METHOD AND SYSTEM FOR PREDICTING THE DEGRADATION STATE OF AN ELECTRIC BATTERY

Description

TECHNICAL FIELD

[0001] The present invention relates to the field of electric batteries. Its application is particularly advantageous in the field of predicting the state of degradation of an electric battery. STATE OF THE ART

[0002] The development of electric batteries, such as high-performance lithium-ion batteries, is a crucial research area, particularly in the context of electric vehicles (EVs) where battery longevity and efficiency directly impact vehicle range and overall performance. However, battery aging, especially in lithium-ion (LiB) batteries, is a complex process that can significantly affect battery performance and lifespan, making it an important consideration in LiB design and operation.

[0003] The aging process of LiB is influenced by several factors, including cycling protocols and storage conditions. Therefore, accurately predicting the aging process and developing effective strategies to mitigate it is difficult. Experimental studies on LiB aging are generally very lengthy and expensive.

[0004] An object of the present invention is therefore to offer a faster and more economical solution to predicting the evolution of the degradation state of an electric battery.

[0005] The other objects, features, and advantages of the present invention will become apparent from an examination of the following description and accompanying drawings. It is understood that other advantages may be incorporated. SUMMARY

[0006] The present invention relates to a method for predicting the state of degradation of at least one electric battery, preferably for an electric vehicle, said method being configured to be executed by at least one electronic system, preferably said electronic system being configured to communicate with at least one database and / or at least one set of sensors, said method comprising: a. A first collection step, preferably executed by at least one first collection module of said electronic system, of at least one piece of information relating to an initial cycling protocol, preferably said initial cycling protocol corresponding to the current cycling protocol of said electric battery, said initial cycling protocol being taken from at least the following plurality of cycling protocols: i. A first cycling protocol, said first cycling protocol being a constant current and constant voltage cycling protocol; ii. A second cycling protocol, said second cycling protocol being a pulsed current cycling protocol; iii. A third cycling protocol, said third cycling protocol being a multi-stage current cycling protocol;b. A selection step, preferably executed by at least one selection module of said electronic system, of at least two prediction cycling protocols taken from said plurality of cycling protocols, including the first and second cycling protocols; c. A second collection step, preferably executed by at least one second collection module of said electronic system, of at least one data point relating to at least one set of parameters, said set of parameters comprising at least one of the following parameters: i. An initial state of charge of the electric battery; ii. A final state of charge of the electric battery; iii. A number N of charge-discharge cycles of said battery; iv. A temperature T of the electric battery; v. A charge rate C, i.e., a charging speed, associated with at least one cycling protocol taken from at least said plurality of cycling protocols;d. A processing step, preferably executed by at least one computing module of said electronic system, by machine learning from at least one pre-trained mathematical model, from said at least one piece of information, from said at least one data point, and from each selected prediction cycling protocol, said processing step comprising at least: i. A step of computing at least one descriptor by applying said pre-trained mathematical model in a multidimensional computing space, said multidimensional computing space being configured such that each parameter of said parameter set defines a dimension of said multidimensional computing space; ii. A step of classifying the state of degradation from said descriptor relative to a set of predetermined classes;iii. A step for calculating the temporal evolution of said degradation state of said electric battery, by calculating a plurality of degradation states of said electric battery, as a function of at least one variable, said variable being taken from at least: A) An initial state of charge of the electric battery; B) A final state of charge of the electric battery; C) A pressure applied to at least a part of the electric battery; D) A number N' of charge-discharge cycles of said electric battery; E) A temperature T' of the electric battery; F) A first charge rate C1 relating to the first cycling protocol; G) A second charge rate C2 relating to the second cycling protocol; H) A third charge rate C3 relating to the third cycling protocol; I) A first discharge rate D1 relating to the first cycling protocol; J) A second discharge rate D2 relating to the second cycling protocol;K) A third discharge rate D3 relating to the third cycling protocol; L) A charge limit; M) A zero-current pause time; N) Another cycling protocol; iv. A step for obtaining a calculated state of the degradation of said electric battery, said calculated state comprising at least one predictive mathematical model, said predictive mathematical model being configured to mathematically represent the evolution of the degradation state of said electric battery as a function of at least said variable, the process comprising at least one; retroactive control step, by at least one management system (40) of the electric battery (10), of at least one variable according to said predictive mathematical model and / or of at least one predetermined setpoint.

[0007] The present invention cleverly uses a mathematical model pre-trained on data from various electric batteries to generate a predictive mathematical model specific to a given electric battery. This results in considerable time and system resource savings because it eliminates the need to rebuild a mathematical model from scratch.

[0008] The present invention improves the quality and reliability of electric batteries. One of the many applications of the present invention is to help electric battery manufacturers improve the quality and reliability of their batteries by reducing the number of premature failures.

[0009] The present invention allows for the optimization of production parameters for an electric battery. Specifically, the manufacturer can use the results of a prediction to adjust production parameters in order to slow battery degradation and thus improve battery quality without relying on cycling tests, which can be both time-consuming and expensive. The present invention is simple and easy to implement, and allows for rapid results, leading to cost savings and improved efficiency in electric battery production.

[0010] The present invention makes it possible to reduce the environmental impact of electric batteries. Indeed, by extending the lifespan of electric batteries, the number of electric batteries required can be reduced, thereby reducing the environmental impact of their production and disposal.

[0011] Furthermore, the present invention enables the optimization of electric battery management. Indeed, by using the results of predicting the evolution of the degradation state of an electric battery, a user can then optimize the management of the electric battery by adjusting the cycling and / or storage conditions, for example, non-restrictively, to extend the battery's lifespan and prevent premature degradation.

[0012] Furthermore, the present invention enables the planning of maintenance and repairs for electric batteries. Indeed, by anticipating battery degradation, a user can plan maintenance and repairs accordingly, which can reduce costs and improve battery availability.

[0013] Finally, the present invention improves the energy efficiency of the electric battery sector. Indeed, by extending the lifespan of an electric battery, a user can reduce the frequency of battery replacement, thereby reducing energy consumption and improving the overall energy efficiency of the sector.

[0014] The present invention also relates to a computer program product, preferably stored on a non-transient memory medium, comprising instructions which, when executed by at least one of a processor, a computer, executes the process according to the present invention.

[0015] The present invention also relates to a non-transient memory medium comprising a computer program product according to the present invention.

[0016] The present invention also relates to an electronic system configured to perform a process according to the present invention, said electronic system comprising at least: a. A communication module configured to communicate with at least one database and / or at least one set of sensors; b. A first data collection module configured to collect, preferably via the communication module, information relating to the initial cycling protocol of the electric battery; c. A selection module configured to select at least one predictive cycling protocol from among a plurality of cycling protocols; d. A second data collection module configured to collect, preferably via the communication module, data relating to a set of parameters; e. A computing module configured to: i. Process, by machine learning from at least one pre-trained mathematical model, said at least one piece of information, said at least one piece of data, and a predictive cycling protocol;ii. Calculate at least one descriptor by applying said pre-trained mathematical model in a multidimensional computational space; iii. Classify the degradation state of said electric battery from said descriptor relative to a set of predetermined classes; iv. Calculate the temporal evolution of said degradation state of said electric battery, by calculating a plurality of degradation states of said electric battery, as a function of at least one variable, said variable being taken from at least: A) An initial state of charge of the electric battery; B) A final state of charge of the electric battery; C) A pressure applied to at least a part of the electric battery; D) A number N' of charge-discharge cycles of said electric battery; E) A temperature T' of the electric battery; F) A first charge rate C1 relating to the first cycling protocol;G) A second charging rate C2 relating to the second cycling protocol; H) A third charging rate C3 relating to the third cycling protocol; I) A first discharge rate D1 relating to the first cycling protocol; J) A second discharge rate D2 relating to the second cycling protocol; K) A third discharge rate D3 relating to the third cycling protocol; L) A charge limit; M) A zero-current pause time; N) Another cycling protocol; v. Generate at least one predictive mathematical model, said predictive mathematical model being configured to mathematically represent the evolution of the degradation state of said electric battery as a function of at least said variable.

[0017] In particular, during the development of the present invention, a machine learning mathematical model capable of classifying aging conditions based on the degradation rate of a LiB was developed. This model was designed using a real-world dataset from a manufacturer of electric vehicle batteries. This development work made it possible to identify patterns and relationships between the aging process and various factors in a broad multi-parameter space. A pre-trained mathematical model was thus designed to predict the degradation rate of LiBs under different cycling conditions. The present invention has generated very promising results for effectively monitoring and mitigating battery aging in electric vehicles.

[0018] It is worth recalling here that machine learning methods allow us to leverage large amounts of data to identify patterns and relationships between various factors, such as the aging process of an electric battery. This type of pre-trained model thus makes it possible, using data specific to a given electric battery, to predict the evolution of its degradation state based on various variables by generating a new predictive mathematical model specific to that battery.

[0019] Thus, the general principle of using machine learning methods to make estimates and predictions related to the life cycle and degradation trajectory of a Li-ion battery is known from prior art, notably from the article "Joint modeling for early predictions of Li-ion battery cycle life and degradation trajectory" (Energy, Volume 277, (2023), 127633, ISSN 0360-5442, by Zhang Chen, Liqun Chen, Zhengwei Ma, Kangkang Xu, Yu Zhou, Wenjing Shen). However, prior art methods still lack precision and reliability. BRIEF DESCRIPTION OF THE FIGURES

[0020] The aims, objects, features and advantages of the invention will become clearer from the detailed description of an embodiment thereof, which is illustrated by the following accompanying drawings in which: [ Fig.1 ] There figure 1represents a prediction method according to an embodiment of the present invention. Fig. 2 ] There figure 2 represents an electronic prediction system according to an embodiment of the present invention. Fig.3 ] There figure 3 represents a feedback loop for controlling the management system of an electric battery according to an embodiment of the present invention. Fig. 4 ] There figure 4 represents obtaining a decision support graph within the framework of an embodiment of the present invention.

[0021] The drawings are provided as examples and are not intended to limit the scope of the invention. They are schematic representations of the principle intended to facilitate understanding of the invention and are not necessarily to scale with practical applications. In particular, the dimensions are not representative of reality. DETAILED DESCRIPTION

[0022] Before proceeding with a detailed review of embodiments of the invention, optional features that may be used in combination or alternatively are listed below:

[0023] According to one example, the present invention includes at least one display step, by at least one display module of said electronic system, of at least one graphical representation of said predictive mathematical model.

[0024] According to one example, the present invention includes at least one step of interaction by a user with said predictive mathematical model via a user interface of said electronic system.

[0025] For example, during the feedback control stage, the battery management system adjusts the battery's charging and discharging parameters. Based on predictions from the mathematical model and the anticipated time evolution, the battery management system (BMS) can decide whether or not to limit the charging rate, and / or, if necessary, introduce a derating factor for discharge, or conversely, not to introduce any limits or even to extend them. The BMS can thus optimize the battery's lifespan and reduce the risk of premature failure.

[0026] As an example, the electric battery management system operationally selects a cycling protocol from among a plurality of cycling protocols. This decision on the cycling protocol can be applicable to the current charging cycle, subsequent charging cycles, or all future charging cycles.

[0027] For example, the BMS can decide to switch from the first cycling protocol (constant current and constant voltage 'CC-CV') to the second pulsed current cycling protocol. Alternatively, the BMS can decide to switch back. Similarly, the BMS can decide to switch to or from the third cycling protocol.

[0028] For example, the electric battery management system (40) schedules maintenance for the electric battery. This allows the vehicle user to be notified of a necessary service visit to the garage regarding the vehicle's traction battery. The user may also be notified of a future date when the battery's remaining intrinsic capacity will reach a certain threshold (e.g., 90% or 85%).

[0029] According to one example, the piloting or feedback control stage is carried out in real time.

[0030] As an example, the pre-trained mathematical model was trained on data from various electric batteries using electric battery data from manufacturers, integrators or users.

[0031] We note here that the pre-trained mathematical model can concern several variants of batteries in the case of electrochemical cells, both in terms of their intrinsic electrochemistry and in terms of the shape factors of the cells (cylindrical, prismatic, 'pouch', etc.).

[0032] The BMS computer will be able to use the branch of the pre-trained mathematical model that best matches the type of electrochemical cells contained in the electric battery of the vehicle of interest.

[0033] According to an example, said set of parameters includes at least the following parameters: An initial state of charge of the electric battery; a number N of charge-discharge cycles of said battery; a temperature T of the electric battery; a charge rate C, i.e., a charging speed, associated with at least one cycling protocol taken from at least said plurality of cycling protocols. This represents, at a minimum, a current state of the electric battery.

[0034] As an example, the selection step (120) of at least two predictive cycling protocols chosen from the plurality of cycling protocols includes three cycling protocols, namely the first, second, and third cycling protocols. The BMS thus has three cycling protocols at its disposal to optimize the lifespan of the electric battery.

[0035] The use of a fourth cycling protocol is also not ruled out.

[0036] According to an example, at least one of the first collection step and the second collection step respectively includes the acquisition of said information and said data respectively from at least one database and / or at least one set of sensors.

[0037] By way of example, the present invention relates to an electronic system that includes: a. at least one display module configured to display at least one graphical representation of said mathematical model, and b. at least one user interface configured to allow at least one interaction by a user with said predictive mathematical model.

[0038] We note that the battery management system, in other words the BMS computer, is on-board and performs operations locally, but also benefits from access to the pre-trained mathematical model which forms a synthesis of a very large set of knowledge accumulated previously during the mission profiles of electric batteries on a large number of vehicles.

[0039] It is noted that it is possible to have at least part of the pre-trained mathematical model stored in the BMS computer's memory. In one implementation example, however, the pre-trained mathematical model is housed on a server or in the cloud and can be made available to the BMS computer via remote communication methods.

[0040] It is observed that the pre-trained mathematical model is constantly updated by the accumulation of new knowledge regarding the mission profile of electric batteries in fleet vehicles currently in operation. Thus, the pre-trained mathematical model and its descriptor(s) are continuously improved, which increases the relevance of the decisions made by the BMS (Battery Management System) based on the model's predicted evolution.

[0041] Furthermore, the following description, listing the principles, aspects, and implementations of the present invention, along with their specific examples, aims to encompass both their structural and functional equivalents, whether currently known or developed in the future. Thus, for example, it will be understood by those skilled in the art that all the functional diagrams herein represent conceptual views of illustrative circuits incorporating the principles of the present invention. Similarly, it will be understood that all the flowcharts and similar diagrams represent various processes that can be substantially represented on computer-readable media and thus executed by a computer or processor, whether or not that computer or processor is explicitly shown.

[0042] The functions of the various elements shown in the figures, including any functional block referred to as a "processor" or "module," can be performed using dedicated hardware as well as hardware capable of executing software in conjunction with a computer program or appropriate instructions. When provided by a processor, the instructions can be delivered by a single dedicated processor, a single shared processor, or by a plurality of individual processors, some of which may be shared. In certain embodiments of the present invention, the processor may be a general-purpose processor, such as a central processing unit (CPU), for example.Furthermore, the explicit use of the term "processor" should not be interpreted as referring exclusively to hardware capable of executing software and may implicitly include, but not be limited to, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), read-only memory (ROM) for storing software, random-access memory (RAM), and non-volatile storage. Other hardware, both conventional and / or custom, may also be included.

[0043] Software modules, or simply modules that are assumed to be software, can be represented here as any combination of flowchart elements or other elements indicating the execution of process steps and / or a textual description. Such modules may be executed by hardware that is expressly or implicitly represented. Furthermore, it should be understood that a module may include, for example, but is not limited to, computer program logic, computer program instructions, software, firmware, hardware circuits, or a combination thereof that provides the required capabilities.

[0044] The present invention proposes using a machine learning model to classify aging cycling conditions based on the degradation rate of electric batteries, and in particular Li-ion (LiB) batteries, and to generate a predictive mathematical model configured to predict the evolution over time of the health status of a given electric battery as a function of various variables and / or parameters. It should be noted that in the remainder of this description, the terms "electric battery," "Li-ion battery," and "LiB" are interpreted interchangeably.

[0045] The present invention cleverly uses a dataset derived from real-world operating conditions of electric batteries under various conditions to identify the relationships between the aging process and different factors across a broad multiparameter range. In particular, the present invention has non-limiting applications in the automotive field.

[0046] As presented below, the present invention makes it possible to predict the degradation rate of an electric battery, preferably of the Li-ion type, under different cycle / storage conditions.

[0047] Unexpectedly, the present invention makes it possible to identify the various factors that influence the aging process of an electric battery. The significant correlations observed during the development of the present invention allow us to determine which variables have the most significant impact on the degradation rate of an electric battery.

[0048] These variables may correspond to specific aging mechanisms, and the results are essential for optimizing the designs of battery management systems capable of effectively monitoring and limiting the aging of an electric vehicle battery, for example, but not limited to electric vehicles. The present invention enables improved performance, greater reliability, and the resolution of durability issues in an electric vehicle battery, for example, but not limited to electric vehicles.

[0049] The present invention provides a valuable tool for predicting, using at least one predictive mathematical model, the performance and lifespan of electric batteries, preferably Li-ion, advantageously in electric vehicles, under various cycling and storage conditions. The ability to identify the factors influencing battery aging and predict its effects can lead to significant improvements in battery design and use, ultimately contributing to the development of more efficient and sustainable electric vehicles.

[0050] Moreover, the present invention finds a very clever application in the control, in the form of a feedback loop for example, of the battery management system known as BMS, from the English acronym "Battery Management System".

[0051] Indeed, for example, by providing real-time predictions of the degradation state of the electric battery, the present invention enables intelligent control of the BMS.

[0052] Indeed, for example, by regularly and / or continuously monitoring the battery's cycling conditions and inputting the data into the present invention—that is, into a mathematical model configured to mathematically represent the evolution of the battery's degradation state—the battery management system (BMS) can use the generated mathematical model to predict the battery's degradation state based on various parameters and thus adjust at least some of these parameters, such as the battery's charging and discharging parameters. This can help optimize the battery's lifespan and reduce the risk of premature failure.

[0053] Furthermore, the present invention can also help the BMS to plan maintenance and repair schedules based on the calculated and therefore predicted rate of degradation. By planning maintenance and repairs in advance, the BMS can reduce the risk of unexpected failures and minimize downtime.

[0054] Thus, according to a preferred embodiment, the present invention relates to a method for predicting the state of degradation of at least one electric battery. Said electric battery being, for example, configured to cooperate with an electric vehicle.

[0055] The method according to the present invention is advantageously configured to be executed by at least one electronic system. This electronic system, referred to as the prediction system, is preferably configured to communicate with at least one database and / or at least a first set of sensors as described below.

[0056] According to one embodiment, and as illustrated by the figure 1 The method for predicting the degradation state of an electric battery 10 includes at least: a. A first collection step 110, performed by at least one first collection module 210 of said electronic system 200, of at least one piece of information relating to an initial cycling protocol, also called the initial charge / discharge protocol. Preferably, said initial cycling protocol corresponds to the current cycling protocol of said electric battery 10. As discussed below, the first initial cycling protocol is taken from at least the following plurality of cycling protocols: i. A first cycling protocol, said first cycling protocol being a constant current, constant voltage cycling protocol; ii. A second cycling protocol, said second cycling protocol being a pulsed current cycling protocol; iii. A third cycling protocol, said third cycling protocol being a multi-stage current cycling protocol;b. A selection step 120, performed by at least one selection module 220 of said electronic system 200, of at least one prediction cycling protocol from said plurality of cycling protocols; Advantageously, two prediction cycling protocols may be selected; c. A second collection step 130, performed by at least one second collection module 230 of said electronic system 200, of at least one data point relating to at least one set of parameters, said set of parameters comprising at least one of the following parameters: i. An initial state of charge of the electric battery 10; ii. A final state of charge of the electric battery 10; iii. A number N of charge-discharge cycles of said battery 10; iv. A temperature T of the electric battery 10; v. A charge rate C, i.e., a charging speed, associated with at least one cycling protocol taken from at least said plurality of cycling protocols;d. A processing step 140, executed by at least one computing module 240 of said electronic system 200, by machine learning from at least one pre-trained mathematical model, from said at least one piece of information, from said at least one data point, and from said prediction cycling protocol. Advantageously, said processing step 140 comprises at least: i. A computing step 141 of at least one descriptor by applying said pre-trained mathematical model in a multidimensional computing space, said multidimensional computing space being configured such that each parameter of said parameter set defines a dimension of said multidimensional computing space; ii. A classification step 142 of the degradation state from said descriptor relative to a set of predetermined classes;iii. A calculation step 143 of the temporal evolution of said degradation state of said electric battery (10), by calculating a plurality of degradation states of said electric battery, as a function of at least one variable, said variable being taken from at least: A) An initial state of charge of the electric battery 10; B) A final state of charge of the electric battery 10; C) A pressure applied to at least a part of the electric battery 10, said pressure being preferably predetermined and advantageously fixed; D) A number N' of charge-discharge cycles of said electric battery 10; E) A temperature T' of the electric battery 10; F) A first charge rate C1 relating to the first cycling protocol; G) A second charge rate C2 relating to the second cycling protocol; H) A third charge rate C3 relating to the third cycling protocol; I) A first discharge rate D1 relating to the first cycling protocol;J) A second discharge rate D2 relating to the second cycling protocol; K) A third discharge rate D3 relating to the third cycling protocol; L) A charge limit; M) A zero-current pause time; N) Another cycling protocol; iv. A step for obtaining 144 a calculated state of the degradation of said electric battery 10, said calculated state comprising at least one predictive mathematical model, said predictive mathematical model being configured to mathematically represent the evolution of the degradation state of said electric battery 10 as a function of at least said variable, preferably by variation of at least said variable.

[0057] This makes it possible to predict the state of degradation of the electric battery 10 according to various variables.

[0058] Thus, the present invention makes it possible, from certain data of the electric battery, to generate a predictive mathematical model of the state of degradation of the electric battery 10. This predictive mathematical model then makes it possible to predict, by the generation of at least one predictive mathematical model, said state of degradation as a function of various variables, thus making it possible to predict the state of degradation of the electric battery as a function of these various variables.

[0059] Advantageously, the predictive mathematical model is a linear and / or non-linear combination of parameters, weighted or unweighted, allowing a mathematical representation in a multidimensional space of the state of degradation of the electric battery 10.

[0060] According to one embodiment, the first collection step 110 is configured to determine the initial cycling protocol of the electric battery 10, that is, its current protocol, i.e., the one currently applied to it or the one last applied to it. The term "cycling protocol" refers to the process used to charge and discharge an electric battery 10; it can also refer to the process used to test and evaluate the performance, lifespan, and safety of an electric battery 10. Preferably, a cycling protocol involves a series of repeated charges and discharges.

[0061] According to one embodiment, we will consider three distinct charging protocols; however, it should be noted that the present invention can perfectly be implemented for other types of charging protocols existing to date or not.

[0062] Thus, according to a non-limiting embodiment, the first data collection step 110 allows for the collection of at least one piece of data relating to the nature of the initial cycling protocol of a given electric battery 10. Advantageously, this initial cycling protocol corresponds to at least one of the following three protocols: i. A first cycling protocol, also called the constant current and constant voltage (CC-CV) cycling protocol. This first cycling protocol is a widely used protocol that involves initially charging the battery with a predetermined constant current until it reaches a predetermined voltage, then switching to a predetermined constant voltage to complete the charging process. This first cycling protocol aims to balance charging speed and voltage control to prevent overcharging. This first cycling protocol is commonly used in various applications and provides efficient and controlled charging. ii. A second cycling protocol, also called the pulsed current cycling protocol.This second cycling protocol consists of applying short bursts of high current to the battery with intermediate rest periods, i.e., pause times (see, for example, the following article: “Lv, H., Huang, X. & Liu, Y. Analysis on pulse charging-discharging strategies for improving capacity retention rates of lithium-ion batteries. Ionics 26, 1749-1770 (2020).”). It is estimated by those skilled in the art that this charging technique improves the battery's capacity and performance, and also reduces its degradation rate. iii. A third cycling protocol, also called the multi-stage current cycling protocol. This third cycling protocol is more complex than the previous two and involves several charging stages with different current levels (see, for example, the following article: “Romain Mathieu, Olivier Briat, Philippe Gyan, Jean-Michel Vinassa.”).Fast charging for electric vehicles applications: Numerical optimization of a multi-stage charging protocol for lithium-ion battery and impact on cycle life. Journal of Energy Storage, 2021, 40, pp.102756.(10.1016 / j.est.2021.102756). (hal-04087488)"). .

[0063] It should be noted that the cycling protocols described above have a different impact on the state of degradation, and therefore the health, of an electric battery 10, and in particular of a LiB. The present invention, during its development, has shown that this state of degradation is strongly correlated with the operating conditions of said electric battery 10.

[0064] Ingeniously, each protocol can be characterized by several specific parameters. Thus, the degradation of electric batteries 10 and in particular LiBs depends, preferably, on a wide set of cycle parameters, i.e. charge-discharge cycle parameters.

[0065] Next, selection step 120 allows the selection of a predictive cycling protocol, that is, a cycling protocol that may differ from the initial cycling protocol and which will serve as a parameter for the future prediction. Thus, the user can decide to predict the degradation state of an electric battery 10 if it is subjected to a particular, i.e., selected, cycling protocol.

[0066] Advantageously, the second data collection step 130 allows for the collection of at least one data point, and preferably a plurality of data points, relating to at least one parameter, and preferably to several parameters. This step allows for the collection of information, data, on the initial state of the electric battery 10, and on its current state. This data is then used to define the state of the electric battery 10.

[0067] We will now describe an example of parameters considered by the present invention, and in particular during the second collection step 130. These various parameters have been cleverly selected because they represent the parameters of an electric battery 10 that can have the most significant impact on its degradation rate: a. Initial state of charge of the electric battery 10: The initial state of charge of the electric battery 10 corresponds to the amount of electrical charge it holds relative to its maximum capacity before undergoing a charge-discharge cycle, i.e., a cycling; b. Final state of charge of the electric battery 10: The final state of charge of the electric battery 10 corresponds to the amount of electrical charge it holds relative to its maximum capacity after undergoing a charge-discharge cycle, i.e., a cycling; c. Number N of charge-discharge cycles of said electric battery 10: The number N of charge / discharge cycles, i.e., cycling, is a parameter that can significantly impact the degradation of a LiB. Indeed, the higher this number, the higher the rate of battery degradation.Indeed, the number of repeated cycles of a battery can lead to several degradation mechanisms that reduce the capacity of the electric battery 10 and its performance over time; d. The temperature T of the electric battery 10: During the development of the present invention, it was observed that the operating temperature of the electric battery 10 can play an important role in understanding its degradation rate. For example, by varying the temperature within a specific range, it is possible to evaluate the performance and behavior of the electric battery 10 under different thermal conditions. e. The charge rate C: The charge rate C corresponds to the charging speed, that is, the rate at which the electric battery 10 is charged. Preferably, this charge rate is a function of the cycling protocol considered.During the development of the present invention, it was observed that the charging rate can have a significant impact on the behavior and degradation patterns of a LiB. Thus, it was observed that higher charging rates can accelerate the aging process and reduce the overall lifespan of the electric battery 10. The development work of the present invention yielded valuable information on the performance of the electric battery 10 under different charging conditions, thereby enabling the development of optimization strategies.

[0068] These five parameters have been identified as having a significant effect on the state of degradation of an electric battery 10.

[0069] Next comes processing step 140. This processing step uses information relating to the initial cycling protocol, the data collected regarding the operating parameters of the electric battery 10, and the choice of the predictive cycling protocol. With these elements and a pre-trained mathematical model, the present invention makes it possible to generate at least one descriptor. This descriptor is then capable of classifying different degradation states of the electric battery 10 according to various variables and parameters. Advantageously, the descriptor classifies the degradation state of the battery 10 according to a set of predetermined classes, such as: a first class called "high loss," a second class called "very high loss," a third class called "low loss," and a fourth class called "moderate loss." The first class can thus correspond to a high degradation rate.The second class may correspond to a very high rate of degradation. The third class may correspond to a low rate of degradation. The fourth class may correspond to a moderate rate of degradation.

[0070] Advantageously, the present invention, by exploiting at least some of these parameters and preferably other so-called secondary parameters, makes it possible to generate a so-called pre-trained mathematical model. Indeed, by using data from manufacturers, integrators, or even users, the development of the present invention has made it possible to design what is called artificial intelligence, that is to say, a pre-trained mathematical model, capable of processing data from parameters relating to an electric battery 10 in order to then design a predictive mathematical model of the degradation state of said electric battery 10, preferably via a prediction of said degradation state as a function of various variables.

[0071] Note that a variable and a parameter can refer to the same data or physical quantity. However, we will speak of a parameter when it is a collected data point and of a variable when it is a data point intended to vary for the calculation of a prediction, for example.

[0072] Preferably, data collection was an important step in building a predictive mathematical model based on machine learning during the development of the present invention. In this type of approach, the quality and quantity of the collected data have a direct impact on the accuracy and effectiveness of the resulting predictive mathematical model. Preferably, the collected data includes a wide range of parameters to provide a comprehensive view of the performance of the electric battery 10 under different conditions, enabling the machine learning model to make accurate predictions, i.e., to generate a predictive mathematical model.

[0073] The parameters considered for the design of the pre-trained mathematical model include the parameters described above, as well as additional or secondary parameters. These additional parameters are advantageously the following:a. A pressure P applied to at least a portion of the electric battery 10, said pressure preferably being predetermined and advantageously fixed. This parameter relates to the application of a pressure P on the electric battery 10 during a cycle. This parameter adds a dimension of mechanical constraints to the mathematical model during its design, relative to the performance of the electric battery 10 and its degradation under conditions of mechanical stress; b. A discharge rate D related to the cycling protocol of the electric battery 10: During the development of the present invention, different discharge rates were applied to evaluate their impact on the aging process of the electric battery 10 and its overall performance. It appears that the discharge rate D has a significant impact on the performance degradation of an electric battery. c.A charge limit L: This involves imposing a limit on the charging of an electric battery, such as 80% of its total charging capacity. It appears that using a charge limit L has a significant impact on the performance degradation of an electric battery. d. A zero-current pause time Tp: This involves imposing a pause time during the charging and / or discharging of an electric battery 10. During this rest period, the electric battery 10 has the opportunity to stabilize and recover from the stress imposed during the previous charge or discharge cycle. The duration of the rest period, i.e., the pause time, can influence the performance of the electric battery 10. It should be noted that excessively short rest periods can lead to increased stress on the electric battery 10, potentially accelerating degradation mechanisms.Conversely, long rest periods can allow for better relaxation of chemical stresses, and thus improve the performance and longevity of the electric battery 10.

[0074] We will now present the variables that can be used to predict the degradation state of an electric battery 10 by the present invention. These variables include parameters that were at least partially used in the design of the pre-trained mathematical model. These so-called training parameters also include the parameters previously discussed. These training parameters become prediction variables when predicting the degradation state of the electric battery 10. These prediction variables are designed to be used as adjustable variables, preferably by a user, when predicting the degradation state of the electric battery 10. These variables are preferably the following: a. An initial state of charge of the electric battery 10; b. A final state of charge of the electric battery 10; c. A pressure P' applied to at least a portion of the electric battery 10, said pressure preferably being predetermined and advantageously fixed; d. A number N' of charge-discharge cycles of said electric battery 10; e. A temperature T' of the electric battery 10; f. A first charge rate C1 relating to the first cycling protocol; g. A second charge rate C2 relating to the second cycling protocol; h. A third charge rate C3 relating to the third cycling protocol; i. A first discharge rate D1 relating to the first cycling protocol; j. A second discharge rate D2 relating to the second cycling protocol; k. A third discharge rate D3 relating to the third cycling protocol; l. A charge limit L'; m. A pause time Tp' at zero current; n. Another cycling protocol;

[0075] It should be noted that for the design of the pre-trained mathematical model, the variables / parameters considered were carefully controlled and varied within a predetermined range to ensure that the data were representative of real-world conditions. This data collection process involved running each protocol multiple times and collecting data at regular intervals. This data was then stored in a database for processing and training at least one machine learning model.To this end, techniques well-known to those in the field (see, for example, the book: "An Introduction to Statistical Learning," Gareth James, Daniela Witten, Trevor Hastie, Robert Tibshirani, Springer New York, NY, 2013, 978-1-4614-7138-7), such as k-means and support vector machines (SVMs), were used to predict the degradation rate of the electric battery 10 as a function of given parameters. This allowed for the design of a pre-trained mathematical model, which serves as the basis for creating the predictive mathematical model according to the present invention. This predictive mathematical model is specifically designed for a given electric battery 10 and not for a so-called "generic" electric battery, as in the case of the pre-trained mathematical model.

[0076] Creating a predictive mathematical model can be long and complex; therefore, the present invention uses a pre-trained mathematical model as a starting point for developing a predictive mathematical model specific to an electric battery 10, this approach allows for considerable savings in time and system resources.

[0077] During the development of the present invention, a data preprocessing step proved very useful. Indeed, once the data was collected, it was preprocessed to adapt it to machine learning algorithms. This included cleaning the data, removing all irrelevant or redundant features, and transforming the data into a digital format that could be fed into the machine learning models.

[0078] Next, the preprocessed data were divided into training and test sets. The k-means technique was used to group similar data points, and then SVM support vector machines were trained on each group or cluster to predict the degradation rate of the electric batteries for a particular group of data.

[0079] As a reminder, unsupervised k-means analysis aims to identify k groups in the dataset by optimizing the distances between each point and data showing similarities and maximizing cluster separations using predefined metrics known to the person in the business.

[0080] The advantage of using a multi-parameter classification is that it can handle data and capture the non-linear relationships between input parameters and the degradation state of the electric battery 10. It provides a flexible and adaptable framework to integrate different functionalities and optimize decision boundaries to accurately classify the degradation of the electric battery 10.

[0081] Once the support vector machine (SVM) is created, it can be used to predict the ranking of new data points. To facilitate understanding and decision-making for the user, a decision diagram, such as the one illustrated in [reference], is provided. figure 4It can be created and used to illustrate the boundaries between different classes in a dataset of the SVM model. The decision diagram 60 is a visual representation of the decision boundaries produced by the SVM classifier. In one embodiment, the decision diagram 60 can be generated by the predictive mathematical model.

[0082] A decision plot can show the decision boundaries that the SVM has learned from the training data. Points on one side of the boundary are classified as belonging to one class, while points on the other side are classified as belonging to a different class. Advantageously, and as described later, this data can be visualized based on different pairs of input parameters, i.e., according to various parameters such as those described previously. Therefore, the user can control the plotting axis, that is, choose which input variables they wish to visualize.

[0083] Advantageously, the method according to the present invention may also include at least one display step, by at least one display module 260 of said electronic system 200, of at least one graphical representation 60 of said second mathematical model. This display step is, for example, very useful for representing the calculated state of degradation to a user via a screen. It is easy to imagine that the present invention could be used in an electric vehicle and that the driver could, via the dashboard, know when the battery 10 is likely to fail and therefore know when maintenance of the battery 10 should be scheduled.

[0084] Preferably, the method according to the present invention may also include a step of interaction by a user with said predictive mathematical model via a user interface 260 of said electronic system 200. In this situation, the user could, for example, vary a variable of the prediction and thus see in real time the evolution of the degradation state of the electric battery 10 over time. For example, the user could charge their battery 10 according to the first cycling protocol, but when they make this prediction and switch to the second cycling protocol, they could then realize that the second cycling protocol might be more advantageous for their battery 10, taking into account the battery and its specific parameters.

[0085] Thus, for example, the development of a graphical user interface (GUI) 260 is a step in the development of the present invention which has made the second predictive mathematical model more accessible to a wider audience, including researchers, engineers and designers working on battery systems, and especially to the end user, for example the driver of an electric vehicle 10. The graphical interface 260 preferably provides an intuitive and user-friendly interface which allows the user to easily control the various input parameters which affect the aging process of lithium-ion batteries 10, i.e. which affect the prediction.

[0086] This tool allows the user to enter their own cycling data and obtain a prediction of the degradation rate of their electric battery 10. The user can also visualize the predicted results in the form of decision graphs 60. This type of graph can include various areas 61, 62, 63, and 64 indicating different scenarios based on input parameters. Thus, the present invention also enables better decision-making. The figure 4 shows an example of the implementation of such a graphical interface 260. In this figure the first class 61 is illustrated, as well as the second class 62, the third class 63 and the fourth class 64, as previously described.

[0087] According to an advantageous embodiment, the method may also include at least one feedback control step, via the BMS 40 for example, of at least one parameter of the battery 10. This embodiment is illustrated in figure 3. In this figure, a feedback control loop 50 is illustrated. Indeed, the electronic system 200 having generated the predictive mathematical model via the process 100, it is then possible for the BMS 40 to exploit this predictive mathematical model to adjust one or more parameters of the electric battery 10, preferably in real time, so as to optimize its performance.

[0088] It should be noted that, according to one embodiment, the first collection step 110 of said information relating to an initial cycling protocol may include a step of acquiring said information from at least one database 20 and / or at least one set of sensors 30.

[0089] According to one embodiment, the electronic system is in communication, via a communication module 260, with a database 20 and / or a set of sensors 30. This set of sensors 30 is configured to allow the measurement, preferably in real time, of various parameters of an electric battery 10.

[0090] The present invention also relates to a computer program product comprising a plurality of instructions which, when executed by at least one processor, execute method 100 according to the present invention.

[0091] The present invention also relates to a non-transient memory medium comprising a computer program product according to the present invention.

[0092] We will now describe, through the figure 2, an electronic system 200 according to an embodiment of the present invention. Preferably, this electronic system 200 comprises at least one processor and advantageously at least one non-transient memory.

[0093] This electronic system 200 is preferably configured to perform process 100. This electronic system 200 advantageously includes at least: a. A communication module 260. This communication module 260 is preferably configured to communicate with at least one database 20 and / or at least one set of sensors 30; b. A first data collection module 210. This first data collection module 210 is advantageously configured to collect, preferably via the communication module 260, at least one piece of information relating to the initial cycling protocol of a given electric battery 10; c. A selection module 220. This selection module 220 is preferably configured to select at least one predictive cycling protocol from among a plurality of cycling protocols; d. A second data collection module 230. This second data collection module 230 may be identical to the first data collection module 210 according to one embodiment.This second data collection module 230 is advantageously configured to collect, preferably via the communication module 260, at least one data point relating to a set of parameters; e. A computing module 240. This computing module 240 advantageously includes at least one pre-trained mathematical model. This computing module 240 is advantageously configured to process said information, said data, and said prediction cycling protocol by machine learning. This computing module 240 is preferably configured to predict a degradation state of said electric battery 10 by generating at least one predictive mathematical model configured to mathematically represent at least one evolution of the degradation state of said electric battery 10.

[0094] Note that a module can include one or more other modules.

[0095] In one embodiment, the electronic system 200 may include a display module. This display module, as previously stated, may be configured to display at least one graphical representation of said predictive mathematical model. This graphical representation may be used to aid in decision-making.

[0096] Preferably, the electronic system 200 may include a user interface 260. This user interface 260 may be configured to allow at least one interaction by a user with the predictive mathematical model. This allows the user, for example, to select the parameters they wish to include in the prediction. It also allows them to vary the value(s) of certain variables to see the impact this has on the prediction and thus on the future state of degradation of the electric battery 10.

[0097] Advantageously, the electronic system 200 may include a control module. This control module is preferably configured to retroactively control at least one parameter of said electric battery 10. According to one embodiment, this control module may be a BMS 40.

[0098] The present invention thus enables an improvement in the performance of an electric battery by anticipating its degradation rate through the generation of a predictive mathematical model allowing predictions.

[0099] The invention is not limited to the embodiments described above and extends to all embodiments covered by the claims.

[0100] All publications referred to in this description are incorporated by reference in their entirety for all purposes of aiding in the understanding of the present invention. Digital references

[0101] 10 Electric Battery 20 Database 30 Sensor Set 40 Electric Battery Management System (BMS) 50 Feedback Loop 60 Decision Graph 61 First Class 62 Second Class 63 Third Class 64 Fourth Class 100 Prediction Process 110 First Collection Stage 120 Selection Stage 130 Second Collection Stage 140 Machine Learning Processing Stage 141 Computation Stage 142 Classification Stage 143 Computation Stage 144 Predictive Mathematical Model Acquisition Stage 200 Electronic System 210 First Collection Module 220 Selection Module 230 Second Collection Module 240 Computation Module 250 Communication Module 260 Graphical User Interface

Claims

1. Method (100) for predicting the degradation state of at least one electric battery (10) in an electric or hybrid motor vehicle, said method (100) being configured to be executed by at least one electronic system (200), said electronic system being configured to communicate with at least one database and / or at least one set of sensors, said method (100) comprising: a. a first collection step (110) of collecting at least one piece of information relating to an initial cycling protocol, said initial cycling protocol being taken from at least the following plurality of cycling protocols: i. a first cycling protocol, said first cycling protocol being a constant current-constant voltage cycling protocol; ii. a second cycling protocol, said second cycling protocol being a pulsed current cycling protocol; iii. a third cycling protocol, said third cycling protocol being a multi-step current cycling protocol; b. a step (120) of selecting at least two prediction cycling protocols taken from said plurality of cycling protocols, including the first cycling protocol and the second cycling protocol; c. a second collection step (130) of collecting at least one datum relating to at least one set of parameters, said set of parameters comprising at least one of the following parameters: i. an initial state of charge of the electric battery; ii. a final state of charge of the electric battery; iii. a number N of charge-discharge cycles of said battery; iv. a temperature T of the electric battery; v. a charge rate C, i.e., a charge speed, associated with at least one cycling protocol taken from at least said plurality of cycling protocols; d. a step (140) of processing, by machine learning based on at least one pre-trained mathematical model, said at least one piece of information, said at least one datum, and each selected prediction cycling protocol, said processing step comprising at least: i. a step (141) of calculating at least one descriptor by applying said pre-trained mathematical model in a multidimensional calculation space, said multidimensional calculation space being configured such that each parameter of said set of parameters defines a dimension of said multidimensional calculation space; ii. a step (142) of classifying the degradation state on the basis of said descriptor relative to a set of predetermined classes; iii. a step (143) of calculating the temporal evolution of said degradation state of said electric battery (10), by calculating a plurality of degradation states of said electric battery, as a function of at least one variable, said variable being taken from at least: • an initial state of charge of the electric battery (10); • a final state of charge of the electric battery (10); • a pressure applied to at least one part of the electric battery (10); • a number N' of charge-discharge cycles of said electric battery (10); • a temperature T' of the electric battery (10); • a first charge rate C1 relating to the first cycling protocol; • a second charge rate C2 relating to the second cycling protocol; • a third charge rate C3 relating to the third cycling protocol; • a first discharge rate D1 relating to the first cycling protocol; • a second discharge rate D2 relating to the second cycling protocol; • a third discharge rate D3 relating to the third cycling protocol; • a charge limit; • a zero-current pause time; • another cycling protocol; iv. a step (144) of obtaining a calculated state of the degradation of said electric battery (10), said calculated state comprising at least one predictive mathematical model, said predictive mathematical model being configured to mathematically represent the evolution of the degradation state of said electric battery (10) as a function of at least said variable, the method comprising at least one step of at least one management system (40) for the electric battery (10) feedback controlling at least one variable as a function of said predictive mathematical model and / or at least one predetermined setpoint.

2. Method (100) according to the preceding claim, comprising at least one step of at least one display module of said electronic system (200) displaying at least one graphical representation (60) of said predictive mathematical model.

3. Method (100) according to either of the preceding claims, comprising at least one step of a user interacting with said predictive mathematical model via a user interface (260) of said electronic system (200).

4. Method (100) according to any of the preceding claims, wherein in the feedback control step, the management system (40) for the electric battery (10) adjusts charge and discharge parameters of the battery.

5. Method (100) according to claim 4, wherein the management system (40) for the electric battery (10) operationally selects a cycling protocol from the plurality of cycling protocols.

6. Method (100) according to claim 4, wherein the management system (40) for the electric battery (10) plans the maintenance schedules of the electric battery.

7. Method (100) according to any of the preceding claims, wherein the feedback control step is performed in real time.

8. Method (100) according to any of the preceding claims, wherein at least one of the first collection step (110) and the second collection step (130) comprises respectively acquiring said piece of information and said datum from at least one database (20) and / or at least one set of sensors (30).

9. Method (100) according to any of the preceding claims, wherein the predictive-trained mathematical model has been trained on data from various electric batteries using electric battery data from manufacturers, integrators, or users.

10. Method (100) according to any of the preceding claims, wherein said set of parameters comprises at least the following parameters: - an initial state of charge of the electric battery; - a number N of charge-discharge cycles of said battery; - a temperature T of the electric battery; - a charge rate C, i.e., a charge speed, associated with at least one cycling protocol taken from at least said plurality of cycling protocols.

11. Method (100) according to any of the preceding claims, wherein the step (120) of selecting at least two prediction cycling protocols taken from the plurality of cycling protocols comprises three cycling protocols, namely the first cycling protocol, the second cycling protocol, and the third cycling protocol.

12. Computer program product, preferably stored on a non-transitory memory medium, comprising instructions, which, when the instructions are performed by at least one from a processor and a computer, executes the method (100) according to any of the preceding claims.

13. Non-transitory memory medium comprising the computer program product according to claim 12.

14. Electronic system (200) configured to execute the method (100) according to any of claims 1 to 6, said electronic system (200) comprising: a. a communication module (250) configured to communicate with at least one database (20) and / or at least one set (30) of sensors; b. a first collection module (210) configured to collect, preferably via the communication module (250), a piece of information relating to the initial cycling protocol of the electric battery (10); c. a selection module (220) configured to select at least two prediction cycling protocols from a plurality of cycling protocols, including the first cycling protocol and the second cycling protocol; d. a second collection module (230) configured to collect, preferably via the communication module (250), a datum relating to a set of parameters; e. a calculation module (240) configured to: i. process, by machine learning based on at least one pre-trained mathematical model, said at least one piece of information and said at least one datum for each selected prediction cycling protocol; ii. calculate at least one descriptor by applying said pre-trained mathematical model in a multidimensional calculation space; iii. classify the degradation state of said electric battery (10) on the basis of said descriptor relative to a set of predetermined classes; iv. calculate the temporal evolution of said degradation state of said electric battery (10), by calculating a plurality of degradation states of said electric battery, as a function of at least one variable, said variable being taken from at least: • an initial state of charge of the electric battery (10); • a final state of charge of the electric battery (10); • a pressure applied to at least one part of the electric battery (10); • a number N' of charge-discharge cycles of said electric battery (10); • a temperature T' of the electric battery (10); • a first charge rate C1 relating to the first cycling protocol; • a second charge rate C2 relating to the second cycling protocol; • a third charge rate C3 relating to the third cycling protocol; • a first discharge rate D1 relating to the first cycling protocol; • a second discharge rate D2 relating to the second cycling protocol; • a third discharge rate D3 relating to the third cycling protocol; • a charge limit; • a zero-current pause time; • another cycling protocol; f. generate at least one predictive mathematical model, said predictive mathematical model being configured to mathematically represent the evolution of the degradation state of said electric battery (10) as a function of at least said variable; and perform at least one step of at least one management system (40) for the electric battery (10) feedback controlling at least one variable as a function of said predictive mathematical model and / or at least one predetermined setpoint.

15. System (200) according to the preceding claim, comprising: a. at least one display module configured to display at least one graphical representation of said mathematical model, and b. at least one user interface (260) configured to allow a user to interact with said predictive mathematical model at least once.