CN-121978537-A - Lithium battery SOC self-adaptive estimation method based on multi-mode fusion

Abstract

The invention discloses a lithium battery SOC self-adaptive estimation method based on multi-mode fusion, which comprises a data acquisition and processing module, a multi-mode fusion estimation module and a self-adaptive parameter optimization module. The method comprises the steps of acquiring multi-source data such as voltage, current and temperature in real time, carrying out online parameter identification by adopting a least square method with a combined self-adaptive forgetting factor and noise covariance, combining with extended Kalman filtering, ampere-hour integration and static calibration, and dynamically fusing according to confidence coefficient to output an optimal SOC estimated value; the method effectively solves the problem that the precision of the traditional single estimation method is reduced under the complex working condition, and remarkably improves the robustness and the accuracy of SOC estimation of the lithium battery under the conditions of full life cycle, wide temperature range and variable load.

Inventors

• DU HAIBO
• ZHAO CHANGYONG
• ZHU MIN
• GONG LONGYAN

Assignees

• 合肥工业大学

Dates

Publication Date

20260505

Application Date

20260306

Claims (3)

1. 1. The lithium battery SOC self-adaptive estimation method based on multi-mode fusion is characterized by comprising the following steps of: Fitting according to lithium battery experimental data A functional relational expression; initializing the state and model of the system; Data is collected and preprocessed in real time; judging whether the static calibration condition is met; under the condition that the static calibration condition is judged to be met, outputting a first SOC estimated value by using a static calibration mode and updating the confidence coefficient under the mode; under the condition that the static calibration condition is not met, estimating the SOC by using a dynamic high-precision calibration mode and updating the confidence coefficient under the mode; Judging whether a dynamic high-precision calibration condition is met; Under the condition that the dynamic high-precision calibration condition is judged to be met, outputting a first SOC estimated value and the confidence coefficient under the dynamic high-precision calibration mode, and then completing the self-updating of the parameters; Under the condition that the dynamic high-precision calibration condition is not met, outputting a first SOC estimated value by using an ampere-hour integral mode and updating the confidence coefficient under the mode; And outputting the second SOC value according to the confidence level arbitration, and writing the state variable into the ROM.
2. 2. The method of claim 1, wherein initializing the state and model of the system comprises: Reading and loading key state data stored in the last operation cycle from the ROM; judging whether the key data are valid or not; under the condition that key data are judged to be effective, initializing a correlation matrix and a model by using the key data; And under the condition that the key data is invalid, initializing a correlation matrix and a model by using preset default parameters.
3. 3. The method of claim 1, wherein outputting the first SOC estimation using the dynamic high-precision calibration mode and updating the confidence level in the mode if the static calibration condition is determined not to be met, comprises: Constructing a battery model; Recursively identifying battery parameters by using a recursion least square method with forgetting factors; Calculating and updating the online covariance estimation of the innovation; evaluating the model uncertainty and generating an adjustment signal; Dynamically generating a joint adjustment factor; Adaptively adjusting forgetting factors and process noise covariance; The first SOC estimation is output based on an extended Kalman filtering algorithm.

Description

Lithium battery SOC self-adaptive estimation method based on multi-mode fusion Technical Field The invention relates to the technical field of lithium batteries, in particular to a lithium battery SOC self-adaptive estimation method based on multi-mode fusion. Background Under the background of rapid popularization of electric automobiles and intelligent energy storage systems, the power battery is used as a core energy storage unit, and the accuracy and reliability of a management system are important. The battery state of charge (SOC) is a key internal state for representing the residual available electric quantity, and the high-precision estimation is carried out on the state of charge, so that the battery is a precondition for realizing high-efficiency energy scheduling, preventing overcharge and overdischarge, guaranteeing the safety of a system and prolonging the service life of the battery. . Currently, the mainstream SOC estimation method mainly comprises an ampere-hour integration method, an open-circuit voltage method, a neural network method and a Kalman filtering method. They all face challenges in different degrees in practical engineering applications. The ampere-hour integration method calculates the throughput of the charge by carrying out time integration on the working current, so as to calculate the SOC change, and has the advantages of definite physical meaning, simple realization and small calculated amount. But its estimation result is severely dependent on the accuracy of the initial SOC, and the error of the initial value will be inherited all the way. And because of inherent measurement accuracy limitation and zero drift of the sensor, tiny deviation in current measurement can be amplified continuously in the long-time integration process to form obvious accumulated errors, so that the SOC estimated value is gradually deviated from a true value, and the method is difficult to be independently used in a scene requiring long-term high-accuracy estimation. The open circuit voltage method is calibrated according to the characteristic that the corresponding relation exists between the terminal voltage and the SOC after the battery is fully placed, and the precision is extremely high under the placed condition. However, in actual operation, the terminal voltage of the battery is affected by the common effects of ohmic internal resistance, electrochemical polarization and concentration polarization, and is difficult to apply to online and real-time dynamic SOC estimation. The neural network method does not depend on the internal mechanism of the battery, but trains a nonlinear mapping model through a large amount of complete test data, so that the SOC is directly deduced according to external characteristics. While exhibiting potential in processing highly nonlinear systems, their performance is highly dependent on coverage and quality of training data. Meanwhile, the complex network model usually accompanies huge calculation overhead and storage requirements, and provides serious challenges for the calculation capacity and memory resources of embedded hardware, so that the large-scale application of the complex network model in BMS products with sensitive cost and limited resources is limited. The Kalman filtering method models the battery as a dynamic system, and optimally estimates the state of the system by constructing a state space equation and utilizing a voltage observation value. It can effectively process measurement noise and implement closed loop correction. But the core bottleneck of this type of algorithm is that its accuracy is highly dependent on the accuracy of the battery model. In practical application, the model parameters of the battery undergo severe and complex time-variation along with the SOC, the operating temperature, the current multiplying power and the health state, and the model cannot accurately describe the dynamic characteristics of the battery. In summary, although the ampere-hour integration method, the open-circuit voltage method, the neural network method and the kalman filter method are currently used as the main current SOC estimation technology, each of them shows unique advantages under specific conditions, but the inherent methodology defects make it difficult to independently meet the daily application requirements of SOC estimation accuracy and robustness under complex dynamic working conditions. Therefore, the method is improved, and a lithium battery SOC self-adaptive estimation method based on multi-mode fusion is provided. Disclosure of Invention The embodiment of the invention provides a lithium battery SOC self-adaptive estimation method based on multi-mode fusion, which enables a system to dynamically select an optimal estimation strategy according to the actual working condition of a battery and realizes high-accuracy estimation of full-scene and full-life-cycle SOC. In order to achieve the above object, an embodiment of the present inve