlier detection and removal to generate the second preprocessed data;

[0040] The second preprocessed data is smoothed to generate the third preprocessed data;

[0041] The third preprocessed data is feature-encoded to generate intermediate feature data.

[0042] Optionally, the step of updating the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model includes:

[0043] The relationship between the battery health state and the number of cycles corresponding to the target life empirical model is approximated by piecewise linearization to generate linear constraints.

[0044] The initial economic scheduling model is updated using the linear constraints to generate the target economic scheduling model.

[0045] The present invention also provides a state estimation system for energy storage batteries, comprising:

[0046] The battery health status change dataset generation module is used to acquire the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a battery health status change dataset.

[0047] The initial lifespan empirical model determination module is used to perform trajectory fitting using the battery health state change dataset to determine the initial lifespan empirical model.

[0048] The target lifetime empirical model generation module is used to update the initial lifetime empirical model based on the parameter identification method and the discharge to cutoff voltage time data corresponding to the battery health state change dataset, and generate the target lifetime empirical model.

[0049] The target economic scheduling model generation module is used to update the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model, and generate the target economic scheduling model.

[0050] The state estimation and economic dispatch data generation module is used to input the battery health state change dataset into the target lifetime empirical model and the target economic dispatch model respectively, and generate the state estimation and economic dispatch data corresponding to the energy storage battery.

[0051] Optionally, the initial lifetime empirical model determination module includes:

[0052] The first fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a single exponential model to generate the first fitting data.

[0053] The second fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a double exponential model to generate second fitting data.

[0054] The third fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a linear model to generate third fitting data.

[0055] The fourth fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a multinomial model to generate the fourth fitting data.

[0056] The fifth fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through the Fairhast algorithm model to generate the fifth fitting data.

[0057] The effect dataset generation module is used to select data from the first fitted data, the second fitted data, the third fitted data, the fourth fitted data, and the fifth fitted data that are greater than a preset effect threshold, and generate an effect dataset;

[0058] The initial lifetime empirical model determination submodule is used to take the model corresponding to the maximum value in the effect dataset as the initial lifetime empirical model.

[0059] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs steps to implement any of the above-described energy storage battery state estimation methods.

[0060] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements any of the above-described energy storage battery state estimation methods.

[0061] As can be seen from the above technical solutions, the present invention has the following advantages:

[0062] This invention acquires capacity data of energy storage batteries, conducts capacity decay experiments using this data, and generates a battery health state change dataset. Trajectory fitting is performed using this dataset to determine an initial lifetime empirical model. Based on parameter identification and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset, the initial lifetime empirical model is updated to generate a target lifetime empirical model. Linear constraints corresponding to the target lifetime empirical model are used to update the initial economic dispatch model to generate a target economic dispatch model. The battery health state change dataset is then input into both the target lifetime empirical model and the target economic dispatch model to generate state estimation and economic dispatch data for the energy storage battery. This invention solves the technical problem of low accuracy in existing battery state estimation methods in scenarios such as sudden changes in maximum usable battery capacity due to battery relaxation, which fail to fully consider the impact of battery polarization effects. Based on discharge-to-cutoff voltage time data, a dynamic decay dual index of constant voltage discharge time and cycle number is formed, fully exploring the aging characteristics of the charge-discharge cycle process, achieving higher accuracy in energy storage system health state estimation, and improving the observability of the energy storage system's health state. The target lifetime empirical model and the target economic dispatch model can not only support the planning and optimized operation of energy storage, but also further study the impact of the characteristic parameters of energy storage lifetime decay on the economic dispatch of the power system. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a flowchart of the steps of a battery state estimation method provided in Embodiment 1 of the present invention;

[0065] Figure 2 This is a flowchart of the steps of a battery state estimation method provided in Embodiment 2 of the present invention;

[0066] Figure 3 This is a graph showing the change of voltage and current over time during the charging and discharging process, corresponding to the extracted discharge-to-cutoff voltage time data features provided in Embodiment 2 of the present invention.

[0067] Figure 4 The diagram shows the original discharge-to-cutoff voltage time data and the discharge-to-cutoff voltage time data after removing abnormal data and reducing noise, as provided in Embodiment 2 of the present invention.

[0068] Figure 5This is a graph showing the voltage change over time at different cycle numbers, provided in Embodiment 2 of the present invention.

[0069] Figure 6 This is a comparison between the battery health status of the energy storage system estimated using the target lifetime empirical model and the actual battery health status, as provided in Embodiment 2 of the present invention.

[0070] Figure 7 This is a schematic diagram of the correlation analysis method between economic dispatch results and energy storage lifetime provided in Embodiment 2 of the present invention;

[0071] Figure 8 This is a structural block diagram of a battery state estimation system provided in Embodiment 3 of the present invention. Detailed Implementation

[0072] This invention provides a method, system, device, and medium for estimating the state of energy storage batteries, which addresses the technical problem that existing battery state estimation methods suffer from low accuracy in estimating the health state of energy storage batteries in scenarios such as sudden changes in maximum usable battery capacity caused by battery relaxation, due to insufficient consideration of the influence of battery polarization effects.

[0073] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0074] Please see Figure 1 , Figure 1 This is a flowchart illustrating the steps of a battery state estimation method provided in Embodiment 1 of the present invention.

[0075] Example 1 of this invention provides a method for estimating the state of an energy storage battery, comprising:

[0076] Step 101: Obtain the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a dataset of battery health status changes.

[0077] In this embodiment of the invention, the battery's state of health (SOH) is obtained as the number of cycles increases through an experimental method to study the degradation characteristics of the energy storage battery. Under a certain charge-discharge regime, data is extracted from the entire process of the energy storage battery's maximum usable capacity decaying from its maximum capacity to a predetermined failure threshold, using experimental methods. This yields a dataset showing the change in the battery's SOH with the increasing number of cycles.

[0078] Step 102: Use the battery health state change dataset to perform trajectory fitting and determine the initial lifetime empirical model.

[0079] In this embodiment of the invention, trajectory fitting is performed using a single exponential model on the battery health state change dataset to generate first fitted data. Trajectory fitting is then performed using a double exponential model on the same dataset to generate second fitted data. Trajectory fitting is then performed using a linear model on the same dataset to generate third fitted data. Trajectory fitting is then performed using a multinomial model on the same dataset to generate fourth fitted data. Trajectory fitting is then performed using the Fairhast algorithm model on the same dataset to generate fifth fitted data. Data exceeding a preset effect threshold from the first, second, third, fourth, and fifth fitted data are selected to generate the effect dataset.

[0080] Step 103: Update the initial lifetime empirical model based on the parameter identification method and the discharge to cutoff voltage time data corresponding to the battery health state change dataset, and generate the target lifetime empirical model.

[0081] In this embodiment of the invention, a parameter identification method is used to identify parameters of the initial lifetime empirical model, generating undetermined coefficients. The initial lifetime empirical model is updated using a battery health state change dataset and the undetermined coefficients to generate an intermediate lifetime empirical model. The discharge-to-cutoff voltage time data corresponding to the battery health state change dataset is preprocessed to generate intermediate feature data. A neural network method is used to denoise the intermediate feature data, generating target feature data. A smooth approximation fitting function is used to characterize the relationship curve between the target feature data and the number of cycles corresponding to the battery health state change dataset, generating relationship curve data. The intermediate lifetime empirical model is updated using the relationship curve data to generate the target lifetime empirical model.

[0082] Step 104: Update the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model.

[0083] In this embodiment of the invention, the relationship between battery health state and cycle count corresponding to the target lifetime empirical model is approximated as piecewise linearized to generate linear constraints. The initial economic scheduling model is then updated using these linear constraints to generate the target economic scheduling model.

[0084] Step 105: Input the battery health status change dataset into the target lifetime empirical model and the target economic dispatch model respectively to generate the corresponding state estimation and economic dispatch data of the energy storage battery.

[0085] In this embodiment of the invention, the state of the energy storage battery is estimated using a target lifetime empirical model on a dataset of battery health state changes, yielding state estimation data. Real-time measured energy storage battery data can be substituted into the target lifetime empirical model for real-time estimation of the energy storage battery state. Economic scheduling analysis is then performed on the battery health state change dataset using the target lifetime empirical model to obtain economic scheduling data, i.e., mitigation strategies, enabling quantitative analysis of the impact of energy storage lifespan degradation characteristics on economic scheduling. Using the state estimation data and economic scheduling data, state estimation and economic scheduling data corresponding to the energy storage battery are constructed.

[0086] In this embodiment of the invention, capacity data of the energy storage battery is acquired, and a capacity decay experiment is conducted using the capacity data to generate a battery health state change dataset. Trajectory fitting is performed using the battery health state change dataset to determine an initial lifetime empirical model. Based on parameter identification and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset, the initial lifetime empirical model is updated to generate a target lifetime empirical model. The initial economic scheduling model is updated using the linear constraints corresponding to the target lifetime empirical model to generate a target economic scheduling model. The battery health state change dataset is input into the target lifetime empirical model and the target economic scheduling model, respectively, to generate state estimation and economic scheduling data for the energy storage battery. This solves the technical problem of low accuracy in existing battery state estimation methods in scenarios such as sudden changes in maximum usable battery capacity due to battery relaxation, which fail to fully consider the influence of battery polarization effects. Based on the discharge-to-cutoff voltage time data, a dynamic decay dual index of constant voltage discharge time and cycle number is formed, fully exploring the aging characteristics of the charge-discharge cycle process, achieving higher accuracy in energy storage system health state estimation, and improving the observability of the energy storage system health state. The target lifetime empirical model and the target economic dispatch model can not only support the planning and optimized operation of energy storage, but also further study the impact of the characteristic parameters of energy storage lifetime decay on the economic dispatch of the power system.

[0087] Please see Figure 2 , Figure 2 This is a flowchart illustrating the steps of a battery state estimation method provided in Embodiment 2 of the present invention.

[0088] Another energy storage battery state estimation method provided in Example 2 of this invention includes:

[0089] Step 201: Obtain the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a dataset of battery health status changes.

[0090] In this embodiment of the invention, the specific implementation process of step 201 is similar to that of step 101, and will not be repeated here.

[0091] Step 202: Use the battery health state change dataset to perform trajectory fitting and determine the initial lifetime empirical model.

[0092] Furthermore, step 202 may include the following sub-steps S11-S17:

[0093] S11. The trajectory is fitted using a single exponential model based on the battery health status change dataset to generate the first fitted data.

[0094] S12. The trajectory is fitted using the battery health status change dataset through a double exponential model to generate the second fitted data.

[0095] S13. Using a linear model, the battery health status change dataset is used to fit the trajectory, generating the third fitting data.

[0096] S14. Using a multinomial model, trajectory fitting is performed on the battery health state change dataset to generate the fourth fitting data.

[0097] S15. Using the Fairhast algorithm model, trajectory fitting is performed on the battery health state change dataset to generate the fifth fitting data.

[0098] S16. Select data from the first, second, third, fourth, and fifth fitted data that are greater than the preset effect threshold, and generate the effect dataset.

[0099] S17. Use the model corresponding to the maximum value in the effect dataset as the initial lifetime empirical model.

[0100] The preset effect threshold refers to the critical value that is set in advance based on actual needs to select the best fitting effect.

[0101] In this embodiment of the invention, to obtain a reasonable mathematical model, single-exponential models, double-exponential models, linear models, polynomial models, and the Verhulst algorithm model are used to repeatedly fit the capacity decay trajectory of the energy storage system battery, obtaining first, second, third, fourth, and fifth fitted data. These are then compared with a preset effect threshold, and data exceeding the preset threshold from the first, second, third, fourth, and fifth fitted data are selected to generate an effect dataset. Next, the function corresponding to the best fitted effect is used to construct an initial lifetime empirical model; that is, the model corresponding to the maximum value in the effect dataset is used as the initial lifetime empirical model.

[0102] Step 203: Use the parameter identification method to identify parameters of the initial life empirical model and generate undetermined coefficients.

[0103] In this embodiment of the invention, assuming the initial lifetime empirical model corresponds to a double exponential model, the values ​​of specific parameters to be determined are first determined using the parameter identification method based on the original double exponential model and the voltage and current changes over time during the specific charging and discharging process, as shown in the following formula, where Cycle is the number of battery cycles, and Q... Discharge The coefficients a, b, c, and d represent the healthy operating state of the battery.

[0104] Q Discharge =a·e b·Cycle +c·e d·Cycle .

[0105] Step 204: Update the initial lifetime empirical model using the battery health state change dataset and undetermined coefficients to generate the intermediate lifetime empirical model.

[0106] Furthermore, step 204 may include the following sub-steps S21-S28:

[0107] S21. Use the probabilities of each random factor in the battery health status change dataset to determine the initial weights of the random factors.

[0108] S22. Substitute the change data corresponding to the initial weights and random factors into the corresponding weighted average functions to construct the initial model parameter set.

[0109] S23. Take the minimum initial model parameter corresponding to each random factor in the initial model parameter set as the target model parameter and generate an intermediate model parameter set.

[0110] S24. Validate the estimation effect of the intermediate model parameter set and generate validation data.

[0111] S25. When the verification data shows an error exceeding the threshold, the initial weights of random factors are dynamically adjusted using a forgetting factor to generate target weights.

[0112] S26. Take the target weight as the initial weight, and jump to execute the steps of substituting the change data corresponding to the initial weight and random factors into the corresponding weighted average function to construct the initial model parameter set.

[0113] S27. When the verification data shows that the errors are all within the allowable range, the intermediate model parameter set at the current moment is used as the target model parameter set.

[0114] S28. Update the initial lifetime empirical model using the target model parameter set and undetermined coefficients to generate the intermediate lifetime empirical model.

[0115] In this embodiment of the invention, parameters are first updated and calibrated using data stored in the Energy Storage System (ESS). Parameter identification is performed based on offline data-driven methods. Then, the parameters are automatically updated periodically according to the designer's intent and the needs of the ESS, and the model order is reselected based on battery type and application characteristics to achieve adaptive parameter calculation. In situations where models and parameters are easily affected by uncertain application environments and undergo significant changes, a least squares algorithm based on a forgetting factor is proposed, derived from the recursive least squares algorithm developed from adaptive filtering theory. This algorithm reduces the information content of old data by incorporating a forgetting factor into the measurement data, thereby creating conditions for supplementing new data. The specific operation steps are as follows: Each random factor is weighted according to its probability of occurrence; that is, the initial weights corresponding to each random factor in the battery health state change dataset are determined using the probabilities of each random factor. Then, the estimation performance is expressed as a weighted average function of each random factor; that is, the initial weights and the change data corresponding to each random factor are substituted into the corresponding weighted average function to construct an initial model parameter set. Finally, the minimum initial model parameter is used to determine the target model parameter set, generating an intermediate model parameter set. Next, the estimation effect of the parameters is verified under each random factor, that is, the estimation effect of the intermediate model parameter set is verified to generate verification data. For cases where the error exceeds the threshold, a forgetting factor is introduced to dynamically adjust its weights and generate target weights. The target weights are used as initial weights, and the process jumps to execute the steps of substituting the initial weights and the changed data corresponding to the random factors into the corresponding weighted average functions to construct the initial model parameter set. The above process is repeated until the model error under all random factors is within the allowable range.

[0116] Step 205: Perform data preprocessing on the discharge to cutoff voltage time data corresponding to the battery health status change dataset to generate intermediate feature data.

[0117] Furthermore, step 205 may include the following sub-steps S31-S35:

[0118] S31. Extract the feature values ​​from the discharge to cutoff voltage time data corresponding to the battery health state change dataset to generate initial feature data.

[0119] S32. Perform repeated observation processing and default value detection and removal on the initial feature data to generate the first preprocessed data.

[0120] S33. Perform outlier detection and removal on the first preprocessed data to generate the second preprocessed data.

[0121] S34. Smooth the second preprocessed data to generate the third preprocessed data.

[0122] S35. Encode the third preprocessed data to generate intermediate feature data.

[0123] In this embodiment of the invention, when the aforementioned model fails to accurately estimate the State of Health (SOH) of the same battery under different environmental conditions, key features of the current and voltage change curves over time during charging and discharging, such as the Discharge Time to Cut-off Voltage (DTCV), are extracted and the characteristics of their data changes are analyzed.

[0124] The extracted data of Discharge Time to Cut-off Voltage (DTCV) is preprocessed, specifically by extracting feature values ​​from the DTCV data corresponding to the battery health state change dataset to generate initial feature data. This initial feature data undergoes further preprocessing, including: data cleaning (repeated observation processing and default value detection and removal) to generate first-level preprocessed data; outlier processing (outlier detection and removal) to generate second-level preprocessed data; noise filtering (smoothing) to generate third-level preprocessed data; and feature encoding to generate intermediate feature data. Furthermore, obvious anomalies can be removed through feature creation and other methods.

[0125] Step 206: Use a neural network method to denoise the intermediate feature data and generate the target feature data.

[0126] In this embodiment of the invention, a neural network method is used to denoise the intermediate feature data to generate target feature data. Then, the relationship between DTCV and the number of iterations is studied to obtain the DTCV variation curve as the number of iterations increases. Finally, the smoothing spine function is used to fit the relationship between DTCV and the number of iterations and incorporate it into the initial lifetime empirical model to form the target lifetime empirical model.

[0127] Step 207: Use a smooth approximation fitting function to characterize the relationship curve between the target feature data and the battery health status change dataset corresponding to the number of cycles, and generate relationship curve data.

[0128] In this embodiment of the invention, when the intermediate lifetime empirical model is a double exponential model, in scenarios where changes in the charging and discharging process or sudden cessation cause relaxation phenomena, the traditional double exponential model is used to accurately estimate the battery's healthy operating status. Key DTCV data features are extracted and denoised to obtain target feature data. Furthermore, curves showing voltage changes over time during charging and discharging under different cycles are constructed. A smoothing spline fitting function is used to characterize the relationship between the target feature data and the cycle number, resulting in relationship curve data.

[0129] Step 208: Update the intermediate lifetime empirical model using relation curve data to generate the target lifetime empirical model.

[0130] In this embodiment of the invention, when the intermediate lifetime empirical model is a bi-exponential model, a smoothing spline fitting function is used to characterize the relationship curve between the target feature data and the cycle number. Finally, this curve is added to the traditional bi-exponential model to form an improved bi-exponential model, as shown in the following formula:

[0131] Q Discharge1 =a·e b·Cycle +c·e d·Cycle(DTCV) .

[0132] Step 209: Update the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model.

[0133] Further, step 209 may include the following sub-steps S41-S42:

[0134] S41. Approximate the relationship between battery health status and cycle number corresponding to the target life empirical model by piecewise linearization to generate linear constraints.

[0135] S42. Update the initial economic scheduling model using linear constraints to generate the target economic scheduling model.

[0136] In this embodiment of the invention, a target lifetime empirical model is incorporated as a constraint into the economic dispatch model of the power system. Since the target lifetime empirical model is a nonlinear equation, directly substituting it into the economic dispatch model would render it unsolvable. Therefore, the relationship between battery health state and cycle count corresponding to the target lifetime empirical model is approximated as piecewise linearized to generate linear constraints. These linear constraints are then incorporated into the initial economic dispatch model for solution, generating the target economic dispatch model. Based on this, correlation analysis is used to obtain the quantitative relationship between the economic dispatch results and energy storage lifetime parameters, further investigating the impact of energy storage lifetime decay characteristics on the economic dispatch of the power system.

[0137] Step 210: Input the battery health status change dataset into the target lifetime empirical model and the target economic dispatch model respectively to generate the corresponding state estimation and economic dispatch data of the energy storage battery.

[0138] In this embodiment of the invention, the specific implementation process of step 210 is similar to that of step 105, and will not be repeated here.

[0139] In this embodiment of the invention, firstly, the battery's state of health (SOH) is obtained as a function of increasing cycle number through experiments on the battery degradation characteristics of the energy storage system. This provides a reference for verifying the accuracy of the improved double-exponential energy storage model. Based on this curve, different function forms are tentatively selected to repeatedly fit the capacity degradation trajectory of the power battery. The double-exponential model is then determined as the SOH estimation model based on the experimental results. Next, a parameter identification method based on neural networks is used to determine the undetermined coefficients, and the accuracy of the double-exponential model under different environments (such as uncertain battery aging conditions, application environment, and other random factors) is verified. Addressing the issue of insufficient accuracy in estimating the SOH of the energy storage system battery, features of the discharge-to-cutoff voltage time data are extracted from the current-voltage time curve of the original charge-discharge process. A curve showing the change of DTCV (Deep Transmission Voltage Variable) as the number of battery cycles increases is constructed. The smoothing spine function is used to fit the relationship curve between DTCV and Cycle and incorporate it into the exponential term of the original double-exponential model, forming an improved double-exponential model for estimating the SOH of the energy storage battery. The numerical examples demonstrate that, compared to the traditional double-exponential model, the proposed model fully considers the impact of battery polarization and exhibits stronger adaptability to scenarios where the maximum available battery capacity changes abruptly due to battery relaxation. Finally, the proposed double-exponential model is incorporated into the original power system economic dispatch, and the economic dispatch results are represented as a piecewise analytical function of energy storage lifetime, thereby quantitatively analyzing the impact of energy storage lifetime degradation characteristics on economic dispatch.

[0140] Specific examples of evaluation using the method of this invention:

[0141] In this section, we first monitor the charging and discharging process of the energy storage system battery to obtain the voltage and current variation curves over time at different cycle numbers, as follows: Figure 3 As shown.

[0142] from Figure 3 As can be seen, under constant current and constant voltage (CCCV) conditions, the charging voltage gradually rises from its initial value to a constant value and then remains essentially constant. The current initially remains at a higher value for a period, then rapidly decreases to a lower value, and finally decreases slowly in a trickle-like manner. At the end of the charging process, the current also drops to zero. During the discharging process, the voltage initially drops rapidly to a certain value and then maintains a gradual downward trend, ending the discharge process when the discharge cutoff voltage is reached. The current remains constant throughout the discharge process.

[0143] Further analysis of the voltage and current curves during the battery charge and discharge process at different cycle numbers reveals that as the number of battery cycles increases, the voltage and current changes during charge and discharge become more and more drastic, and the battery life also decreases significantly.

[0144] When improving the original exponential double-exponential model, it can be found that the characteristics of the discharge to cutoff voltage time change significantly with the increase of the Cycle number, such as... Figure 5 As shown, it is clear that the charging voltage rises faster in the last cycle compared to the first cycle, and there are repeated voltage fluctuations as the charging is nearing completion. During the discharge process, the voltage drop rate increases significantly, the charging and discharging time is significantly shortened, and the battery's durability becomes increasingly severe.

[0145] To fully account for the influence of the cycle number on the charge-discharge voltage variation curve, the characteristics of the discharge-to-cutoff voltage time data for different cycle numbers were extracted and analyzed, such as... Figure 4 As shown in the small blue circle, the original data has a certain degree of dispersion and contains some noisy data. By processing it, removing outliers and reducing noise, the improved data can be obtained, as shown in the red curve.

[0146] Furthermore, the relationship between the discharge-to-cutoff voltage time characteristic and the number of cycles was fitted using the smoothing spline function, and then integrated into the exponential term of the double exponential model to form an improved double exponential model for estimating the battery's healthy operating state. The comparison between the estimated results and the actual results is shown in the figure below. Figure 6As shown in the figure, the improved double-exponential model fully considers the relaxation phenomenon caused by changes in the charging and discharging process or sudden cessation, and has a high estimation accuracy for the healthy operating state of the battery.

[0147] Finally, the analytical relationship between energy storage battery planning parameters and economic dispatch results is considered, as shown in the following formula. A schematic diagram of the correlation analysis method between economic dispatch results and energy storage planning parameters is shown below. Figure 7 As shown, the relationship between economic dispatch results and energy storage lifetime is not a simple linear one. In power systems with large-scale energy storage batteries, the value of energy storage lifetime determines its correlation with economic dispatch results and has a direct impact on the dispatch outcome. Therefore, accurate energy storage lifetime values ​​are crucial for the usability of economic dispatch results. Finally, the economic dispatch result is represented as a piecewise analytical function of energy storage lifetime, thereby quantitatively analyzing the impact of energy storage lifetime decay characteristics on economic dispatch. Figure 7 middle, For the optimal solution, E m equal e m equal F m Let G be the objective function of the economic scheduling model. m equal g m equal The meanings of A and Y are explained above, and CR m For the feasible region of the parameters, C 1,life These are the energy storage planning parameters for the first energy storage device (ESS). minF = Kx

[0148] stAx≤Bw+C;

[0149]

[0150] Where F m Let K be the objective function of economic scheduling; K be a constant coefficient matrix; and w be the planning parameters. This is the optimal solution; A, B, and C are all constant parameter matrices, where... Y represents an active constraint; N represents an inactive constraint; T represents transpose. B Y C Y The meaning is as described above; G m equal g m equal All are constant matrices.

[0151] Please see Figure 8 , Figure 8 This is a structural block diagram of a battery state estimation system provided in Embodiment 3 of the present invention.

[0152] Example 3 of this invention provides a state estimation system for an energy storage battery, comprising:

[0153] The battery health status change dataset generation module 801 is used to obtain the capacity data of the energy storage battery, conduct capacity decay experiments using the capacity data, and generate a battery health status change dataset.

[0154] The initial lifetime empirical model determination module 802 is used to perform trajectory fitting using a battery health state change dataset to determine the initial lifetime empirical model.

[0155] The target lifetime empirical model generation module 803 is used to update the initial lifetime empirical model based on the parameter identification method and the discharge to cutoff voltage time data corresponding to the battery health state change dataset, and generate the target lifetime empirical model.

[0156] The target economic scheduling model generation module 804 is used to update the initial economic scheduling model by adopting the linear constraints corresponding to the target lifetime empirical model, and generate the target economic scheduling model.

[0157] The state estimation and economic dispatch data generation module 805 is used to input the battery health state change dataset into the target lifetime empirical model and the target economic dispatch model respectively, and generate the state estimation and economic dispatch data corresponding to the energy storage battery.

[0158] Optionally, the initial lifetime empirical model determination module 802 includes:

[0159] The first fitting data generation module is used to perform trajectory fitting using a single exponential model with a battery health state change dataset to generate the first fitting data.

[0160] The second fitting data generation module is used to perform trajectory fitting using a battery health state change dataset through a double exponential model to generate the second fitting data.

[0161] The third fitting data generation module is used to perform trajectory fitting using a linear model with a dataset of changes in battery health status, and generate the third fitting data.

[0162] The fourth fitting data generation module is used to perform trajectory fitting using a multinomial model on a dataset of changes in battery health status, and generate the fourth fitting data.

[0163] The fifth fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through the Fairhast algorithm model to generate the fifth fitting data.

[0164] The effect dataset generation module is used to select data from the first, second, third, fourth, and fifth fitted data that are greater than the preset effect threshold, and generate the effect dataset.

[0165] The initial lifetime empirical model determination submodule is used to select the model corresponding to the maximum value in the effect dataset as the initial lifetime empirical model.

[0166] Optionally, the target life empirical model generation module 803 includes:

[0167] The module for generating undetermined coefficients is used to identify parameters of the initial life empirical model using the parameter identification method and generate undetermined coefficients.

[0168] The intermediate lifetime empirical model generation module is used to update the initial lifetime empirical model using a battery health state change dataset and undetermined coefficients to generate an intermediate lifetime empirical model.

[0169] The intermediate feature data generation module is used to preprocess the discharge to cutoff voltage time data corresponding to the battery health state change dataset to generate intermediate feature data.

[0170] The target feature data generation module is used to denoise the intermediate feature data using a neural network method to generate target feature data.

[0171] The relation curve data generation module is used to use a smooth approximation fitting function to characterize the relationship curve between the target feature data and the battery health status change dataset corresponding to the number of cycles, and generate relation curve data.

[0172] The target life empirical model generation submodule is used to update the intermediate life empirical model using relation curve data to generate the target life empirical model.

[0173] Optionally, the intermediate lifetime empirical model generation module may perform the following steps:

[0174] The initial weights of the random factors are determined by using the probabilities of each random factor in the battery health status change dataset.

[0175] The initial model parameter set is constructed by substituting the change data corresponding to the initial weights and random factors into the corresponding weighted average function;

[0176] The minimum initial model parameter corresponding to each random factor in the initial model parameter set is used as the target model parameter to generate an intermediate model parameter set.

[0177] The intermediate model parameter set is used to validate the estimation effect and generate validation data.

[0178] When the verification data shows an error exceeding a threshold, a forgetting factor is used to dynamically adjust the initial weights of random factors to generate target weights.

[0179] The target weights are used as initial weights, and the process jumps to execute the steps of substituting the change data corresponding to the initial weights and random factors into the corresponding weighted average functions to construct the initial model parameter set.

[0180] When the verification data shows that the errors are all within the allowable range, the intermediate model parameter set at the current moment is used as the target model parameter set;

[0181] The initial lifetime empirical model is updated using the target model parameter set and undetermined coefficients to generate the intermediate lifetime empirical model.

[0182] Optionally, the intermediate feature data generation module may perform the following steps:

[0183] Extract feature values ​​from the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate initial feature data;

[0184] The initial feature data is subjected to repeated observation processing and default value detection and removal to generate the first preprocessed data;

[0185] The first preprocessed data is subjected to outlier detection and removal to generate the second preprocessed data;

[0186] The second preprocessed data is smoothed to generate the third preprocessed data;

[0187] The third preprocessed data is feature-encoded to generate intermediate feature data.

[0188] Optionally, the target economic scheduling model generation module 804 includes:

[0189] The linear constraint generation module is used to approximate the piecewise linearization of the relationship between battery health state and cycle number corresponding to the target life empirical model, thereby generating linear constraints.

[0190] The target economic scheduling model generation submodule is used to update the initial economic scheduling model with linear constraints and generate the target economic scheduling model.

[0191] This invention also provides an electronic device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the energy storage battery state estimation method as described in any of the above embodiments.

[0192] The memory can be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. The memory has storage space for program code used to perform any of the method steps described above. For example, the storage space for program code may include individual program codes for implementing the various steps in the methods described above. This program code can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the energy storage battery state estimation method described above.

[0193] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the energy storage battery state estimation method as described in any of the above embodiments.

[0194] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0195] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0196] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0197] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0198] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0199] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating the state of an energy storage battery, characterized in that, include: Acquire the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a dataset of battery health status changes. The battery health state change dataset was used to perform trajectory fitting to determine the initial lifetime empirical model; The initial lifetime empirical model is updated based on the parameter identification method and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate the target lifetime empirical model. The initial economic scheduling model is updated using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model; The battery health status change dataset is input into the target lifetime empirical model and the target economic scheduling model, respectively, to generate the state estimation and economic scheduling data corresponding to the energy storage battery.

2. The energy storage battery state estimation method according to claim 1, characterized in that, The step of using the battery health state change dataset to perform trajectory fitting and determine the initial lifetime empirical model includes: The battery health status change dataset is used to perform trajectory fitting using a single exponential model to generate the first fitted data. The battery health status change dataset is used to perform trajectory fitting using a dual exponential model to generate a second fitted data; A third set of fitted data is generated by using the battery health state change dataset with a linear model to fit the trajectory. The fourth fitting data is generated by using the battery health state change dataset to perform trajectory fitting through a multinomial model. The fifth fitted data is generated by using the battery health state change dataset through the Fairhast algorithm model to perform trajectory fitting. Select data from the first, second, third, fourth, and fifth fitted data that are greater than a preset effect threshold to generate an effect dataset; The model corresponding to the maximum value in the effect dataset is used as the initial lifetime empirical model.

3. The energy storage battery state estimation method according to claim 1, characterized in that, The step of updating the initial lifetime empirical model based on the parameter identification method and the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate the target lifetime empirical model includes: The parameter identification method is used to identify the parameters of the initial lifetime empirical model and generate undetermined coefficients. The initial lifetime empirical model is updated using the battery health state change dataset and the undetermined coefficients to generate an intermediate lifetime empirical model. The discharge-to-cutoff voltage time data corresponding to the battery health state change dataset are preprocessed to generate intermediate feature data. The intermediate feature data is denoised using a neural network method to generate the target feature data. A smooth approximation fitting function is used to characterize the relationship curve between the target feature data and the number of cycles corresponding to the battery health state change dataset, thereby generating relationship curve data; The intermediate lifetime empirical model is updated using the relationship curve data to generate the target lifetime empirical model.

4. The energy storage battery state estimation method according to claim 3, characterized in that, The step of updating the initial lifetime empirical model using the battery health state change dataset and the undetermined coefficients to generate an intermediate lifetime empirical model includes: The initial weights of the random factors are determined by using the probabilities of each random factor in the battery health status change dataset. The initial model parameter set is constructed by substituting the change data corresponding to the initial weights and the random factors into the corresponding weighted average function; The minimum initial model parameter corresponding to each random factor in the initial model parameter set is used as the target model parameter to generate an intermediate model parameter set. The intermediate model parameter set is used to verify the estimation effect, and verification data is generated. When the verification data shows an error exceeding a threshold, a forgetting factor is used to dynamically adjust the initial weights of the random factors to generate target weights. The target weight is used as the initial weight, and the process jumps to the step of substituting the change data corresponding to the initial weight and the random factor into the corresponding weighted average function to construct the initial model parameter set. When the verification data shows that all errors are within the allowable range, the intermediate model parameter set at the current moment is used as the target model parameter set. The initial lifetime empirical model is updated using the target model parameter set and the undetermined coefficients to generate an intermediate lifetime empirical model.

5. The energy storage battery state estimation method according to claim 3, characterized in that, The step of preprocessing the discharge-to-cutoff voltage time data corresponding to the battery health state change dataset to generate intermediate feature data includes: Extract feature values ​​from the discharge-to-cutoff-voltage time data corresponding to the battery health state change dataset to generate initial feature data; The initial feature data is subjected to repeated observation processing and default value detection and removal to generate the first preprocessed data; The first preprocessed data is subjected to outlier detection and removal to generate the second preprocessed data; The second preprocessed data is smoothed to generate the third preprocessed data; The third preprocessed data is feature-encoded to generate intermediate feature data.

6. The energy storage battery state estimation method according to claim 1, characterized in that, The step of updating the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model to generate the target economic scheduling model includes: The relationship between the battery health state and the number of cycles corresponding to the target life empirical model is approximated by piecewise linearization to generate linear constraints. The initial economic scheduling model is updated using the linear constraints to generate the target economic scheduling model.

7. A state estimation system for energy storage batteries, characterized in that, include: The battery health status change dataset generation module is used to acquire the capacity data of the energy storage battery, conduct a capacity decay experiment using the capacity data, and generate a battery health status change dataset. The initial lifespan empirical model determination module is used to perform trajectory fitting using the battery health state change dataset to determine the initial lifespan empirical model. The target lifetime empirical model generation module is used to update the initial lifetime empirical model based on the parameter identification method and the discharge to cutoff voltage time data corresponding to the battery health state change dataset, and generate the target lifetime empirical model. The target economic scheduling model generation module is used to update the initial economic scheduling model using the linear constraints corresponding to the target lifetime empirical model, and generate the target economic scheduling model. The state estimation and economic dispatch data generation module is used to input the battery health state change dataset into the target lifetime empirical model and the target economic dispatch model respectively, and generate the state estimation and economic dispatch data corresponding to the energy storage battery.

8. The energy storage battery state estimation system according to claim 7, characterized in that, The initial lifetime empirical model determination module includes: The first fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a single exponential model to generate the first fitting data. The second fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a double exponential model to generate second fitting data. The third fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a linear model to generate third fitting data. The fourth fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through a multinomial model to generate the fourth fitting data. The fifth fitting data generation module is used to perform trajectory fitting using the battery health state change dataset through the Fairhast algorithm model to generate the fifth fitting data. The effect dataset generation module is used to select data from the first fitted data, the second fitted data, the third fitted data, the fourth fitted data, and the fifth fitted data that are greater than a preset effect threshold, and generate an effect dataset; The initial lifetime empirical model determination submodule is used to take the model corresponding to the maximum value in the effect dataset as the initial lifetime empirical model.

9. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the steps of the energy storage battery state estimation method as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the energy storage battery state estimation method as described in any one of claims 1 to 6.

Images