EP4264301A1

EP4264301A1 - Method for estimating the lifespan of an energy storage system

Info

Publication number: EP4264301A1
Application number: EP21839174.6A
Authority: EP
Inventors: Jean-Marie Klein; Franck AL SHAKARCHI; Arnaud Delaille
Original assignee: Powerup; Commissariat a lEnergie Atomique CEA; Commissariat a lEnergie Atomique et aux Energies Alternatives CEA
Current assignee: Powerup; Commissariat a lEnergie Atomique et aux Energies Alternatives CEA
Priority date: 2020-12-21
Filing date: 2021-12-16
Publication date: 2023-10-25
Also published as: FR3118310B1; FR3118310A1; WO2022136098A1

Abstract

One aspect of the invention relates to a method for estimating the lifespan of an energy storage system comprising a plurality of electrochemical cells, the method comprising the following steps: • - defining (S21) an ageing model for the storage system; • - collecting (S22) first values of a state of health, values of a state of charge and data on the operation of the storage system while the storage system is in use, the storage system being subjected to use conditions, the operation data comprising current, temperature and voltage values; • - updating (S23) parameters of the ageing model according to the first state of health values, the state of charge values and the operation data; • - determining (S24) a thermal profile and electrical load profiles under use conditions from the operation data; • - simulating (S25) the ageing of the storage system using the updated parameters of the ageing model, predetermined thermal and electrical load profiles, and the thermal and electrical load profiles under use conditions, which yields a lifespan projection for the storage system.

Description

DESCRIPTION

TITLE: PROCEDURE FOR ESTIMATING THE LIFETIME OF AN ENERGY STORAGE SYSTEM

TECHNICAL AREA

The technical field of the invention is that of electrochemical energy storage systems. The invention relates more particularly to a method for estimating the lifetime of a storage system, in particular to anticipate the maintenance of this system.

STATE OF THE ART

[0002] An electrochemical energy storage system comprises one or more batteries capable of storing electrical energy (in a chemical form) and releasing it in due time.

[0003] There are two types of electrochemical storage systems: stationary storage systems and on-board storage systems.

[0004] Stationary storage makes it possible in particular to ensure the balance between production and consumption of electricity on an electrical distribution network, and in particular to compensate for the variability in the production of renewable energies (solar, wind, etc.) . For example, the electricity produced in excess on a very sunny day can be restored in the evening when demand is greater. Stationary storage also contributes to guaranteeing the quality of the electricity distribution network by limiting the fluctuations caused by the intermittency of the production of renewable energies. Finally, stationary storage makes it possible to meet the needs of isolated sites, which are poorly supplied or not supplied by the distribution networks.

[0005] Stationary storage systems are mainly medium- or high-power large-scale storage systems (of the order of several hundred kW to several tens of MW), having high energies (of the order of several hundreds of kWh to several tens of MWh).

[0006] Conversely, on-board storage systems have a smaller amount of energy (from a few Wh to a few tens of kWh) and lower power (from a few W to a few hundred kW). They are dedicated to applications mobiles. They are mainly used in transport, in particular in electric and rechargeable hybrid vehicles, and in portable electronic devices (telephones, tablets, computers, etc.).

[0007] The batteries of the storage system make it possible to convert the energy of a chemical reaction into electrical energy. There are many battery technologies. Mention may be made, by way of example, of lead-acid batteries, lithium batteries (lithium-metal when the negative electrode is made of metallic lithium, lithium-ion when the lithium remains in the ionic state, lithium polymer when the electrolyte is a polymer), nickel batteries (nickel-cadmium, nickel-metal hydride), sodium-sulfur batteries, magnesium batteries and metal-air batteries.

[0008] Lithium batteries have one of the highest energy density and specific energy and therefore represent a preferred technology for many applications.

[0009] The performance of batteries, in particular their capacity and their electrical resistance, deteriorate over time and, for certain battery technologies such as lithium batteries, even when they are not used. There are thus two modes of battery degradation: calendar degradation: this degradation occurs constantly during the life of the battery, whether it is subjected to electrical stress or not; degradation during cycling: this degradation occurs when the battery is subjected to electrical stress, during charging or discharging.

[0010] The evolution of the performance of the batteries can be monitored thanks to the implementation of indicators such as the state of health or “SOH” (for “State Of Health”). This state of health can be calculated from physical parameters measured directly on the battery.

[0011] It is also possible to estimate the lifetime of a battery, and more generally of an electrochemical storage system, by simulating its aging in connection with electrical stress profiles and a temperature profile. To do this, a previously parameterized aging model is used. [0012] Empirical aging models are the easiest to use. They are parameterized using a database made up of aging test results. They can take into account both calendar degradation and battery cycling degradation. For each degradation mode, several factors influence the degradation rates. The factors cited most often in the literature are the temperature, the state of charge of the battery or "SOC" (for "State Of Charge" in English), the level of electrical stress (in other words, the electrical current ) and the range of SOC (ASOC).

[0013] The documents [“Degradation of lithium ion batteries employing graphite negatives and nickel-cobalt-manganese oxide + spinel manganese oxide positives: Part 1, aging mechanisms and life estimation”, J. WANG et al., Journal of Power Sources, Flight. 269, p. 937-948, 2014] and ["Development of an empirical aging model for Li-ion batteries and application to assess the impact of Vehicle-to-Grid strategies on battery lifetime", Applied energy, Vol. 172, p. 398-407, 2016] describe several examples of an empirical aging model.

Figure 1 is a block diagram of a typical method for estimating the lifetime of a storage system using an empirical aging model.

[0015] Aging tests are carried out in the laboratory on a multitude of electrochemical cells during a preliminary step S1 1 . Electrochemical cells undergo charge and discharge cycles, at different charge and discharge rates and at different temperatures. The parameters of the aging model, which depend on these current regimes and on the temperature, are then determined during a step S12 on the basis of these tests, for example by statistical adjustment of the model to the results of the tests. Then, the aging of the storage system is simulated during a step S13 using the parameterized model. To perform this simulation, current and voltage electrical stress profiles and a temperature profile are applied. The simulation step S13 provides a lifetime projection of the storage system, typically in the form of a decrease in the state of health (SOH) over time.

The service life estimation method according to FIG. 1 is based exclusively on measurements carried out in the laboratory. The conditions in which these measurements are taken do not represent the actual operation of the storage system, the conditions of which are by nature random. The aging model, although representative of an electrochemical cell technology, is therefore not calibrated to faithfully reproduce the behavior of the storage system in real operation.

[0017] Furthermore, the electrical stress profiles and the thermal profile used during the simulation step are general profiles, made up of average values.

[0018] As a result, the lifetime projection obtained using this method is approximate, which prevents in particular from planning the maintenance of the storage system in a detailed manner and/or from optimizing the use of the system. storage to prolong its life.

SUMMARY OF THE INVENTION

[0019] There is therefore a need to provide a method for estimating the lifetime of an energy storage system that is more accurate.

[0020] According to a first aspect of the invention, this need tends to be satisfied by providing a method for estimating a lifetime of an energy storage system comprising a plurality of electrochemical cells, said method comprising the following steps: a) define an aging model of the storage system; b) collecting first values of a state of health, values of a state of charge and operation data of the storage system during use of the storage system, the storage system being subjected to conditions in use, the operation data comprising current, temperature and voltage values; c) updating parameters of the aging model according to the first values of the state of health, the values of the state of charge and the operation data; d) determining a thermal profile and electrical stress profiles under conditions of use from the operation data; e) simulate the aging of the storage system using the updated parameters of the aging model, the predetermined thermal and electrical stress profiles, and the thermal and electrical stress profiles under conditions of use, from which a duration projection results life of the storage system.

[0021] Advantageously, the estimation method further comprises, before the step of updating the parameters of the aging model, the following steps: collecting test data from a characterization test of the storage system; determine a second storage system health value from the test data; and correcting the first health state values according to the second health state value.

[0022] Preferably, the step of defining the aging model comprises the selection of an aging model and the determination of the parameters of the aging model selected from the results of aging tests. The results of aging tests can come from aging tests carried out on the scale of an electrochemical cell and/or from aging tests carried out on the scale of a module comprising several interconnected electrochemical cells.

In one embodiment of the estimation method, the step of updating the parameters of the aging model comprises the following sub-steps: dividing an operating duration of the storage system into time slices and determining average operating conditions over each time slot from the operating data and the state of charge values; calculating for each time slice a state-of-health error at the end of the time-slice, said state-of-health error being equal to the absolute value of the difference between the first value of the state-of-health and a health value simulated using the operation data on the time slice and the aging model to be updated; comparing the state of health error of each time slice to a threshold value; and when the state-of-health error of at least one time slot is greater than the threshold value, the following sub-steps: modifying aging test results with regard to the first state-of-health values and time slice operation data; and determining aging model parameter values from the modified aging test results.

[0024] According to a development of this mode of implementation, the sub-step of modifying the results of aging tests comprises the following operations: for each time slot, identifying among the results of aging tests results of so-called neighboring tests carried out at test conditions close to the average operating conditions of said time slot; identifying for each of the neighboring test results a time window having a duration equal to that of the time slice and starting at a time when the test result reaches the first health state value at the start of the time slice; determining for each of the neighboring test results a health status value at the end of the time window; weighting for each time slice the health state values at the end of the time windows to obtain an experimental health state value at the end of the time slice; calculating a transfer function between the experimental values of state of health at the end of the time slice and the first values of state of health at the end of the time slice; apply the transfer function to the results of aging tests.

The transfer function is preferably a polynomial function.

[0026] In a variant implementation mode, the step of updating the parameters of the aging model comprises the following sub-steps: dividing an operation duration of the storage system into time slices; calculating for each time slice a state-of-health error at the end of the time-slice, said state-of-health error being equal to the absolute value of the difference between the first value of the state-of-health and a state value simulated health using the operation data on the time slice and the aging model to be updated; compare the state-of-health error of each time slice to a threshold value; and when the state of health error of at least one time slice is greater than the threshold value, the following sub-step: modifying the values of the parameters of the aging model until the error d health of each time slice is less than the threshold value.

The time slots can be of identical duration. Alternatively, the time slots are of variable duration and determined so that the operation data are substantially identical on the same time slot and different between the time slots.

Preferably, steps b) to e) are repeated several times when using the storage system.

To simulate the aging of the storage system, the predetermined electrical and thermal stress profiles are advantageously replaced by the electrical and thermal stress profiles under conditions of use as the operation data is collected. The estimation method according to the first aspect of the invention may also have one or more of the characteristics below, considered individually or according to all technically possible combinations: the storage system is stationary; the storage system is a battery storage system (BESS); the storage system includes a lithium battery.

A second aspect of the invention relates to a data processing device comprising means for implementing an estimation method according to the first aspect of the invention.

A third aspect of the invention relates to a computer program product comprising instructions which, when the program is executed by a computer, lead the latter to implement an estimation method according to the first aspect of the 'invention.

A fourth aspect of the invention relates to a computer-readable data carrier, on which the computer program product according to the third aspect of the invention is recorded.

BRIEF DESCRIPTION OF FIGURES

Other features and advantages of the invention will emerge clearly from the description given below, by way of indication and in no way limiting, with reference to the following figures.

[0035] [Fig. 1], previously described, schematically represents a method for estimating the lifetime of a storage system according to the prior art;

[0036] [Fig. 2] schematically represents the steps of a method for estimating the lifetime of a storage system by means of an aging model, according to a first embodiment of the invention;

[0037] [Fig. 3] schematically represents the steps of a method for estimating the lifetime of a storage system by means of an aging model, according to a second embodiment of the invention;

[0038] [Fig. 4] schematically represents a preferred mode of implementation of the step of recalibrating the aging model. [0039] [Fig. 5] schematically represents a sub-step of the recalibration of the aging model, comprising the division of an operation duration into time slices.

[0040] [Fig. 6] schematically represents another sub-step of the recalibration of the aging model, including the modification of aging test results.

[0041] [Fig. 7] schematically represents an operation of the test results modification sub-step according to figure 6.

[0042] For greater clarity, identical or similar elements are identified by identical reference signs in all of the figures.

DETAILED DESCRIPTION

Figure 2 is a block diagram (or block diagram) of a method for estimating the lifetime of an energy storage system, according to a first mode of implementation of the invention .

The energy storage system can be an on-board storage system or a stationary storage system. It comprises a plurality of electrochemical cells (also called electrochemical accumulators) connected in series and/or in parallel to form one or more batteries. The electrochemical cells are preferably of the same type, for example lithium accumulators (and belonging to the same variant of lithium accumulators: lithium-metal, lithium-ion, lithium polymer, lithium-ion-polymer, etc. ).

[0045] By way of example, the storage system can be used to supply energy to an electric or hybrid transport vehicle (on-board system), to absorb the surplus electricity production within a photovoltaic power plant (system stationary) or to supply an isolated site (stationary system).

The estimation method is particularly suitable for battery storage systems designated by the acronym “BESS” (for “Battery Energy Storage System”). A BESS is a large-scale stationary storage system generally coupled to a renewable energy production unit (solar, wind, etc.). The system includes several battery sections, each battery section comprising battery assemblies called "racks", each rack comprising a plurality of modules connected to each other, themselves made up of cells assembled in series and in parallel.

When in use, the storage system (whether stationary or onboard) is subject to the constraints of a so-called “terrain” environment and to conditions of use.

[0048] The constraints of the field environment are for example the presence of connection elements and protection elements between the electrochemical cells, between the different batteries of a rack, between the different racks of a section of batteries (in the case of a BESS) ..., the presence of a casing around the electrochemical cells of each battery, and the presence of electronic systems such as the BMS ("Battery Management System") and the EMS (“Energy Management System”). These constraints related to the actual use of the storage system have an impact on the performance of the system, particularly in terms of internal resistance (and therefore heating), capacity and lifespan.

The conditions of use are defined by various parameters such as the charging and discharging regimes to which the electrochemical cells are subjected, the cycling and rest phases of the electrochemical cells and the temperature surrounding the electrochemical cells. These conditions of use may be different between the batteries that make up the storage system, or even (in terms of temperature) between the electrochemical cells of the same battery.

The term “cycling phases” refers to the phases during which the electrochemical cells of the storage system undergo charge and discharge cycles, that is to say the phases during which the storage system is used. In this case, the storage system is said to be "in cycled mode".

[0051] The term “resting phases” refers to the phases during which the electrochemical cells of the storage system are not called upon. In this case, the storage system is said to be “in calendar mode”.

The storage system is advantageously equipped with at least one electronic management system, commonly called "BMS" for "Battery Management System" in English. The electronic management system makes it possible to control the operation of the storage system, in particular by regulating the load and discharge. A storage system can include several electronic management systems. For example, a BESS can have one BMS per module and one BMS per rack.

The electronic management system is also configured to collect storage system operation data and to calculate indicators, such as the state of health (SOH) and the state of charge (SOC) of the storage system. , from the operation data. The operating data are preferably current I(t), temperature (T(t)) and voltage (U(t)) values measured over time (t) at the batteries. The indicators are preferably calculated by a microprocessor belonging to the electronic management system.

The collected and calculated data are preferably stored in a database in communication with the microprocessor of the electronic management system. The database can be integrated into the electronic management system or, on the contrary, be separated from it.

The estimation of the lifetime of the storage system using the method according to the invention is based on the use of an empirical aging model. Advantageously, this aging model (also called “state of health evolution model”) takes into account both the calendar degradation and the cycling degradation of the storage system.

The following formulas constitute an example of an empirical aging model.

[0057] [Math. 1 ]

The state of health (SOH) is expressed as a percentage and corresponds to the ratio between the maximum discharge capacity Qmax and the initial capacity Qo of the storage system. The maximum discharge capacity Qmax represents the quantity of electrical energy that the storage system can supply when it is fully charged, at a given moment in the life of the system. The initial capacity Qo is the initial discharge capacity of the storage system, that is to say when the storage system is new. It can be measured or equated to the rated capacity specified by the storage system manufacturer. The closer the discharge capacity Qmax is to the initial capacity Qo of the storage system, the better its state of health. The state of health is therefore representative of an irreversible loss of autonomy of the storage system.

The state of health is linked to the losses of dQioss capacity in the following way:

[0060] [Math. 2]

The capacity losses dQioss are defined as the sum of the capacity losses at rest and those in cycling:

[0062] [Math. 3]

[0063] where: the term represents the capacity losses during the phases of rest (function of time t); the term represents the capacity losses during the phases cycling (depending on the quantity Qth of Ah accumulated during successive discharges).

An example of an equation for each of the preceding quantities is given by the following formulas:

[0065] [Math. 4]

[0066] [Math. 5]

[0067] where: Qioss is the capacity lost by the storage system, expressed in ampere-hours (Ah);

Jcai is a first factor for accelerating the aging of the storage system in calendar mode, called “rate of degradation in calendar mode” and expressed in Ah. sec ¹ ;

Jcyc is a first factor of acceleration of the aging of the storage system in cycled mode, called “rate of degradation in cycled mode” and expressed in Ah. Ah ^-1 ;

Acai is a second factor in accelerating the aging of the calendar mode storage system; and

Acyc is a second factor for accelerating storage system aging in cycled mode.

The Jcai and Acai parameters depend on the state of charge (SOC) and the temperature T at rest, while the Jcyc and Acyc parameters depend on the SOC and the current and temperature conditions during cycling, in other words discharge current ID, charge current le, discharge temperature TD and charge temperature Te.

This example of an aging model is described in detail in document [B. PILIPILI MATADI, Doctoral thesis from the University of Grenoble Alpes, "Study of the aging mechanisms of Li-ion batteries in low-temperature cycling and high-temperature storage: understanding of the origins and modeling of aging", 2017],

[0070] Referring to FIG. 2, the estimation method includes steps S21 to S25 leading to a projection of the lifetime of the storage system. A lifetime projection can be represented by a decrease in SOH or storage system capacity as a function of time in the future. Such a lifetime projection makes it possible to anticipate the end of life of the storage system, to program one or more storage system maintenance actions and to modify the management rules of the storage system in order to extend its lifetime. (and therefore improve its economic profitability). One or more SOH (or capacity) thresholds, called maintenance thresholds, can be defined to determine at which instants in the life of the system the maintenance actions will be performed. Similarly, an end-of-life threshold can be set to determine when the storage system should be replaced. A maintenance action or the replacement of the system is undertaken at the moment when the SOH (or the capacity) reaches the considered threshold, according to the lifetime projection.

Steps S21 to S25 are preferably coded in the form of a program and implemented by a computer (by executing the program).

The step S21 of the estimation method consists in defining the aging model which will be used to estimate the lifetime.

In a preferred mode of implementation, step S21 for defining the aging model comprises the selection of an aging model (whose parameters are unknown) and the determination of the parameters of the selected aging model (by example the terms Jcai, Jcyc, Acai and A _cyc in the previous model example) from test results.

The test results come from aging tests (also called "endurance tests") carried out in the laboratory (typically in a climatic chamber) on the scale of an electrochemical cell and/or a module comprising several interconnected electrochemical cells. These aging tests consist in aging, individually or in modules, a multitude of electrochemical cells of the same type as those of the storage system, at different charge and discharge currents and at different temperatures, the electrochemical cells having different target SOCs ( at the end of discharge and at the end of charge). Typically, each aging test corresponds to a charging current value Ie, a discharging current value ID, a temperature value T and a target SOC value.

These aging test results include data representative of the evolution of the SOH or of the capacity of the cell (or of the module, depending on the type of test) as a function of time. The parameters of the model can be determined for the first time in a conventional way, for example by statistical adjustment of the model to the test results.

In a variant implementation, the step S21 for defining the aging model consists of selecting a generic aging model that has already been configured.

[0079] In step S22, data is collected during the use of the storage system. In use, the electrochemical cells of the storage system are subject to conditions of use which can be very different from laboratory conditions. These conditions of use correspond to the actual operation of the storage system (placed in its field environment).

The data collected is the storage system operation data, namely the current I(t), the voltage U(t) and the temperature T(t), as well as the first values of the SOH and that of storage system SOC values. The first values of the SOH and the values of the SOC are so-called “field” values, since they are collected when the system is in use (in other words in its field environment). They are denoted respectively SOHBMs(t) and SOCBMs(t) in Figure 2.

The data collected is preferably that supplied by the electronic management system of the storage system (and recorded in the database).

The data is advantageously collected opportunistically during the actual operation of the storage system, during a charge or a discharge (partial or total). In other words, the data can be acquired when the conditions necessary for obtaining them are met. For example, a field value of SOH can be obtained when the storage system performs, during its use, a cycle comprising a full charge (SOC = 100%) followed by a full discharge (SOC = 0% ).

[0083] We also speak of "during operation" (or "on-line") collection for the SOH/SOC values and "continuous" collection for the operation data (current, voltage, temperature), the latter being collected at a frequency higher sampling, typically of the order of a second, a minute or an hour.

The step S23 of the estimation method is a recalibration step (also called “recalibration”) of the aging model. During this step S23, the parameters of the aging model are updated according to the field values of SOH (SOHBMS(I)), the field values of SOC (SOCBMS(I)) and the operation data (l(t) , U(t), T(t)) collected during step S22. The updated values of the model parameters are denoted Jcai', Jcyc', Acai' and Acyc' in Figure 2.

In a preferred embodiment of the recalibration step S23 described later in relation to FIG. 4, the updating of the parameters of the model can be accomplished by also considering the results of laboratory tests, noted SOH (le, ID, T, SOC) (le, ID, T and SOC being the laboratory conditions under which the tests are carried out).

[0086] The taking into account of the field data of the storage system, alone or in addition to the test results, in the determination of the parameters of the model makes it possible to have a predictive aging model specific to the storage system and representative of his actual behavior on the pitch. Indeed, the constraints of the field environment (for example related to the mode of connection of the cells, batteries, racks... and to the conditioning of the cells) are now taken into account in the determination of the parameters of the aging model.

The updating step S23 can be performed at each new data item collected during the collection step S22 (for example at each new value of SOH) or be repeated at a predetermined updating frequency, for example every months, quarters or semesters.

Thus, the aging model, previously static because established once and for all on the basis of an aging test campaign, becomes quasi-dynamic (or evolving).

The estimation method is remarkable in that the operation data are also used to determine, during a step S24, profiles (or scenarios) of electrical stress lop(t) and Uo _P (t ) (current and voltage demand respectively) and a thermal profile T _op (t) under conditions of use. The electrical stress profiles l _op (t) and U _op (t) consist of current or voltage values (charging and/or discharging) distributed over time, while the thermal profile Top(t) consists of temperature values distributed over time. All of these values belong to the operation data collected in step S22.

The step S24 for determining the profiles in conditions of use lop(t), Uop(t) and Top(t) preferably comprises operations of filtering (to discard the inconsistent values) and of concatenation of the data d 'operation.

Finally, in step S25, the aging of the storage system is simulated using the aging model using the parameters updated during step S23. This simulation is accomplished by feeding the model with predetermined electrical and thermal stress profiles lgen(t), Ugen(t) and T _ge n(t) on the one hand, and the profiles in use conditions lop(t) , Uop(t) and T _op (t) determined in step S24 on the other hand.

[0092] Unlike the predetermined electrical and thermal stress profiles lgen(t), Ugen(t) and Tgen(t), which are general simulation profiles consisting of average values of current, voltage and temperature respectively, the profiles in conditions of use lop(t), Uop(t) and T _op (t) represent a history of the use of the storage system.

[0093] The fact of feeding the aging model with these historical data makes the simulation of aging dynamic and more faithful to the behavior of the storage system observed in the field. The resulting lifetime projection is then more reliable and accurate.

The collection step S22, the recalibration step S23, the step S24 for determining the profiles under conditions of use and the simulation step S25 are advantageously repeated several times during the use of the system. of storage.

As indicated previously, data collection (step S22) can be performed continuously. The profiles in use conditions lop(t), Uop(t) and T _op (t) are preferably constructed (during step S24) and used in simulation (during step S25) as measurement of operation data collection. At the beginning of the life of the storage system, in the absence of operational data, an aging simulation can only be performed with the electrical and thermal stress profiles lgen(t), Ugen(t) and Tgen( t) predetermined. Then, these predetermined profiles are progressively replaced by the electrical and thermal stress profiles under conditions of use l _op (t), U _op (t) and T _op (t) (in other words, the operating data gradually replace the average values). The more operation data collected, the greater the usage history of the storage system and the greater the share of operation data in the simulation profiles.

[0096] Thus, the lifetime projection of the storage system becomes increasingly relevant as the data is collected.

FIG. 3 is a block diagram of a method for estimating the lifetime of an energy storage system, according to a second embodiment of the invention.

In addition to the steps S21 to S25 described above, the estimation method according to the second mode of implementation comprises: a step S31 of collecting test data, these test data coming from a characterization test the storage system; a step S32 for determining, from the test data, a second value of the SOH, called test and denoted SOHtest in FIG. 3; and a step S33 of correcting at least part of the field values of the state of health SOHBMs(t) as a function of the test value SOHtest.

The storage system characterization test, also called standardized test or "check-up", is a test carried out in the field which interrupts the use of the storage system. For this reason, it is carried out very punctually, for example every 3 months. The collection of test data therefore takes place at a much lower frequency than that of operation data. Its main purpose is to detect a drift in the measurement of transaction data by the electronic management system (“continuous” measurement during use). This test preferably includes a full charge followed by a full discharge. It can include up to 3 complete charge and discharge cycles, the values retained being those of the last cycle.

[00100] The test data collected during step S31 are of the same type as the operation data, namely values of the current I, of the voltage U and of the temperature T of the storage system (in charge and in discharge). They are denoted Itest, Utest and Ttest respectively.

[00101] The SOHtest test value of the state of health can be determined in step S32 by various calculation methods known to those skilled in the art and similar to those implemented in battery management systems (BMS ). By way of non-limiting examples, advanced calculation methods are described in documents EP3080624 and EP3080625. These methods are advantageous because they allow the phase-free SOH to be determined from full-charge and full-discharge cycles.

[00102] The correction step S33 can also be seen as a step for determining a so-called real state of health, denoted SOHreai(t), the values of which (field values, possibly corrected) are assumed to be exact. It preferably includes an interpolation operation (linear or quadratic) between the field values SOHBMs(t) and the test value SOHtest.

[00103] The test value SOHtest is a precise value of SOH. Correcting all or part of the SOHBMs(t) field values using the SOHtest test value improves the accuracy of the SOH values used to recalibrate the aging model. This correction compensates for the error in the SOH calculated by the electronic management system, due to measurement drift. The parameters of the aging model are then updated even more precisely and the reliability of the lifetime projection obtained by simulation is maximized.

[00104] FIG. 4 illustrates a preferred mode of implementation of the recalibration step S23 of the aging model.

[00105] The recalibration step S23 firstly comprises a sub-step S41 which can be broken down into two operations, performed successively or simultaneously: the division of an operation duration of the storage system into N time slots, N being a natural integer greater than or equal to 2; and determining average operating conditions over each time slice from the operating data (l(t), U(t) and T(t)) and the field values of SOC (SOCBMS(I)). [00106] The duration of operation corresponds to a period during which the storage system has been used normally and for which we have field values of SOH (SOHBMS(I)), possibly corrected (SOHréei(t)), values SOC terrain (SOCBMs(t)) and operation data (l(t), U(t), T(t)).

[00107] The average operating conditions on each slice include an average value of the charging current Icmay(i), an average value of the discharging current iDmoy(i), an average value of the temperature Tmoy(i) and a value d SOCavg(i) state of charge. It is also possible to distinguish between an average temperature value on charge and an average temperature value on discharge.

[00108] Advantageously, the time slices are of variable duration and determined so that the operation data and the field values of SOC are substantially identical on the same slice and different between the time slices. A homogeneity of the operating data, and in particular of the charging current (le), of the discharging current (ID) and of the temperature (T), and a homogeneity of the field values of SOC on each time slice are sought during the slicing.

[00109] In the following description, the different time slices are indexed by means of an index i, i being a natural integer varying from 1 to N. The instant t at which the slice of index i begins is denoted “ti -i >> and the instant t at which the slice of index i ends is denoted “ti”. Thus, the first slice (i = 1) starts at t = to and ends at t = ti.

In an exemplary implementation represented by FIG. 5, the sub-step S41 for cutting into time slices and determining the average operating conditions comprises the following operations:

551: calculate a center of gravity of the operation, C _gr (0), by averaging the operation data l(t), T(t) and the field values of the state of charge SOCBMs(t) over the entire duration of operation: Cgr(O) = {lcmavg(0), lDavg(0), Tavg(0) SOCavg(O)}.

552: define a minimum slice duration Ttr and a slice duration increment step Ptr; for each slice of index i (i varying from 1 to N): 553: initialize the slice end time ti as being the sum of the slice start time ti-i and the minimum slice duration Ttr (to being the start time of the operation duration);

554: check if the end time of slice ti is greater than or equal to the end time of the operation duration top; if this condition is met:

555: set the end of slice time ti equal to the end time of operation duration top (ti = t _op ) and terminate the slicing process; otherwise :

556: calculate a center of gravity of slice i, Cgr(i), by averaging the operation data l(t), T(t) and the field values of state of charge SOCBMs(t) over the entire duration of slice i, i.e. between ti-i and ti: Cgr(i) = {lcmavg(i), iDavg(i), Tavg(i) SOCavg(i)}.

557: calculating the distance between the centers of gravity C _gr (i) and Cgr(i-1 ) and comparing it to a predetermined minimum distance D _gr ; and, when the distance C _gr (i)-C ₈ r(i-1 ) is less than the minimum distance D _gr :

558: increment the slice end time ti by the incrementation step Ptr (ti = ti + Ptr); and looping back to verification step S54.

When the distance C _gr (i) and C _gr (i-1) is greater than or equal to the minimum distance D _gr , then the search for slice i is over.

[00112] The search for the next slice is started by looping back to step S53 after having incremented the slice index i (i=i+1).

[00113] In a variant implementation of sub-step S31, the duration of the operation is divided into time slots of identical duration, in particular to take into account a certain “seasonality” of the operation. For exemple, an operation duration of one year can be divided into 4 slices of 3 months, each slice corresponding to a season of the year.

[00114] Centered distributions of operating conditions can be determined for each time slice during sub-step S31. Each distribution is characterized by a standard deviation value associated with the average operating condition (lcmavg(i), iDavg(i), Tavg(i) or SOCavg(i) depending on the distribution).

[00115] Again with reference to FIG. 4, the recalibration step S23 then comprises a sub-step S42 consisting in calculating, for each time slice, a state of health error Eson(ti) at the end of the time slice. . The state of health error EsoH(ti) is equal to the absolute value of the difference between the field value of SOH at the time of the end of the slice ti (SOHBMs(ti) or SOHréei(ti) according to the mode of implementation; see Figs.2-3) and a simulated state of health value SOHsim(ti) at the time of end of slice ti.

[00116] [Math. 6]

Eson(ti) ⁼ \ OH _BMS (ti) — SOH _sim (ti \

[00117] The simulated state of health value SOHsim(ti) is obtained by simulation using the operation data on the time slice and the aging model that it is sought to recalibrate (represented by the parameters Jcai, Jcyc , Acai and Acyc when it is the first iteration of the recalibration step S23, otherwise by the parameters Jcai', Jcyc', Acai' and Acyc' previously updated).

[00118] Then, during a sub-step S43, the state of health error Eson(ti) of each time slot is compared with a predetermined threshold value Eth, called “acceptable error of SOH”.

[00119] When the state of health error Eson(ti) of all the time slots is less than or equal to the threshold value Eth, it is considered that the aging simulation is sufficiently close to the aging observed in the field during the duration of operation and therefore that it is not useful to update the parameters of the model. The step S32 for recalibrating the aging model then ends without having modified the model parameter values. [00120] On the other hand, when the state of health error Eson(ti) of one or more time slots is greater than the threshold value Eth, the model parameter values must be modified.

[00121] The recalibration step S23 then comprises (cf. FIG. 4): a sub-step S44 for modifying the SOH aging test results (le, ID, T, SOC), according to the field values d the state of health SOHBMs(t)/SOHréei(ti) and average operating conditions over the time slices; and a so-called sub-step S45 of relearning the parameters of the model, consisting in determining new values of parameters from the results of modified aging tests SOH′ (le, ID, T, SOC).

[00122] FIG. 6 represents an example of implementation of sub-step S44 for modifying the results of SOH aging tests (le, ID, T, SOC).

[00123] During a first operation S61 , so-called neighboring test results are identified for each time slot among all the aging test results. The results of tests close to slice i, indexed by the index j ranging from 1 to 16 and denoted SOH[i,j] (le, ID, T, SOC), are results of tests carried out under conditions of tests (charging current, discharging current, temperature and SOC) close to the average operating conditions of the considered time slot.

[00124] Slot i is characterized by an average value of the charging current Icmay(i), an average value of the discharging current iDmoy(i), an average value of the temperature Tmoy(i) and an average value of the state of SOCavg(i) load. For each of these 4 average operating conditions, the 2 test results whose test conditions are closest are identified. 16 test results are then obtained (ie revolution of SOH for 16 sets of test conditions) which frame the center of gravity of the wafer C _gr (i).

[00125] Then, during an operation S62, a time window is identified for each of the neighboring test results SOH[i,j] having a duration Di equal to that of the time slice (of index i) and starting at a time td [i, j] at which the test result reaches the field value SOHBMs(ti-i) (or SOHréei(ti-i)), i.e. the field value of SOH at the start time ti-i of the time slice. [00126] Figure 7 shows two examples of test results close to the slice of index i, SOH[i,1] (le, ID, T = 20°C, SOC) and SOH[i,2] ( le, ID, T = 30°C, SOC), corresponding to two values (called "limits") close to the same average operating condition, here the temperature with T = 20°C and T= 30°C ( Tmean(i) being for example 25°C). The time window of the neighboring test SOH[i,1 ] (le, ID, T = 20°C, SOC) starts at time td [i, 1 ] (where SOH[i,1 ] = SOHBMs(ti -i)) and ends at time tf[i, 1]. The time window of the neighboring test SOH[i,2] (le, ID, T = 30°C, SOC) starts at time td[i, 2] (where SOH[i,2] = SOHBMs(ti -i)) and ends at time tf[i, 2].

Then, during an operation S63, for each of the neighboring test results SOH[i,j] (le, ID, T, SOC), the state of health value SOH[i,j] is determined. (tf [i,j]) reached at the end of the time window.

At the end of this operation S63, 16 SOH values are available for each time slice, denoted SOH[i,j] (tf[i,j]) and called SOH test values (because drawn neighboring test results).

[00129] The different SOH test values of each time slot are then weighted during an operation S64 to obtain a single SOH value at the end of the time slot, this value being qualified as experimental and denoted SOHexp(ti) . This experimental value of SOH can be compared to the field value of SOH at the end of the time slice, SOHBMS (ti), in order to determine to what extent the results of neighboring tests are representative of the aging actually observed in the field.

[00130] To perform the weighting operation S64 of the SOH test values on each time slot, coefficients are preferably applied which, when applied to neighboring test conditions, make it possible to obtain the average conditions operation of the time slice.

During the following operation S65, a transfer function is calculated between the N experimental values of state of health SOHexp(ti) and the N field values of state of health SOHBMs(ti). The transfer function is preferably a polynomial function.

Finally, during a last operation S66, the same transfer function is applied to the initial SOH (le, ID, T, SOC) test results, thus obtaining the modified test results SOH' (le, ID, T, SOC) more in line with the behavior of the storage system during the operation period.

[00133] Again with reference to FIG. 4, the sub-step S45 of relearning the parameters of the model, from the modified test results SOH' (le, ID, T, SOC), can be accomplished in the same way. way as the initial determination of the parameters of the model from the results of initial tests SOH (le, ID, T, SOC) (step S21), for example by statistical adjustment.

[00134] The preferred mode of implementation of the recalibration step S23 represented by FIG. 4 can be implemented easily and executed quickly. It significantly improves the accuracy of the aging model and the resulting lifetime projections. In particular, the division of the duration of operation into time slices increases the probability of triggering a modification of the parameters of the model.

As a variant of steps S44 and S45 (not represented by the figures), the values of the parameters of the model (eg Jcai, Jcyc, Acai and A _C yc) can be modified so that the error of state of health Eson(ti) of each time slice is less than the threshold value Eth. This variant is equivalent to solving an optimization problem. Optimization software, implementing non-linear mathematical methods, can be used for this purpose.

[00136] In this variant, no use is made of average operating conditions. These therefore do not need to be determined during sub-step S31 (in particular in the case of a division into slices of identical duration).

[00137] An additional objective of the optimization may be to minimize the variation of the parameters of the model compared to the previous values.

Claims

[Claim 1] A method for estimating a lifetime of an energy storage system comprising a plurality of electrochemical cells, said method comprising the following steps: a) defining (S21) an aging model of the storage system; b) collecting (S22) first values of a state of health (SOHBMS(I)), values of a state of charge (SOCBMS(I)) and operation data of the storage system during a use of the storage system, the storage system being subject to conditions of use, the operation data comprising values of current (l(t)), temperature (T(t)) and voltage (U( t)); c) updating (S23) parameters of the aging model according to the first values of the state of health, the state of charge values (SOCBMs(t)) and the operation data; d) determining (S24) a thermal profile (T _op (t)) and electrical stress profiles (l _op (t), U _op (t)) under conditions of use from the operation data; e) simulating (S25) the aging of the storage system using the updated parameters of the aging model, the predetermined thermal and electrical stress profiles (lgen(t), Ugen(t), T _ge n(t)), and the thermal and electrical stress profiles under conditions of use (lop(t), Uop(t), Top(t)), from which results a projection of the service life of the storage system.

[Claim 2] Method according to claim 1, further comprising, before the step (S23) of updating the parameters of the aging model, the following steps:

- collecting (S31) test data from a characterization test of the storage system;

- determining (S32) a second value of the state of health (SOHtest) of the storage system from the test data; and correcting (S33) the first values of the state of health (SOHBMS(I)) according to the second value of state of health (SOHtest).

[Claim 3] Method according to one of Claims 1 and 2, in which the step (S21 ) of defining the aging model comprises the selection of an aging model and the determination of the parameters of the selected aging model from aging test results (SOH (le, ID, T, SOC)).

[Claim 4] A method according to claim 3, wherein the aging test results (SOH (le, ID, T, SOC)) come from aging tests carried out at the scale of an electrochemical cell and/or aging tests carried out on the scale of a module comprising several interconnected electrochemical cells.

[Claim 5] Method according to any one of claims 1 to 4, in which the step (S23) of updating the parameters of the aging model comprises the following sub-steps:

- dividing (S41) an operating duration of the storage system into time slices and determining average operating conditions (lcmavg(i), iDavg(i), Tavg(i), SOCavg(i)) on each time slice from operation data and state of charge values (SOCBMS(I));

- calculating (S42) for each time slice a state of health error (EsoH(ti)) at the end of the time slice, said state of health error being equal to the absolute value of the difference between the first value of the health status (SOHBMs(ti)) and a health status value (SOHsim(ti)) simulated using the operation data on the time slice and the aging model to be updated;

- comparing (S43) the state of health error (Eson(ti)) of each time slice with a threshold value (Eth); and when the state-of-health error of at least one time slice is greater than the threshold value, the following substeps:

- modifying (S44) results of aging tests (SOH (le, ID, T, SOC)) with regard to the first state-of-health values (SOHBMS(I)) and the operation data on the time slots ; and - determining (S45) values of parameters of the aging model from the results of modified aging tests (SOH' (le, ID, T, SOC)).

[Claim 6] A method according to claim 5, wherein the substep (S44) of modifying the aging test results (SOH (le, ID, T, SOC)) comprises the following operations:

- for each time slot, identifying (S61 ) among the results of aging tests results of so-called neighboring tests (SOH[i,j]) performed at test conditions close to the average operating conditions (lcmay( i), iDavg(i), Tavg(i), SOCavg(i)) of said time slice;

- identifying (S62) for each of the neighboring test results (SOH[i,j]) a time window having a duration (Di) equal to that of the time slice and starting at a time (td [i,j]) at which the test result reaches the first health status value (SOHBMs(ti-i)) at the start of the time slice;

- determining (S63) for each of the neighboring test results (SOH[i,j]) a state of health value (SOH[i,j] (tf[i,j])) at the end of the window temporal;

- weighting (S64) for each time slice the state of health values (SOH[i,j] (tf [i,j])) at the end of the time windows to obtain an experimental state of health value (SOHexp (ti)) at the end of the time slice;

- calculating (S65) a transfer function between the experimental state of health values (SOHexp(ti)) at the end of the time slice and the first state of health values (SOHBMs(ti)) at the end of the time slice;

- applying (S66) the transfer function to the results of aging tests (SOH (le, ID, T, SOC)).

[Claim 7] A method according to claim 6, wherein the transfer function is a polynomial function.

[Claim 8] Method according to any one of claims 1 to 4, in which the step (S23) of updating the parameters of the aging model comprises the following sub-steps:

- dividing (S41) an operation duration of the storage system into time slices;

- calculating (S42) for each time slice a state of health error (EsoH(ti)) at the end of the time slice, said state of health error being equal to the absolute value of the difference between the first value of the the state of health (SOHréei(ti)) and a state of health value (SOHsim(ti)) simulated using the operation data on the time slice and the aging model to be updated;

- comparing (S43) the state of health error (Eson(ti)) of each time slice with a threshold value (Eth); and when the state-of-health error of at least one time slice is greater than the threshold value, the following sub-step:

- modify the values of the parameters of the aging model until the state-of-health error (Eson(ti)) of each time slice is lower than the threshold value (Eth).

[Claim 9] A method according to any of claims 5 to 8, wherein the time slots are of identical duration.

[Claim 10] A method according to any one of claims 5 to 8, wherein the time slots are of variable duration and determined such that the operation data is substantially identical within a same time slot and different between time slots.

[Claim 1 1 ] A method according to any one of claims 1 to 10, wherein steps b) to e) are repeated several times during use of the storage system.

[Claim 12] Method according to claim 1 1 , in which, to simulate the aging of the storage system, the predetermined electrical and thermal stress profiles (lgen(t), Ugen(t), T _ge n(t)) are replaced by the profiles of electrical and thermal stress under conditions of use (lop(t), Uo _P (t), T _op (t)) as the operation data is collected.

[Claim 13] A method according to any of claims 1 to 12, wherein the storage system is stationary, preferably a battery storage system (BESS).

[Claim 14] A method according to any of claims 1 to 13, wherein the storage system includes a lithium battery.

[Claim 15] Data processing device comprising means for implementing an estimation method according to any one of Claims 1 to 14.

[Claim 16] Computer program product comprising instructions which, when the program is executed by a computer, cause the latter to carry out a method according to any one of claims 1 to 14.

[Claim 17] A computer-readable data carrier on which the computer program product according to claim 16 is recorded.