CN-121995224-A - Lithium battery SOC and SOH joint estimation method and system based on iterative denoising

CN121995224ACN 121995224 ACN121995224 ACN 121995224ACN-121995224-A

Abstract

The invention provides a lithium battery SOC and SOH joint estimation method and system based on iterative denoising, belonging to the technical field of lithium battery management, wherein the method comprises the following steps of S1, acquiring charging fragment data of a kth round of a lithium battery and leaving a factory to nominal capacity; the method comprises the steps of S2, inputting charge segment data into an SOH estimation model to obtain an SOH estimated value, S3, calculating the current available capacity based on the SOH estimated value and the factory nominal capacity, S4, correcting the SOC based on the current available capacity through an inverse ampere-hour integration method to obtain a physical correction SOC, S5, inputting the physical correction SOC, the SOH estimated value and k into a residual denoising network to obtain SOC noise data, correcting the physical correction SOC based on the SOC noise data to obtain a residual correction SOC, S6, taking the residual correction SOC as the SOC of the k-1 round to conduct iterative denoising, and outputting the final residual correction SOC and the SOH estimated value after iteration is completed. The method has the advantage that the accuracy and the robustness of SOC and SOH estimation are greatly improved.

Inventors

CHEN WENPING
LIANG QIHUI
YE YING

Assignees

福建星云软件技术有限公司

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1. The lithium battery SOC and SOH joint estimation method based on iterative denoising is characterized by comprising the following steps: step S1, setting an iteration round identifier, setting an initial value of the iteration round identifier as k, and acquiring charging fragment data of a kth round of a lithium battery and factory nominal capacity, wherein the charging fragment data comprises charging voltage, charging current, charging temperature and SOC; S2, inputting the charging fragment data into an SOH estimation model constructed based on a transducer network to obtain an SOH estimation value of a kth round; Step S3, calculating the current available capacity of the kth round based on the SOH estimated value and the factory nominal capacity; s4, correcting the SOC of the kth round based on the current available capacity by an inverse ampere-hour integration method to obtain a physical correction SOC; S5, inputting the physical correction SOC, the SOH estimated value and k into a residual error denoising network to obtain the SOC noise data of the kth round, and correcting the physical correction SOC based on the SOC noise data to obtain a residual error correction SOC; And S6, taking the residual correction SOC as the SOC of the k-1 th round, further constructing the charging fragment data of the k-1 th round, updating the value of the iteration round identifier to be k-1, carrying out iterative denoising of the k-1 th round until the value of the iteration round identifier is 1, and outputting the final residual correction SOC and SOH estimated value.
2. The method for joint estimation of lithium battery SOC and SOH based on iterative denoising of claim 1, wherein in step S1, the expression of the charging fragment data of the kth round is: ; Wherein, the Charge segment data representing the kth round; A charging voltage at time t; a charging current at time t; the charging temperature at time t is represented; SOC indicating the kth wheel and the time t; representing a sequence length of the charging fragment data; The charging voltage, charging current and charging temperature remain unchanged throughout each iteration of the round.
3. The method for joint estimation of lithium battery SOC and SOH based on iterative denoising of claim 1, wherein in step S2, the calculation formula of the SOH estimation value of the kth round is: ; Wherein, the SOH estimation values representing the kth round; charge segment data representing the kth round; () Representing an SOH estimation model; in the step S3, the calculation formula of the current available capacity of the kth round is: ; Wherein, the Representing the current available capacity of the kth round; SOH estimation values representing the kth round; indicating the factory nominal capacity.
4. The method for joint estimation of lithium battery SOC and SOH based on iterative denoising of claim 1, wherein in step S4, the calculation formula of the physical correction SOC is: ; ; Wherein, the Representing the physical correction SOC of the kth round and the t moment; Indicating a charging start time; Indicating the charging end time; Representation of Initial SOC at time; Representing the current available capacity of the kth round; the integration of the charging current from the charging start time to the present time is shown.
5. The method for joint estimation of lithium battery SOC and SOH based on iterative denoising of claim 1, wherein in step S5, the calculation formula of the SOC noise data of the kth round is: ; Wherein, the SOC noise data representing the kth round; representing a residual denoising network; Representing the physical correction SOC of the kth round; SOH estimation values representing the kth round; the calculation formula of the residual correction SOC is as follows: ; Wherein, the Representing the residual correction SOC of the kth round, namely the SOC in the charging fragment data of the kth-1 round; The residual denoising network is constructed based on a deep learning model, a training process is carried out in combination with the SOH estimation model, an end-to-end optimization strategy is adopted, a loss function comprises the mean square error of a final SOC estimation value and an SOC real label and the mean square error of the final SOC estimation value and the SOC real label, gradients are propagated reversely through K rounds of iteration, and model parameters of the SOH estimation model and the residual denoising network are optimized synchronously; The iteration round k is introduced in the residual denoising network through timestep embedding mechanisms.
6. The lithium battery SOC and SOH joint estimation system based on iterative denoising is characterized by comprising the following modules: The initialization module is used for setting an iteration round identifier, setting an initial value of the iteration round identifier as k, and acquiring charging fragment data of a kth round of the lithium battery and factory nominal capacity, wherein the charging fragment data comprises charging voltage, charging current, charging temperature and SOC; The SOH estimation module is used for inputting the charging fragment data into an SOH estimation model constructed based on a transducer network to obtain SOH estimation values of a kth round; the current available capacity estimation module is used for calculating the current available capacity of the kth round based on the SOH estimated value and the factory nominal capacity; the SOC physical correction module is used for correcting the SOC of the kth round based on the current available capacity through an inverse ampere-hour integration method to obtain a physical correction SOC; the SOC residual error correction module is used for inputting the physical correction SOC, the SOH estimated value and k into a residual error denoising network to obtain the SOC noise data of the kth round, and correcting the physical correction SOC based on the SOC noise data to obtain a residual error correction SOC; And the iterative denoising module is used for taking the residual correction SOC as the SOC of the k-1 th round, further constructing the charging fragment data of the k-1 th round, updating the value of the iteration round identifier to be k-1, carrying out iterative denoising of the k-1 th round until the value of the iteration round identifier is 1, and outputting the final residual correction SOC and SOH estimated value.
7. The lithium battery SOC and SOH joint estimation system based on iterative denoising as set forth in claim 6, wherein in the initialization module, the expression of the k-th round of charging fragment data is: ; Wherein, the Charge segment data representing the kth round; A charging voltage at time t; a charging current at time t; the charging temperature at time t is represented; SOC indicating the kth wheel and the time t; representing a sequence length of the charging fragment data; The charging voltage, charging current and charging temperature remain unchanged throughout each iteration of the round.
8. The lithium battery SOC and SOH joint estimation system based on iterative denoising as set forth in claim 6, wherein in the SOH estimation module, the calculation formula of the SOH estimation value of the kth round is: ; Wherein, the SOH estimation values representing the kth round; charge segment data representing the kth round; representing an SOH estimation model; in the current available capacity estimation module, the calculation formula of the current available capacity of the kth round is as follows: ; Wherein, the Representing the current available capacity of the kth round; SOH estimation values representing the kth round; indicating the factory nominal capacity.
9. The lithium battery SOC and SOH joint estimation system based on iterative denoising as set forth in claim 6, wherein in the SOC physical correction module, the calculation formula of the physical correction SOC is: ; ; Wherein, the Representing the physical correction SOC of the kth round and the t moment; Indicating a charging start time; Indicating the charging end time; Representation of Initial SOC at time; Representing the current available capacity of the kth round; the integration of the charging current from the charging start time to the present time is shown.
10. The lithium battery SOC and SOH joint estimation system based on iterative denoising as set forth in claim 6, wherein in the SOC residual error correction module, the calculation formula of the k-th round of SOC noise data is: ; Wherein, the SOC noise data representing the kth round; representing a residual denoising network; Representing the physical correction SOC of the kth round; SOH estimation values representing the kth round; the calculation formula of the residual correction SOC is as follows: ; Wherein, the Representing the residual correction SOC of the kth round, namely the SOC in the charging fragment data of the kth-1 round; The residual denoising network is constructed based on a deep learning model, a training process is carried out in combination with the SOH estimation model, an end-to-end optimization strategy is adopted, a loss function comprises the mean square error of a final SOC estimation value and an SOC real label and the mean square error of the final SOC estimation value and the SOC real label, gradients are propagated reversely through K rounds of iteration, and model parameters of the SOH estimation model and the residual denoising network are optimized synchronously; The iteration round k is introduced in the residual denoising network through timestep embedding mechanisms.

Description

Lithium battery SOC and SOH joint estimation method and system based on iterative denoising Technical Field The invention relates to the technical field of lithium battery management, in particular to a lithium battery SOC and SOH joint estimation method and system based on iterative denoising. Background In the technical field of lithium battery management, accurate estimation of State of Charge (SOC) and State of Health (SOH) is a core problem of ensuring safe operation, prolonging service life and optimizing energy scheduling efficiency of a battery system. SOC is used to characterize the real-time remaining power of a battery, and its common estimation method is the ampere-hour integration method, but the accuracy of this method is severely dependent on the accuracy of the initial SOC and the correct assessment of the maximum available capacity of the current battery. As the battery ages cyclically, its maximum available capacity decays continuously, and if the integration operation is still performed with a fixed nominal capacity, systematic drift of the SOC estimation results will occur, and errors will accumulate continuously over time. SOH, as a key parameter reflecting the degree of battery aging, is generally defined as the ratio of the current maximum available capacity to the factory nominal capacity. The conventional SOH estimation methods are mostly based on open circuit voltage platforms, internal resistance change rules or empirical aging models, but these methods usually take known or accurate SOC as a premise, and in practical application, the SOC itself has noise or drift, so that a coupling estimation problem of interdependence between SOC and SOH is formed, that is, accurate estimation of SOH needs to be conditioned on high-precision SOC input, and correction of SOC depends on correct SOH values. In recent years, deep learning methods have been gradually applied to joint estimation of SOC and SOH, such as predicting battery state directly from sensor data of voltage, current, temperature, etc., using a cyclic neural network (RNN) or a time series convolution network (TCN). The data driving method has the capability of capturing nonlinear dynamic characteristics of the battery, but has obvious limitations that firstly, most methods model SOC and SOH estimation as a one-way process, lack a dynamic feedback mechanism and are sensitive to noise in the input SOC, secondly, the existing methods can not effectively distinguish signal-to-noise ratio differences of different characteristics, namely direct measurement values of voltage, current, temperature and the like, although sensor noise is contained, dynamic response is basically reliable, and the SOC becomes a main error source due to integral accumulation effect, if all the characteristics are denoised indiscriminately, not only calculation burden is increased, but also effective information related to aging is possibly weakened, and further the accuracy of SOH estimation is influenced. Meanwhile, the diffusion model shows excellent iterative denoising capability in the fields of image, voice generation and the like, gradually approximates real data distribution through multi-step denoising and denoising processes, and controls denoising strength by means of time-step embedding. However, such methods have not been explored effectively in battery state estimation. The prior art lacks a mechanism for guiding SOC denoising by taking SOH as condition information, and also fails to construct a closed-loop optimization path of denoising-feedback-correction, so that the estimated error is difficult to realize round-by-round suppression and self-correction. In summary, the prior art mainly has the following disadvantages: 1. SOC estimation is typically based on fixed capacity parameters, failing to dynamically adjust the capacity benchmark as the battery ages, resulting in a constant accumulation of estimated bias over long term use; 2. SOH estimation models often take a noisy SOC sequence as an input, noise interferes with feature expression, and model robustness is affected; 3. Most of the existing joint estimation methods are of a one-time modeling structure, lack of an iterative denoising mechanism, and cannot realize a gradual purification process similar to a diffusion model; 4. Data-driven methods tend to ignore electrochemical physical constraints of the battery itself, possibly generating SOC trajectories that do not conform to actual dynamics (e.g., SOC rises non-monotonically during charging); 5. the model cannot distinguish between a coarse adjustment stage and a fine adjustment stage without introducing a perception mechanism for an iteration stage, and overcorrection or result oscillation is easy to cause. Therefore, how to provide a method and a system for jointly estimating the SOC and the SOH of a lithium battery based on iterative denoising, so as to improve the accuracy and the robustness of the estimation of the SOC and the SO